a review and meta-analysis of the effect of weight loss on ......weight loss: all-cause mortality:...
TRANSCRIPT
A review and meta-analysis of the effect of weight loss on all-causemortality risk
Mary Harrington1 Sigrid Gibson2 and Richard C Cottrell31The Sugar Bureau London WC2B 5JJ UK
2Sig-Nurture Ltd Guildford Surrey GU1 2TF UK3World Sugar Research Organisation London SW1V 3LX UK
Overweight and obesity are associated with increased morbidity and mortality although the rangeof body weights that is optimal for health is controversial It is less clear whether weight lossbenefits longevity and hence whether weight reduction is justified as a prime goal for allindividuals who are overweight (normally defined as BMI 25 kgm2) The purpose of thepresent review was to examine the evidence base for recommending weight loss by diet andlifestyle change as a means of prolonging life An electronic search identified twenty-six eligibleprospective studies that monitored subsequent mortality risk following weight loss by lifestylechange published up to 2008 Data were extracted and further analysed by meta-analysis givingparticular attention to the influence of confounders Moderator variables such as reason forweight loss (intentional unintentional) baseline health status (healthy unhealthy) baseline BMI(normal overweight obese) method used to estimate weight loss (measured weight lossreported weight loss) and whether models adjusted for physical activity (adjusted dataunadjusted data) were used to classify subgroups for separate analysis Intentional weight loss perse had a neutral effect on all-cause mortality (relative risk (RR) 1middot01 P frac14 0middot89) while weight losswhich was unintentional or ill-defined was associated with excess risk of 22 to 39 Intentionalweight loss had a small benefit for individuals classified as unhealthy (with obesity-related riskfactors) (RR 0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028) especially unhealthy obese (RR 0middot84 (95 CI0middot73 0middot97) P frac14 0middot018) but appeared to be associated with slightly increased mortality forhealthy individuals (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) and for those who were overweightbut not obese (RR 1middot09 (95 CI 1middot02 1middot17) P frac14 0middot008) There was no evidence for weight lossconferring either benefit or risk among healthy obese In conclusion the available evidence doesnot support solely advising overweight or obese individuals who are otherwise healthy to loseweight as a means of prolonging life Other aspects of a healthy lifestyle especially exercise anddietary quality should be considered However well-designed intervention studies are neededclearly to disentangle the influence of physical activity diet strategy and body composition inorder to define appropriate advice to those populations that might be expected to benefit
Weight loss All-cause mortality Meta-analyses
Introduction
Weight loss has been reported to result in several healthbenefits such as significant improvements in CVD riskfactors (blood pressure lipid profiles glucose toler-ance)(12) It may therefore be reasonable to expect thatweight loss would lead to decreased mortality in the longterm Indeed this seems to be the case in obese individualswith serious medical complications(34) or when substantialweight loss has followed surgical procedures(5) Howeverthe long-term effects of more moderate degrees of weight
loss for those who are not severely obese and do not haveco-morbidities are unclear Many prospective studies showconflicting results while some recent studies indicateeither excess(6 ndash 9) or unchanged mortality(10) followingweight loss Reviews of the data suggest that inconsistentresults might be due to failure to control for knownconfounding factors (for example underlying diseaseintention to lose weight)(1112) while also noting that manyof the existing studies were not specifically designed to testthe hypothesis that weight loss increases or decreases
Abbreviation RR relative risk
Corresponding author Dr Richard C Cottrell fax thorn 1 20 7834 4137 email rcottrellwsroorg
Nutrition Research Reviews (2009) 22 93ndash108
q The Authors 2009
doi101017S0954422409990035
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Table 1 Prospective studies of weight loss and mortality
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Harris et al(1988)(38)
Apparentlyhealthy menand women
0ndash9 BMIgain
Men weightloss intentionunspecified
ndash 55ndash65 ndash $10 BMI Baseline weight 1middot9 1middot1 3middot2 Smokersex-smokers
FraminghamHeart Study
0ndash9 BMI 1middot4 1middot0 1middot9
Change inBMI from age55 to 65 years
Women weightloss intentionunspecified
ndash 55ndash65 ndash $10 BMI 1middot8 1middot2 2middot60ndash9 BMI 1middot1 0middot8 1middot4
Pamuk et al(1992)(39)
Apparentlyhealthy menand women
5 maxi-mum
weight loss
Men weightloss intentionunspecified
68 45ndash74 26ndash 29kgm2
6middot5 weight Age race smokingparity pre-existingillness maximumBMI
1middot7 1middot1 2middot7 Deaths infirst 5 years
NHANES I 279 26 kgm2 9middot4 weight 1middot4 1middot0 1middot897 26 kgm2 20middot3 weight 2middot4 1middot7 3middot5172 $29 kgm2 8middot8 weight 0middot7 0middot5 1middot0204 26ndash 29
kgm29middot7 weight 0middot8 0middot6 1middot1
34 $29 kgm2 18middot1 weight 0middot8 0middot4 1middot568 26ndash 29
kgm220 weight 1middot1 0middot7 1middot6
105 26 kgm2 23middot2 weight 1middot4 1middot0 2middot0Women weightloss intentionunspecified
38 45ndash74 26ndash 29kgm2
6middot6 weight 1middot4 0middot6 2middot9
223 26 kgm2 10middot2 weight 1middot6 1middot0 2middot6128 26 kgm2 20middot3 weight 2middot2 1middot3 3middot6230 $29 kgm2 9middot3 weight 1middot0 0middot7 1middot4168 26ndash 29
kgm29middot8 weight 1middot5 1middot0 2middot1
97 $29 kgm2 20middot9 weight 1middot4 0middot9 2middot180 26ndash 29
kgm220middot9 weight 1middot4 0middot9 2middot1
151 26 kgm2 25middot5 weight 1middot9 1middot3 2middot6Lee ampPaffenbar-
ger(1992)(1)
Apparentlyhealthy men
Weightstable
Men weightloss intentionunspecified
1293 58 78 kg 5 kg weight Age heightsmoking physicalactivity
1middot57 1middot341middot84
Baseline CVDcancer
Harvard AlumniHealth Study
2730 58 78 kg 1ndash5 kgweight
1middot26 1middot101middot46
Higgins et al(1993)(40)
Apparentlyhealthy menand women
No changein BMI
Men ndash 45middot5 27middot1 kgm2 0middot52 kg Age BMI systolicblood pressurecholesterolglucose intoleranceleft ventricularhypertrophysmoking
1middot33 1middot061middot68
Deaths infirst 4 years
FraminghamStudy
Women ndash 45middot8 25middot7 kgm2 0middot39 kg 1middot28 0middot981middot68
Chaturvedi ampFuller(1995)(41)
NIDDM menand women
Weight stable European menand women
252 (52deaths)
35ndash55 26 kgm2 2 kgm2 Age sex duration ofdiagnosed diabetes
3middot05 1middot267middot36
ndash
WHO study Weight lossintentionunspecified
35ndash55 26ndash29kgm2
2 kgm2 2middot02 1middot004middot08
35ndash55 $29 kgm2 2 kgm2 0middot84 0middot401middot74
M
Harrin
gto
net
al
94
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Iribarrenetal(1995)(42)
Apparentlyhealthy men
Loss of 2middot5 kgto gainof 2middot4 kg
Men Weightloss intentionunspecified
744 54middot6 25middot2 kgm2 4middot5 kg Age average weightsmoking alcoholconsumptionphysical activitytotal energy intakeemploymentpre-existingillness
1middot21 1middot021middot43
Deaths in first5 years
Honolulu HeartProgram
928 54middot2 24middot3 kgm2 2middot6ndash4middot5kg
1middot29 1middot101middot51
Manson et al(1995)(43)
Apparentlyhealthywomen
Weight stable(4 kg) sinceage 18 years
Women Weightloss intentionunspecified
Samplesize notreported
Age BMI smoking Baseline CVDcancer deathsin first 4 years
Nursesrsquo HealthStudy
16 deaths 30ndash55 ndash $10 kg 0middot7 0middot4 1middot454 deaths 30ndash55 ndash 4ndash9 kg 1middot2 0middot9 1middot6
Wallace et al(1995)(44)
Unhealthymen
Non-weightlosers
Unintentional 54 (175non-weight
losers)
73middot9 80middot8 kg 5middot6 kg Age BMI tobaccouse
hypertension healthstatus cholesterolalbumin levels
2middot83 1middot385middot81
Diseases thataffect nutritionalstatus orbody weight
Williamsonet al(1995)(45)
Healthy andunhealthywomen
No weightchange
Healthy Age baseline BMIeducation alcoholintake physicalactivity healthconditions
Deaths in first3 years
AmericanCancerPreventionStudy
Unintentional 942 52middot9 30middot9 kgm2 4middot9 kgm2 1middot20 0middot931middot55
Intentional1ndash19 lbs
2745 51middot7 30middot4 kgm2 3middot1 kgm2 1middot12 0middot941middot33
Intentional$20 lbs
3018 50middot8 33middot1 kgm2 6middot5 kgm2 0middot98 0middot821middot17
UnhealthyUnintentional 812 55middot3 31middot9 kgm2 5middot6 kgm2 1middot00 0middot83
1middot20Intentional
1ndash19 lbs1550 53middot8 31middot5 kgm2 3middot0 kgm2 0middot80 0middot68
0middot94Intentional
$20 lbs2598 53middot7 34middot8 kgm2 7middot0 kgm2 0middot81 0middot71
0middot92Yaari ampGoldbourt(1998)(46)
Unhealthymen
IsraeliIschemicHeartDiseaseStudy
Weight stable Weight lossintentional(dieters)
2471 Notgiven
78middot2 kg $ 5 kg5 kg
Age BMI systolicblood
pressure cholesterolsmoking diabetescancer history ofmyocardialinfarctionangina chroniclung diseasebaseline dietingperipheralartery disease
1middot301middot3
1middot021middot65
1middot021middot65
ndash
Weig
ht
loss
and
all-cause
mo
rtality9
5
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
French et al(1999)(47)
Apparentlyhealthywomen
Never $20 lbsweight loss
Intentional 4300 66middot6 30middot3 kgm2 $9middot1 kg Age BMI waisthipratio educationmarital statussmoking oestrogenuse cancer diabetesangina stroke heartattack hypertension
1middot18 0middot941middot48
ndash
Iowa WomenrsquosHealth Study
Unintentional 5008 68middot0 26middot1 kgm2 $9middot1 kg 1middot33 1middot131middot57
Williamsonet al
(1999)(48)
Healthy andunhealthy
men
No weightchange
Healthy Age BMI smokingeducation alcoholintake physicalactivity healthcomplaints
BMI 27 kgm2non-Caucasian
AmericanCancerPreventionStudy
Unintentional 1474 52middot0 29middot2 kgm2 3middot2 kgm2 1middot04 0middot911middot19
Intentional1ndash19 lbs
2834 51middot5 29middot0 kgm2 1middot8 kgm2 1middot09 0middot981middot21
Intentional$20 lbs
2610 51middot5 31middot4 kgm2 4middot5 kgm2 1middot07 0middot961middot20
UnhealthyUnintentional 917 54middot4 29middot7 kgm2 4middot2 kgm2 1middot15 1middot04
1middot27Intentional
1ndash19 lbs1310 53middot4 29middot1 kgm2 1middot9 kgm2 1middot01 0middot91
1middot12Intentional
$20 lbs2614 53middot6 31middot6 kgm2 4middot9 kgm2 1middot02 0middot94
1middot11Williamson
et al(2000)(4)
Unhealthymen andwomen
No or unknownweight change
Intentional 1669 54middot5 33middot5 kgm2 5middot8 kgm2 Age sex BMI racesmoking educationalcohol intakephysical activitydisease history
0middot75 0middot670middot84
BMI 27 kgm2
AmericanCancerPreventionStudy
Unintentional 649 55middot6 31middot8 kgm2 5middot9 kgm2 0middot98 0middot851middot13
Newmanet al
(2001)(49)
Older menand women
Weight stableie weightwithin ^5 of baseline
Weight lossintentionunspecified
126 deaths 77middot4 27 kgm2 $5 weight
Age sex racecognitive functionmedication smokingwaist circumferencemobility impairment
1middot67 1middot292middot15
Living in aninstitutionwheelchair usecancer treat-
ment62 deaths 77middot4 $5
weight1middot66 1middot18
2middot33Plus interimillness
Wanna-methee
et al(2002)(50)
Apparentlyhealthy men
Weight stable Weight lossintentionunspecified
950 40ndash59years
26middot6kgm2
2middot11kgm2
Age social classsmoking physicalactivity BMI CVDcancer poorhealth diabetes
1middot34 1middot091middot63
ndash
British RegionalHeart Study
M
Harrin
gto
net
al
96
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Gregg et al(2003)(31)
Overweightor obesemen andwomen
Weight stable Overall(unspecified)
1931 Over 35years(mean54middot1)
30middot8kgm2
7 kg Age sex racesmoking educationBMI self-ratedhealth diabetesacute and chronicconditions functionallimitations dueto CVD or cancerhospital bed days
1middot09 0middot901middot32
BaselineBMI 25
kgm2
NHIS UScohort
Unintentional 188 ndash ndash 6middot9 kg 1middot31 1middot011middot70
Retrospectiveweight
change9-yearfollow-up
Intentional 827 ndash ndash 7middot1 kg 0middot76 0middot600middot97
Gregg et al(2004)(51)
Unhealthy menand
women(diabetics)from NHIS US
cohort(9-yearfollow-up)
Weight stable Overall(unspecified)
629 ndash 33middot0kgm2
6middot80 kg Age sex race BMIsmoking educationself-rated healthdiabetes medicationlength of diseasefunctional limitationhypertension strokeheart disease retinaldisease neuropathyhospital days doctorvisits
1middot19 0middot9 1middot47 BaselineBMI 25
kgm2
Unintentional 365 ndash ndash 6middot80 kg 1middot58 1middot082middot31
Intentional 34 ndash ndash 6middot80 kg 0middot83 0middot631middot08
Maru et al(2004)(52)
Healthy women Weight stableie 5
weightchange
Moderate weightloss
531 50ndash66years
MedianBMI25middot4 kgm2
5ndash9 weight
Age smoking BMI 1middot14 1middot1 1middot6 Medication forhypertensionCVD diabetesrestriction diet
DOM Dutchcohort
Severe weightloss
108 IQR 23middot3ndash 27middot8kgm2
10ndash14 weight
0middot9 0middot5 1middot4
Weight changein
1-year follow-up
Extreme weightloss
43 $15 weight 0middot8 0middot4 1middot8
Median follow-up
17 yearsDiaz et al(2005)(6)
Apparentlyhealthy
men andwomen
Weight stable Weight lossintentionunspecified
711 51middot6 30middot8 kgm2 5middot55 kgm2 Age sex race BMIsmoking healthstatus poor healthincapacitated
3middot36 2middot474middot55
Diabetes CVDcancer
NHANES I andfollow-up
Droslashyvoldet al
(2005)(7)
Apparentlyhealthy
men andwomen
Weight stableie change inBMI 0middot1per year
Weight lossintentionunspecified
Age BMI systolicblood pressureblood pressuremedication smokingalcohol intakephysicalactivity maritalstatuseducation
CVD diabetescancer
Nord-Troslashnde-lag Health
Study
Men 1319 54middot3 26middot9 kgm2 2middot2 kgm2 1middot6 1middot4 1middot8Women 1971 54middot0 27middot7 kgm2 2middot7 kgm2 1middot7 1middot5 2middot0
Weig
ht
loss
and
all-cause
mo
rtality9
7
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Elliott et al(2005)(53)
Apparentlyhealthy
women
Weight changefrom 21middot81to thorn1middot36 kg
Weight lossintentionunspecified
964 42ndash81years
ndash 116middot58ndash1middot81 kg
Social class BMIparity smokinghormonereplacementtherapy
0middot96 0middot651middot43
ndash
Oral Contra-ception
StudySoslashrensen
et al(2005)(8)
Apparentlyhealthy
men andwomen
Weight stable Intentional 398 41middot5 27middot4 kgm2 1middot21 kgm2 Age sex BMIhypertensionsmoking alcoholphysical activitylife satisfactionwork status drugs
1middot87 1middot222middot87
Angina myocar-dial
infarctiondiabetes CVDlung diseasehypertensionprescription
drugsunemployment
Finnish TwinCohort
Unintentional 728 42middot6 26middot72 kgm2 1middot09 kgm2 1middot17 0middot821middot66
Wanna-methee
et al(2005)(10)
Apparentlyhealthy
men
No weightchange
Unintentional 527 40ndash59 25middot6 kgm2 3middot91 kgm2 Age smoking socialclass physicalactivityalcohol intakeobesityperceived healthstatus CVD cancerhypertension stroke
1middot71 1middot332middot19
ndash
British RegionalHeart Study
Intentional 342 40ndash59 28middot0 kgm2 2middot37 kgm2 1middot00 0middot911middot10
Intentionalpersonalreason
178 40ndash59 26middot9 kgm2 2middot31 kgm2 0middot59 0middot341middot00
Intentionalphysicianrsquosadvice
164 40ndash59 28middot5 kgm2 2middot44 kgm2 1middot37 0middot961middot94
Breeze et al(2006)(54)
Apparentlyhealthy
men
Minimal weightchange ie
loss0ndash3 kg orgain 0ndash3 kg
Weight lossintentionunspecified
554 40ndash69 ndash $10 kg Age marital statusemploymentsmoking respiratorysymptoms heartdisease indicatorsdiastolic bloodpressure totalcholesterol
1middot88 1middot6 2middot2 ndash
WhitehallCohort
1190 40ndash69 ndash 4ndash9 kg 1middot26 1middot1 1middot5
Nilsson et al(2002)(55)
HealthySwedishmen (n 5194)
Weight stable(^0middot1 kgm2)
Weight lossunspecified(no directquestion)
464 47 years(38ndash52years)
22ndash25kgm2
ndash Age 1middot39 0middot981middot95
Cancer deathsdisease atbaseline
deathsin year 1 offollow-up
482 26 thornkgm2
1middot71 1middot182middot47
M
Harrin
gto
net
al
98
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
relative risk (RR) of all-cause mortality(1314) Methodo-logical problems have also been identified for example themethod by which the weight loss was achieved has usuallynot been reported (although dietary energy restriction islikely to have been a major factor) while weight changesbefore and after the recording periods have usually not beendetermined(12 ndash 16)
In light of the current obesity epidemic and the resultingfocus on encouraging those with BMI above 25 kgm2 tolose weight by changing their diet and lifestyle(17) it isimportant to establish whether the long-term effects ofweight loss benefit life expectancy The current advice fromthe UK Department of Healthrsquos Obesity Care Programme isfor those who are overweight or obese to reduce energyintake and increase physical activity as a method of lifestylemodification(18) Further treatment and advice may need toconsider a broad spectrum of evidence so as not to rule outpotential investigations that identify subgroups of patientsor certain conditions where weight loss may be detrimentalto health and increase mortality(19)
The aim of the present study was to examine the availableevidence of the impact of weight loss as a lifestyleintervention on the RR of all-cause mortality and toquantify this using meta-analysis Data were pooled in anumber of different ways in order to examine the influenceof a number of possible confounders Meta-analysis wasused to provide a more objective appraisal of the evidenceintegrating data from multiple prospective cohort studies toincrease the power and precision of estimates of effect andreducing the likelihood of false negative results(2021)
Methods
Search strategy
A literature search was carried out independently by twoinvestigators to identify prospective cohort studies thatevaluated the effect of weight loss as a lifestyle interventionon mortality risk A web search was undertaken onPubMedMedline and ScienceDirect databases Articlespublished between 1987 and 2008 and in the Englishlanguage were included Search terms included lsquoweightBMI loss change mortality intentional unintentionalrelative risk prospective and cohortrsquo Identified citationsand abstracts were obtained from journals libraries orauthors A hand-search of the bibliographies of retrievedpapers and linked articles was also carried out
Data selection
Inclusion criteria were prospective studies in English ofadults (men andor women) with data on body weight andweight loss over more than 1 year Studies needed to presentRR of mortality and associated 95 CI for the group thatlost weight relative to a comparable reference group wholost minimal or no weight Drug treatment studies andstudies that measured weight loss following bariatricsurgery were excluded as the aim was to assess the effectof lifestyle interventions Twenty-six publications wereidentified that met the inclusion criteria Data on RR ofmortality and 95 confidence limits were extracted for allW
edic
ketal
(2002)
(56)
Calif
orn
ia
US
AW
eig
ht
sta
ble
(loss
10
lbs
or
gain
)
Weig
ht
loss
unspecifi
ed
71
years
atsta
rtof
12-y
ears
mort
alit
yfo
llow
-up
10
lbs
Age
curr
ent
and
past
sm
okin
g
exerc
ise
less
than
10
years
earlie
r
ndash
Healthy
men
628
26
1middot3
81middot0
61middot8
Healthy
wom
en
933
24
1middot7
61
middot33
2middot3
4D
iabetic
men
140
26
3middot6
62
middot15
6middot2
4D
iabetic
wom
en
90
25
1middot6
50middot7
3middot8
7In
tern
ational
weig
ht
loss
Tota
l
642
Healthy
men
1middot2
70
middot94
1middot7
1H
ealthy
Wom
en
1middot2
80middot9
5
1middot7
3
NH
AN
ES
N
ationalH
ealth
and
Nutr
itio
nE
xam
ination
Surv
ey
NID
DM
non-insulin
-dependent
dia
bete
sm
elli
tus
NH
IS
NationalH
ealth
Inte
rvie
wS
urv
ey
DO
M
Dia
gnostisch
Onderz
oek
Mam
macarc
inoom
(Dia
gnostic
Investigation
into
Bre
ast
Cancer)
IQ
R
inte
rquart
ilera
nge
1
lbfrac14
0middot4
536
kg
Weight loss and all-cause mortality 99
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Table 1 Prospective studies of weight loss and mortality
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Harris et al(1988)(38)
Apparentlyhealthy menand women
0ndash9 BMIgain
Men weightloss intentionunspecified
ndash 55ndash65 ndash $10 BMI Baseline weight 1middot9 1middot1 3middot2 Smokersex-smokers
FraminghamHeart Study
0ndash9 BMI 1middot4 1middot0 1middot9
Change inBMI from age55 to 65 years
Women weightloss intentionunspecified
ndash 55ndash65 ndash $10 BMI 1middot8 1middot2 2middot60ndash9 BMI 1middot1 0middot8 1middot4
Pamuk et al(1992)(39)
Apparentlyhealthy menand women
5 maxi-mum
weight loss
Men weightloss intentionunspecified
68 45ndash74 26ndash 29kgm2
6middot5 weight Age race smokingparity pre-existingillness maximumBMI
1middot7 1middot1 2middot7 Deaths infirst 5 years
NHANES I 279 26 kgm2 9middot4 weight 1middot4 1middot0 1middot897 26 kgm2 20middot3 weight 2middot4 1middot7 3middot5172 $29 kgm2 8middot8 weight 0middot7 0middot5 1middot0204 26ndash 29
kgm29middot7 weight 0middot8 0middot6 1middot1
34 $29 kgm2 18middot1 weight 0middot8 0middot4 1middot568 26ndash 29
kgm220 weight 1middot1 0middot7 1middot6
105 26 kgm2 23middot2 weight 1middot4 1middot0 2middot0Women weightloss intentionunspecified
38 45ndash74 26ndash 29kgm2
6middot6 weight 1middot4 0middot6 2middot9
223 26 kgm2 10middot2 weight 1middot6 1middot0 2middot6128 26 kgm2 20middot3 weight 2middot2 1middot3 3middot6230 $29 kgm2 9middot3 weight 1middot0 0middot7 1middot4168 26ndash 29
kgm29middot8 weight 1middot5 1middot0 2middot1
97 $29 kgm2 20middot9 weight 1middot4 0middot9 2middot180 26ndash 29
kgm220middot9 weight 1middot4 0middot9 2middot1
151 26 kgm2 25middot5 weight 1middot9 1middot3 2middot6Lee ampPaffenbar-
ger(1992)(1)
Apparentlyhealthy men
Weightstable
Men weightloss intentionunspecified
1293 58 78 kg 5 kg weight Age heightsmoking physicalactivity
1middot57 1middot341middot84
Baseline CVDcancer
Harvard AlumniHealth Study
2730 58 78 kg 1ndash5 kgweight
1middot26 1middot101middot46
Higgins et al(1993)(40)
Apparentlyhealthy menand women
No changein BMI
Men ndash 45middot5 27middot1 kgm2 0middot52 kg Age BMI systolicblood pressurecholesterolglucose intoleranceleft ventricularhypertrophysmoking
1middot33 1middot061middot68
Deaths infirst 4 years
FraminghamStudy
Women ndash 45middot8 25middot7 kgm2 0middot39 kg 1middot28 0middot981middot68
Chaturvedi ampFuller(1995)(41)
NIDDM menand women
Weight stable European menand women
252 (52deaths)
35ndash55 26 kgm2 2 kgm2 Age sex duration ofdiagnosed diabetes
3middot05 1middot267middot36
ndash
WHO study Weight lossintentionunspecified
35ndash55 26ndash29kgm2
2 kgm2 2middot02 1middot004middot08
35ndash55 $29 kgm2 2 kgm2 0middot84 0middot401middot74
M
Harrin
gto
net
al
94
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Iribarrenetal(1995)(42)
Apparentlyhealthy men
Loss of 2middot5 kgto gainof 2middot4 kg
Men Weightloss intentionunspecified
744 54middot6 25middot2 kgm2 4middot5 kg Age average weightsmoking alcoholconsumptionphysical activitytotal energy intakeemploymentpre-existingillness
1middot21 1middot021middot43
Deaths in first5 years
Honolulu HeartProgram
928 54middot2 24middot3 kgm2 2middot6ndash4middot5kg
1middot29 1middot101middot51
Manson et al(1995)(43)
Apparentlyhealthywomen
Weight stable(4 kg) sinceage 18 years
Women Weightloss intentionunspecified
Samplesize notreported
Age BMI smoking Baseline CVDcancer deathsin first 4 years
Nursesrsquo HealthStudy
16 deaths 30ndash55 ndash $10 kg 0middot7 0middot4 1middot454 deaths 30ndash55 ndash 4ndash9 kg 1middot2 0middot9 1middot6
Wallace et al(1995)(44)
Unhealthymen
Non-weightlosers
Unintentional 54 (175non-weight
losers)
73middot9 80middot8 kg 5middot6 kg Age BMI tobaccouse
hypertension healthstatus cholesterolalbumin levels
2middot83 1middot385middot81
Diseases thataffect nutritionalstatus orbody weight
Williamsonet al(1995)(45)
Healthy andunhealthywomen
No weightchange
Healthy Age baseline BMIeducation alcoholintake physicalactivity healthconditions
Deaths in first3 years
AmericanCancerPreventionStudy
Unintentional 942 52middot9 30middot9 kgm2 4middot9 kgm2 1middot20 0middot931middot55
Intentional1ndash19 lbs
2745 51middot7 30middot4 kgm2 3middot1 kgm2 1middot12 0middot941middot33
Intentional$20 lbs
3018 50middot8 33middot1 kgm2 6middot5 kgm2 0middot98 0middot821middot17
UnhealthyUnintentional 812 55middot3 31middot9 kgm2 5middot6 kgm2 1middot00 0middot83
1middot20Intentional
1ndash19 lbs1550 53middot8 31middot5 kgm2 3middot0 kgm2 0middot80 0middot68
0middot94Intentional
$20 lbs2598 53middot7 34middot8 kgm2 7middot0 kgm2 0middot81 0middot71
0middot92Yaari ampGoldbourt(1998)(46)
Unhealthymen
IsraeliIschemicHeartDiseaseStudy
Weight stable Weight lossintentional(dieters)
2471 Notgiven
78middot2 kg $ 5 kg5 kg
Age BMI systolicblood
pressure cholesterolsmoking diabetescancer history ofmyocardialinfarctionangina chroniclung diseasebaseline dietingperipheralartery disease
1middot301middot3
1middot021middot65
1middot021middot65
ndash
Weig
ht
loss
and
all-cause
mo
rtality9
5
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
French et al(1999)(47)
Apparentlyhealthywomen
Never $20 lbsweight loss
Intentional 4300 66middot6 30middot3 kgm2 $9middot1 kg Age BMI waisthipratio educationmarital statussmoking oestrogenuse cancer diabetesangina stroke heartattack hypertension
1middot18 0middot941middot48
ndash
Iowa WomenrsquosHealth Study
Unintentional 5008 68middot0 26middot1 kgm2 $9middot1 kg 1middot33 1middot131middot57
Williamsonet al
(1999)(48)
Healthy andunhealthy
men
No weightchange
Healthy Age BMI smokingeducation alcoholintake physicalactivity healthcomplaints
BMI 27 kgm2non-Caucasian
AmericanCancerPreventionStudy
Unintentional 1474 52middot0 29middot2 kgm2 3middot2 kgm2 1middot04 0middot911middot19
Intentional1ndash19 lbs
2834 51middot5 29middot0 kgm2 1middot8 kgm2 1middot09 0middot981middot21
Intentional$20 lbs
2610 51middot5 31middot4 kgm2 4middot5 kgm2 1middot07 0middot961middot20
UnhealthyUnintentional 917 54middot4 29middot7 kgm2 4middot2 kgm2 1middot15 1middot04
1middot27Intentional
1ndash19 lbs1310 53middot4 29middot1 kgm2 1middot9 kgm2 1middot01 0middot91
1middot12Intentional
$20 lbs2614 53middot6 31middot6 kgm2 4middot9 kgm2 1middot02 0middot94
1middot11Williamson
et al(2000)(4)
Unhealthymen andwomen
No or unknownweight change
Intentional 1669 54middot5 33middot5 kgm2 5middot8 kgm2 Age sex BMI racesmoking educationalcohol intakephysical activitydisease history
0middot75 0middot670middot84
BMI 27 kgm2
AmericanCancerPreventionStudy
Unintentional 649 55middot6 31middot8 kgm2 5middot9 kgm2 0middot98 0middot851middot13
Newmanet al
(2001)(49)
Older menand women
Weight stableie weightwithin ^5 of baseline
Weight lossintentionunspecified
126 deaths 77middot4 27 kgm2 $5 weight
Age sex racecognitive functionmedication smokingwaist circumferencemobility impairment
1middot67 1middot292middot15
Living in aninstitutionwheelchair usecancer treat-
ment62 deaths 77middot4 $5
weight1middot66 1middot18
2middot33Plus interimillness
Wanna-methee
et al(2002)(50)
Apparentlyhealthy men
Weight stable Weight lossintentionunspecified
950 40ndash59years
26middot6kgm2
2middot11kgm2
Age social classsmoking physicalactivity BMI CVDcancer poorhealth diabetes
1middot34 1middot091middot63
ndash
British RegionalHeart Study
M
Harrin
gto
net
al
96
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Gregg et al(2003)(31)
Overweightor obesemen andwomen
Weight stable Overall(unspecified)
1931 Over 35years(mean54middot1)
30middot8kgm2
7 kg Age sex racesmoking educationBMI self-ratedhealth diabetesacute and chronicconditions functionallimitations dueto CVD or cancerhospital bed days
1middot09 0middot901middot32
BaselineBMI 25
kgm2
NHIS UScohort
Unintentional 188 ndash ndash 6middot9 kg 1middot31 1middot011middot70
Retrospectiveweight
change9-yearfollow-up
Intentional 827 ndash ndash 7middot1 kg 0middot76 0middot600middot97
Gregg et al(2004)(51)
Unhealthy menand
women(diabetics)from NHIS US
cohort(9-yearfollow-up)
Weight stable Overall(unspecified)
629 ndash 33middot0kgm2
6middot80 kg Age sex race BMIsmoking educationself-rated healthdiabetes medicationlength of diseasefunctional limitationhypertension strokeheart disease retinaldisease neuropathyhospital days doctorvisits
1middot19 0middot9 1middot47 BaselineBMI 25
kgm2
Unintentional 365 ndash ndash 6middot80 kg 1middot58 1middot082middot31
Intentional 34 ndash ndash 6middot80 kg 0middot83 0middot631middot08
Maru et al(2004)(52)
Healthy women Weight stableie 5
weightchange
Moderate weightloss
531 50ndash66years
MedianBMI25middot4 kgm2
5ndash9 weight
Age smoking BMI 1middot14 1middot1 1middot6 Medication forhypertensionCVD diabetesrestriction diet
DOM Dutchcohort
Severe weightloss
108 IQR 23middot3ndash 27middot8kgm2
10ndash14 weight
0middot9 0middot5 1middot4
Weight changein
1-year follow-up
Extreme weightloss
43 $15 weight 0middot8 0middot4 1middot8
Median follow-up
17 yearsDiaz et al(2005)(6)
Apparentlyhealthy
men andwomen
Weight stable Weight lossintentionunspecified
711 51middot6 30middot8 kgm2 5middot55 kgm2 Age sex race BMIsmoking healthstatus poor healthincapacitated
3middot36 2middot474middot55
Diabetes CVDcancer
NHANES I andfollow-up
Droslashyvoldet al
(2005)(7)
Apparentlyhealthy
men andwomen
Weight stableie change inBMI 0middot1per year
Weight lossintentionunspecified
Age BMI systolicblood pressureblood pressuremedication smokingalcohol intakephysicalactivity maritalstatuseducation
CVD diabetescancer
Nord-Troslashnde-lag Health
Study
Men 1319 54middot3 26middot9 kgm2 2middot2 kgm2 1middot6 1middot4 1middot8Women 1971 54middot0 27middot7 kgm2 2middot7 kgm2 1middot7 1middot5 2middot0
Weig
ht
loss
and
all-cause
mo
rtality9
7
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Elliott et al(2005)(53)
Apparentlyhealthy
women
Weight changefrom 21middot81to thorn1middot36 kg
Weight lossintentionunspecified
964 42ndash81years
ndash 116middot58ndash1middot81 kg
Social class BMIparity smokinghormonereplacementtherapy
0middot96 0middot651middot43
ndash
Oral Contra-ception
StudySoslashrensen
et al(2005)(8)
Apparentlyhealthy
men andwomen
Weight stable Intentional 398 41middot5 27middot4 kgm2 1middot21 kgm2 Age sex BMIhypertensionsmoking alcoholphysical activitylife satisfactionwork status drugs
1middot87 1middot222middot87
Angina myocar-dial
infarctiondiabetes CVDlung diseasehypertensionprescription
drugsunemployment
Finnish TwinCohort
Unintentional 728 42middot6 26middot72 kgm2 1middot09 kgm2 1middot17 0middot821middot66
Wanna-methee
et al(2005)(10)
Apparentlyhealthy
men
No weightchange
Unintentional 527 40ndash59 25middot6 kgm2 3middot91 kgm2 Age smoking socialclass physicalactivityalcohol intakeobesityperceived healthstatus CVD cancerhypertension stroke
1middot71 1middot332middot19
ndash
British RegionalHeart Study
Intentional 342 40ndash59 28middot0 kgm2 2middot37 kgm2 1middot00 0middot911middot10
Intentionalpersonalreason
178 40ndash59 26middot9 kgm2 2middot31 kgm2 0middot59 0middot341middot00
Intentionalphysicianrsquosadvice
164 40ndash59 28middot5 kgm2 2middot44 kgm2 1middot37 0middot961middot94
Breeze et al(2006)(54)
Apparentlyhealthy
men
Minimal weightchange ie
loss0ndash3 kg orgain 0ndash3 kg
Weight lossintentionunspecified
554 40ndash69 ndash $10 kg Age marital statusemploymentsmoking respiratorysymptoms heartdisease indicatorsdiastolic bloodpressure totalcholesterol
1middot88 1middot6 2middot2 ndash
WhitehallCohort
1190 40ndash69 ndash 4ndash9 kg 1middot26 1middot1 1middot5
Nilsson et al(2002)(55)
HealthySwedishmen (n 5194)
Weight stable(^0middot1 kgm2)
Weight lossunspecified(no directquestion)
464 47 years(38ndash52years)
22ndash25kgm2
ndash Age 1middot39 0middot981middot95
Cancer deathsdisease atbaseline
deathsin year 1 offollow-up
482 26 thornkgm2
1middot71 1middot182middot47
M
Harrin
gto
net
al
98
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
relative risk (RR) of all-cause mortality(1314) Methodo-logical problems have also been identified for example themethod by which the weight loss was achieved has usuallynot been reported (although dietary energy restriction islikely to have been a major factor) while weight changesbefore and after the recording periods have usually not beendetermined(12 ndash 16)
In light of the current obesity epidemic and the resultingfocus on encouraging those with BMI above 25 kgm2 tolose weight by changing their diet and lifestyle(17) it isimportant to establish whether the long-term effects ofweight loss benefit life expectancy The current advice fromthe UK Department of Healthrsquos Obesity Care Programme isfor those who are overweight or obese to reduce energyintake and increase physical activity as a method of lifestylemodification(18) Further treatment and advice may need toconsider a broad spectrum of evidence so as not to rule outpotential investigations that identify subgroups of patientsor certain conditions where weight loss may be detrimentalto health and increase mortality(19)
The aim of the present study was to examine the availableevidence of the impact of weight loss as a lifestyleintervention on the RR of all-cause mortality and toquantify this using meta-analysis Data were pooled in anumber of different ways in order to examine the influenceof a number of possible confounders Meta-analysis wasused to provide a more objective appraisal of the evidenceintegrating data from multiple prospective cohort studies toincrease the power and precision of estimates of effect andreducing the likelihood of false negative results(2021)
Methods
Search strategy
A literature search was carried out independently by twoinvestigators to identify prospective cohort studies thatevaluated the effect of weight loss as a lifestyle interventionon mortality risk A web search was undertaken onPubMedMedline and ScienceDirect databases Articlespublished between 1987 and 2008 and in the Englishlanguage were included Search terms included lsquoweightBMI loss change mortality intentional unintentionalrelative risk prospective and cohortrsquo Identified citationsand abstracts were obtained from journals libraries orauthors A hand-search of the bibliographies of retrievedpapers and linked articles was also carried out
Data selection
Inclusion criteria were prospective studies in English ofadults (men andor women) with data on body weight andweight loss over more than 1 year Studies needed to presentRR of mortality and associated 95 CI for the group thatlost weight relative to a comparable reference group wholost minimal or no weight Drug treatment studies andstudies that measured weight loss following bariatricsurgery were excluded as the aim was to assess the effectof lifestyle interventions Twenty-six publications wereidentified that met the inclusion criteria Data on RR ofmortality and 95 confidence limits were extracted for allW
edic
ketal
(2002)
(56)
Calif
orn
ia
US
AW
eig
ht
sta
ble
(loss
10
lbs
or
gain
)
Weig
ht
loss
unspecifi
ed
71
years
atsta
rtof
12-y
ears
mort
alit
yfo
llow
-up
10
lbs
Age
curr
ent
and
past
sm
okin
g
exerc
ise
less
than
10
years
earlie
r
ndash
Healthy
men
628
26
1middot3
81middot0
61middot8
Healthy
wom
en
933
24
1middot7
61
middot33
2middot3
4D
iabetic
men
140
26
3middot6
62
middot15
6middot2
4D
iabetic
wom
en
90
25
1middot6
50middot7
3middot8
7In
tern
ational
weig
ht
loss
Tota
l
642
Healthy
men
1middot2
70
middot94
1middot7
1H
ealthy
Wom
en
1middot2
80middot9
5
1middot7
3
NH
AN
ES
N
ationalH
ealth
and
Nutr
itio
nE
xam
ination
Surv
ey
NID
DM
non-insulin
-dependent
dia
bete
sm
elli
tus
NH
IS
NationalH
ealth
Inte
rvie
wS
urv
ey
DO
M
Dia
gnostisch
Onderz
oek
Mam
macarc
inoom
(Dia
gnostic
Investigation
into
Bre
ast
Cancer)
IQ
R
inte
rquart
ilera
nge
1
lbfrac14
0middot4
536
kg
Weight loss and all-cause mortality 99
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Iribarrenetal(1995)(42)
Apparentlyhealthy men
Loss of 2middot5 kgto gainof 2middot4 kg
Men Weightloss intentionunspecified
744 54middot6 25middot2 kgm2 4middot5 kg Age average weightsmoking alcoholconsumptionphysical activitytotal energy intakeemploymentpre-existingillness
1middot21 1middot021middot43
Deaths in first5 years
Honolulu HeartProgram
928 54middot2 24middot3 kgm2 2middot6ndash4middot5kg
1middot29 1middot101middot51
Manson et al(1995)(43)
Apparentlyhealthywomen
Weight stable(4 kg) sinceage 18 years
Women Weightloss intentionunspecified
Samplesize notreported
Age BMI smoking Baseline CVDcancer deathsin first 4 years
Nursesrsquo HealthStudy
16 deaths 30ndash55 ndash $10 kg 0middot7 0middot4 1middot454 deaths 30ndash55 ndash 4ndash9 kg 1middot2 0middot9 1middot6
Wallace et al(1995)(44)
Unhealthymen
Non-weightlosers
Unintentional 54 (175non-weight
losers)
73middot9 80middot8 kg 5middot6 kg Age BMI tobaccouse
hypertension healthstatus cholesterolalbumin levels
2middot83 1middot385middot81
Diseases thataffect nutritionalstatus orbody weight
Williamsonet al(1995)(45)
Healthy andunhealthywomen
No weightchange
Healthy Age baseline BMIeducation alcoholintake physicalactivity healthconditions
Deaths in first3 years
AmericanCancerPreventionStudy
Unintentional 942 52middot9 30middot9 kgm2 4middot9 kgm2 1middot20 0middot931middot55
Intentional1ndash19 lbs
2745 51middot7 30middot4 kgm2 3middot1 kgm2 1middot12 0middot941middot33
Intentional$20 lbs
3018 50middot8 33middot1 kgm2 6middot5 kgm2 0middot98 0middot821middot17
UnhealthyUnintentional 812 55middot3 31middot9 kgm2 5middot6 kgm2 1middot00 0middot83
1middot20Intentional
1ndash19 lbs1550 53middot8 31middot5 kgm2 3middot0 kgm2 0middot80 0middot68
0middot94Intentional
$20 lbs2598 53middot7 34middot8 kgm2 7middot0 kgm2 0middot81 0middot71
0middot92Yaari ampGoldbourt(1998)(46)
Unhealthymen
IsraeliIschemicHeartDiseaseStudy
Weight stable Weight lossintentional(dieters)
2471 Notgiven
78middot2 kg $ 5 kg5 kg
Age BMI systolicblood
pressure cholesterolsmoking diabetescancer history ofmyocardialinfarctionangina chroniclung diseasebaseline dietingperipheralartery disease
1middot301middot3
1middot021middot65
1middot021middot65
ndash
Weig
ht
loss
and
all-cause
mo
rtality9
5
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
French et al(1999)(47)
Apparentlyhealthywomen
Never $20 lbsweight loss
Intentional 4300 66middot6 30middot3 kgm2 $9middot1 kg Age BMI waisthipratio educationmarital statussmoking oestrogenuse cancer diabetesangina stroke heartattack hypertension
1middot18 0middot941middot48
ndash
Iowa WomenrsquosHealth Study
Unintentional 5008 68middot0 26middot1 kgm2 $9middot1 kg 1middot33 1middot131middot57
Williamsonet al
(1999)(48)
Healthy andunhealthy
men
No weightchange
Healthy Age BMI smokingeducation alcoholintake physicalactivity healthcomplaints
BMI 27 kgm2non-Caucasian
AmericanCancerPreventionStudy
Unintentional 1474 52middot0 29middot2 kgm2 3middot2 kgm2 1middot04 0middot911middot19
Intentional1ndash19 lbs
2834 51middot5 29middot0 kgm2 1middot8 kgm2 1middot09 0middot981middot21
Intentional$20 lbs
2610 51middot5 31middot4 kgm2 4middot5 kgm2 1middot07 0middot961middot20
UnhealthyUnintentional 917 54middot4 29middot7 kgm2 4middot2 kgm2 1middot15 1middot04
1middot27Intentional
1ndash19 lbs1310 53middot4 29middot1 kgm2 1middot9 kgm2 1middot01 0middot91
1middot12Intentional
$20 lbs2614 53middot6 31middot6 kgm2 4middot9 kgm2 1middot02 0middot94
1middot11Williamson
et al(2000)(4)
Unhealthymen andwomen
No or unknownweight change
Intentional 1669 54middot5 33middot5 kgm2 5middot8 kgm2 Age sex BMI racesmoking educationalcohol intakephysical activitydisease history
0middot75 0middot670middot84
BMI 27 kgm2
AmericanCancerPreventionStudy
Unintentional 649 55middot6 31middot8 kgm2 5middot9 kgm2 0middot98 0middot851middot13
Newmanet al
(2001)(49)
Older menand women
Weight stableie weightwithin ^5 of baseline
Weight lossintentionunspecified
126 deaths 77middot4 27 kgm2 $5 weight
Age sex racecognitive functionmedication smokingwaist circumferencemobility impairment
1middot67 1middot292middot15
Living in aninstitutionwheelchair usecancer treat-
ment62 deaths 77middot4 $5
weight1middot66 1middot18
2middot33Plus interimillness
Wanna-methee
et al(2002)(50)
Apparentlyhealthy men
Weight stable Weight lossintentionunspecified
950 40ndash59years
26middot6kgm2
2middot11kgm2
Age social classsmoking physicalactivity BMI CVDcancer poorhealth diabetes
1middot34 1middot091middot63
ndash
British RegionalHeart Study
M
Harrin
gto
net
al
96
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Gregg et al(2003)(31)
Overweightor obesemen andwomen
Weight stable Overall(unspecified)
1931 Over 35years(mean54middot1)
30middot8kgm2
7 kg Age sex racesmoking educationBMI self-ratedhealth diabetesacute and chronicconditions functionallimitations dueto CVD or cancerhospital bed days
1middot09 0middot901middot32
BaselineBMI 25
kgm2
NHIS UScohort
Unintentional 188 ndash ndash 6middot9 kg 1middot31 1middot011middot70
Retrospectiveweight
change9-yearfollow-up
Intentional 827 ndash ndash 7middot1 kg 0middot76 0middot600middot97
Gregg et al(2004)(51)
Unhealthy menand
women(diabetics)from NHIS US
cohort(9-yearfollow-up)
Weight stable Overall(unspecified)
629 ndash 33middot0kgm2
6middot80 kg Age sex race BMIsmoking educationself-rated healthdiabetes medicationlength of diseasefunctional limitationhypertension strokeheart disease retinaldisease neuropathyhospital days doctorvisits
1middot19 0middot9 1middot47 BaselineBMI 25
kgm2
Unintentional 365 ndash ndash 6middot80 kg 1middot58 1middot082middot31
Intentional 34 ndash ndash 6middot80 kg 0middot83 0middot631middot08
Maru et al(2004)(52)
Healthy women Weight stableie 5
weightchange
Moderate weightloss
531 50ndash66years
MedianBMI25middot4 kgm2
5ndash9 weight
Age smoking BMI 1middot14 1middot1 1middot6 Medication forhypertensionCVD diabetesrestriction diet
DOM Dutchcohort
Severe weightloss
108 IQR 23middot3ndash 27middot8kgm2
10ndash14 weight
0middot9 0middot5 1middot4
Weight changein
1-year follow-up
Extreme weightloss
43 $15 weight 0middot8 0middot4 1middot8
Median follow-up
17 yearsDiaz et al(2005)(6)
Apparentlyhealthy
men andwomen
Weight stable Weight lossintentionunspecified
711 51middot6 30middot8 kgm2 5middot55 kgm2 Age sex race BMIsmoking healthstatus poor healthincapacitated
3middot36 2middot474middot55
Diabetes CVDcancer
NHANES I andfollow-up
Droslashyvoldet al
(2005)(7)
Apparentlyhealthy
men andwomen
Weight stableie change inBMI 0middot1per year
Weight lossintentionunspecified
Age BMI systolicblood pressureblood pressuremedication smokingalcohol intakephysicalactivity maritalstatuseducation
CVD diabetescancer
Nord-Troslashnde-lag Health
Study
Men 1319 54middot3 26middot9 kgm2 2middot2 kgm2 1middot6 1middot4 1middot8Women 1971 54middot0 27middot7 kgm2 2middot7 kgm2 1middot7 1middot5 2middot0
Weig
ht
loss
and
all-cause
mo
rtality9
7
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Elliott et al(2005)(53)
Apparentlyhealthy
women
Weight changefrom 21middot81to thorn1middot36 kg
Weight lossintentionunspecified
964 42ndash81years
ndash 116middot58ndash1middot81 kg
Social class BMIparity smokinghormonereplacementtherapy
0middot96 0middot651middot43
ndash
Oral Contra-ception
StudySoslashrensen
et al(2005)(8)
Apparentlyhealthy
men andwomen
Weight stable Intentional 398 41middot5 27middot4 kgm2 1middot21 kgm2 Age sex BMIhypertensionsmoking alcoholphysical activitylife satisfactionwork status drugs
1middot87 1middot222middot87
Angina myocar-dial
infarctiondiabetes CVDlung diseasehypertensionprescription
drugsunemployment
Finnish TwinCohort
Unintentional 728 42middot6 26middot72 kgm2 1middot09 kgm2 1middot17 0middot821middot66
Wanna-methee
et al(2005)(10)
Apparentlyhealthy
men
No weightchange
Unintentional 527 40ndash59 25middot6 kgm2 3middot91 kgm2 Age smoking socialclass physicalactivityalcohol intakeobesityperceived healthstatus CVD cancerhypertension stroke
1middot71 1middot332middot19
ndash
British RegionalHeart Study
Intentional 342 40ndash59 28middot0 kgm2 2middot37 kgm2 1middot00 0middot911middot10
Intentionalpersonalreason
178 40ndash59 26middot9 kgm2 2middot31 kgm2 0middot59 0middot341middot00
Intentionalphysicianrsquosadvice
164 40ndash59 28middot5 kgm2 2middot44 kgm2 1middot37 0middot961middot94
Breeze et al(2006)(54)
Apparentlyhealthy
men
Minimal weightchange ie
loss0ndash3 kg orgain 0ndash3 kg
Weight lossintentionunspecified
554 40ndash69 ndash $10 kg Age marital statusemploymentsmoking respiratorysymptoms heartdisease indicatorsdiastolic bloodpressure totalcholesterol
1middot88 1middot6 2middot2 ndash
WhitehallCohort
1190 40ndash69 ndash 4ndash9 kg 1middot26 1middot1 1middot5
Nilsson et al(2002)(55)
HealthySwedishmen (n 5194)
Weight stable(^0middot1 kgm2)
Weight lossunspecified(no directquestion)
464 47 years(38ndash52years)
22ndash25kgm2
ndash Age 1middot39 0middot981middot95
Cancer deathsdisease atbaseline
deathsin year 1 offollow-up
482 26 thornkgm2
1middot71 1middot182middot47
M
Harrin
gto
net
al
98
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
relative risk (RR) of all-cause mortality(1314) Methodo-logical problems have also been identified for example themethod by which the weight loss was achieved has usuallynot been reported (although dietary energy restriction islikely to have been a major factor) while weight changesbefore and after the recording periods have usually not beendetermined(12 ndash 16)
In light of the current obesity epidemic and the resultingfocus on encouraging those with BMI above 25 kgm2 tolose weight by changing their diet and lifestyle(17) it isimportant to establish whether the long-term effects ofweight loss benefit life expectancy The current advice fromthe UK Department of Healthrsquos Obesity Care Programme isfor those who are overweight or obese to reduce energyintake and increase physical activity as a method of lifestylemodification(18) Further treatment and advice may need toconsider a broad spectrum of evidence so as not to rule outpotential investigations that identify subgroups of patientsor certain conditions where weight loss may be detrimentalto health and increase mortality(19)
The aim of the present study was to examine the availableevidence of the impact of weight loss as a lifestyleintervention on the RR of all-cause mortality and toquantify this using meta-analysis Data were pooled in anumber of different ways in order to examine the influenceof a number of possible confounders Meta-analysis wasused to provide a more objective appraisal of the evidenceintegrating data from multiple prospective cohort studies toincrease the power and precision of estimates of effect andreducing the likelihood of false negative results(2021)
Methods
Search strategy
A literature search was carried out independently by twoinvestigators to identify prospective cohort studies thatevaluated the effect of weight loss as a lifestyle interventionon mortality risk A web search was undertaken onPubMedMedline and ScienceDirect databases Articlespublished between 1987 and 2008 and in the Englishlanguage were included Search terms included lsquoweightBMI loss change mortality intentional unintentionalrelative risk prospective and cohortrsquo Identified citationsand abstracts were obtained from journals libraries orauthors A hand-search of the bibliographies of retrievedpapers and linked articles was also carried out
Data selection
Inclusion criteria were prospective studies in English ofadults (men andor women) with data on body weight andweight loss over more than 1 year Studies needed to presentRR of mortality and associated 95 CI for the group thatlost weight relative to a comparable reference group wholost minimal or no weight Drug treatment studies andstudies that measured weight loss following bariatricsurgery were excluded as the aim was to assess the effectof lifestyle interventions Twenty-six publications wereidentified that met the inclusion criteria Data on RR ofmortality and 95 confidence limits were extracted for allW
edic
ketal
(2002)
(56)
Calif
orn
ia
US
AW
eig
ht
sta
ble
(loss
10
lbs
or
gain
)
Weig
ht
loss
unspecifi
ed
71
years
atsta
rtof
12-y
ears
mort
alit
yfo
llow
-up
10
lbs
Age
curr
ent
and
past
sm
okin
g
exerc
ise
less
than
10
years
earlie
r
ndash
Healthy
men
628
26
1middot3
81middot0
61middot8
Healthy
wom
en
933
24
1middot7
61
middot33
2middot3
4D
iabetic
men
140
26
3middot6
62
middot15
6middot2
4D
iabetic
wom
en
90
25
1middot6
50middot7
3middot8
7In
tern
ational
weig
ht
loss
Tota
l
642
Healthy
men
1middot2
70
middot94
1middot7
1H
ealthy
Wom
en
1middot2
80middot9
5
1middot7
3
NH
AN
ES
N
ationalH
ealth
and
Nutr
itio
nE
xam
ination
Surv
ey
NID
DM
non-insulin
-dependent
dia
bete
sm
elli
tus
NH
IS
NationalH
ealth
Inte
rvie
wS
urv
ey
DO
M
Dia
gnostisch
Onderz
oek
Mam
macarc
inoom
(Dia
gnostic
Investigation
into
Bre
ast
Cancer)
IQ
R
inte
rquart
ilera
nge
1
lbfrac14
0middot4
536
kg
Weight loss and all-cause mortality 99
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
French et al(1999)(47)
Apparentlyhealthywomen
Never $20 lbsweight loss
Intentional 4300 66middot6 30middot3 kgm2 $9middot1 kg Age BMI waisthipratio educationmarital statussmoking oestrogenuse cancer diabetesangina stroke heartattack hypertension
1middot18 0middot941middot48
ndash
Iowa WomenrsquosHealth Study
Unintentional 5008 68middot0 26middot1 kgm2 $9middot1 kg 1middot33 1middot131middot57
Williamsonet al
(1999)(48)
Healthy andunhealthy
men
No weightchange
Healthy Age BMI smokingeducation alcoholintake physicalactivity healthcomplaints
BMI 27 kgm2non-Caucasian
AmericanCancerPreventionStudy
Unintentional 1474 52middot0 29middot2 kgm2 3middot2 kgm2 1middot04 0middot911middot19
Intentional1ndash19 lbs
2834 51middot5 29middot0 kgm2 1middot8 kgm2 1middot09 0middot981middot21
Intentional$20 lbs
2610 51middot5 31middot4 kgm2 4middot5 kgm2 1middot07 0middot961middot20
UnhealthyUnintentional 917 54middot4 29middot7 kgm2 4middot2 kgm2 1middot15 1middot04
1middot27Intentional
1ndash19 lbs1310 53middot4 29middot1 kgm2 1middot9 kgm2 1middot01 0middot91
1middot12Intentional
$20 lbs2614 53middot6 31middot6 kgm2 4middot9 kgm2 1middot02 0middot94
1middot11Williamson
et al(2000)(4)
Unhealthymen andwomen
No or unknownweight change
Intentional 1669 54middot5 33middot5 kgm2 5middot8 kgm2 Age sex BMI racesmoking educationalcohol intakephysical activitydisease history
0middot75 0middot670middot84
BMI 27 kgm2
AmericanCancerPreventionStudy
Unintentional 649 55middot6 31middot8 kgm2 5middot9 kgm2 0middot98 0middot851middot13
Newmanet al
(2001)(49)
Older menand women
Weight stableie weightwithin ^5 of baseline
Weight lossintentionunspecified
126 deaths 77middot4 27 kgm2 $5 weight
Age sex racecognitive functionmedication smokingwaist circumferencemobility impairment
1middot67 1middot292middot15
Living in aninstitutionwheelchair usecancer treat-
ment62 deaths 77middot4 $5
weight1middot66 1middot18
2middot33Plus interimillness
Wanna-methee
et al(2002)(50)
Apparentlyhealthy men
Weight stable Weight lossintentionunspecified
950 40ndash59years
26middot6kgm2
2middot11kgm2
Age social classsmoking physicalactivity BMI CVDcancer poorhealth diabetes
1middot34 1middot091middot63
ndash
British RegionalHeart Study
M
Harrin
gto
net
al
96
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Gregg et al(2003)(31)
Overweightor obesemen andwomen
Weight stable Overall(unspecified)
1931 Over 35years(mean54middot1)
30middot8kgm2
7 kg Age sex racesmoking educationBMI self-ratedhealth diabetesacute and chronicconditions functionallimitations dueto CVD or cancerhospital bed days
1middot09 0middot901middot32
BaselineBMI 25
kgm2
NHIS UScohort
Unintentional 188 ndash ndash 6middot9 kg 1middot31 1middot011middot70
Retrospectiveweight
change9-yearfollow-up
Intentional 827 ndash ndash 7middot1 kg 0middot76 0middot600middot97
Gregg et al(2004)(51)
Unhealthy menand
women(diabetics)from NHIS US
cohort(9-yearfollow-up)
Weight stable Overall(unspecified)
629 ndash 33middot0kgm2
6middot80 kg Age sex race BMIsmoking educationself-rated healthdiabetes medicationlength of diseasefunctional limitationhypertension strokeheart disease retinaldisease neuropathyhospital days doctorvisits
1middot19 0middot9 1middot47 BaselineBMI 25
kgm2
Unintentional 365 ndash ndash 6middot80 kg 1middot58 1middot082middot31
Intentional 34 ndash ndash 6middot80 kg 0middot83 0middot631middot08
Maru et al(2004)(52)
Healthy women Weight stableie 5
weightchange
Moderate weightloss
531 50ndash66years
MedianBMI25middot4 kgm2
5ndash9 weight
Age smoking BMI 1middot14 1middot1 1middot6 Medication forhypertensionCVD diabetesrestriction diet
DOM Dutchcohort
Severe weightloss
108 IQR 23middot3ndash 27middot8kgm2
10ndash14 weight
0middot9 0middot5 1middot4
Weight changein
1-year follow-up
Extreme weightloss
43 $15 weight 0middot8 0middot4 1middot8
Median follow-up
17 yearsDiaz et al(2005)(6)
Apparentlyhealthy
men andwomen
Weight stable Weight lossintentionunspecified
711 51middot6 30middot8 kgm2 5middot55 kgm2 Age sex race BMIsmoking healthstatus poor healthincapacitated
3middot36 2middot474middot55
Diabetes CVDcancer
NHANES I andfollow-up
Droslashyvoldet al
(2005)(7)
Apparentlyhealthy
men andwomen
Weight stableie change inBMI 0middot1per year
Weight lossintentionunspecified
Age BMI systolicblood pressureblood pressuremedication smokingalcohol intakephysicalactivity maritalstatuseducation
CVD diabetescancer
Nord-Troslashnde-lag Health
Study
Men 1319 54middot3 26middot9 kgm2 2middot2 kgm2 1middot6 1middot4 1middot8Women 1971 54middot0 27middot7 kgm2 2middot7 kgm2 1middot7 1middot5 2middot0
Weig
ht
loss
and
all-cause
mo
rtality9
7
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Elliott et al(2005)(53)
Apparentlyhealthy
women
Weight changefrom 21middot81to thorn1middot36 kg
Weight lossintentionunspecified
964 42ndash81years
ndash 116middot58ndash1middot81 kg
Social class BMIparity smokinghormonereplacementtherapy
0middot96 0middot651middot43
ndash
Oral Contra-ception
StudySoslashrensen
et al(2005)(8)
Apparentlyhealthy
men andwomen
Weight stable Intentional 398 41middot5 27middot4 kgm2 1middot21 kgm2 Age sex BMIhypertensionsmoking alcoholphysical activitylife satisfactionwork status drugs
1middot87 1middot222middot87
Angina myocar-dial
infarctiondiabetes CVDlung diseasehypertensionprescription
drugsunemployment
Finnish TwinCohort
Unintentional 728 42middot6 26middot72 kgm2 1middot09 kgm2 1middot17 0middot821middot66
Wanna-methee
et al(2005)(10)
Apparentlyhealthy
men
No weightchange
Unintentional 527 40ndash59 25middot6 kgm2 3middot91 kgm2 Age smoking socialclass physicalactivityalcohol intakeobesityperceived healthstatus CVD cancerhypertension stroke
1middot71 1middot332middot19
ndash
British RegionalHeart Study
Intentional 342 40ndash59 28middot0 kgm2 2middot37 kgm2 1middot00 0middot911middot10
Intentionalpersonalreason
178 40ndash59 26middot9 kgm2 2middot31 kgm2 0middot59 0middot341middot00
Intentionalphysicianrsquosadvice
164 40ndash59 28middot5 kgm2 2middot44 kgm2 1middot37 0middot961middot94
Breeze et al(2006)(54)
Apparentlyhealthy
men
Minimal weightchange ie
loss0ndash3 kg orgain 0ndash3 kg
Weight lossintentionunspecified
554 40ndash69 ndash $10 kg Age marital statusemploymentsmoking respiratorysymptoms heartdisease indicatorsdiastolic bloodpressure totalcholesterol
1middot88 1middot6 2middot2 ndash
WhitehallCohort
1190 40ndash69 ndash 4ndash9 kg 1middot26 1middot1 1middot5
Nilsson et al(2002)(55)
HealthySwedishmen (n 5194)
Weight stable(^0middot1 kgm2)
Weight lossunspecified(no directquestion)
464 47 years(38ndash52years)
22ndash25kgm2
ndash Age 1middot39 0middot981middot95
Cancer deathsdisease atbaseline
deathsin year 1 offollow-up
482 26 thornkgm2
1middot71 1middot182middot47
M
Harrin
gto
net
al
98
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
relative risk (RR) of all-cause mortality(1314) Methodo-logical problems have also been identified for example themethod by which the weight loss was achieved has usuallynot been reported (although dietary energy restriction islikely to have been a major factor) while weight changesbefore and after the recording periods have usually not beendetermined(12 ndash 16)
In light of the current obesity epidemic and the resultingfocus on encouraging those with BMI above 25 kgm2 tolose weight by changing their diet and lifestyle(17) it isimportant to establish whether the long-term effects ofweight loss benefit life expectancy The current advice fromthe UK Department of Healthrsquos Obesity Care Programme isfor those who are overweight or obese to reduce energyintake and increase physical activity as a method of lifestylemodification(18) Further treatment and advice may need toconsider a broad spectrum of evidence so as not to rule outpotential investigations that identify subgroups of patientsor certain conditions where weight loss may be detrimentalto health and increase mortality(19)
The aim of the present study was to examine the availableevidence of the impact of weight loss as a lifestyleintervention on the RR of all-cause mortality and toquantify this using meta-analysis Data were pooled in anumber of different ways in order to examine the influenceof a number of possible confounders Meta-analysis wasused to provide a more objective appraisal of the evidenceintegrating data from multiple prospective cohort studies toincrease the power and precision of estimates of effect andreducing the likelihood of false negative results(2021)
Methods
Search strategy
A literature search was carried out independently by twoinvestigators to identify prospective cohort studies thatevaluated the effect of weight loss as a lifestyle interventionon mortality risk A web search was undertaken onPubMedMedline and ScienceDirect databases Articlespublished between 1987 and 2008 and in the Englishlanguage were included Search terms included lsquoweightBMI loss change mortality intentional unintentionalrelative risk prospective and cohortrsquo Identified citationsand abstracts were obtained from journals libraries orauthors A hand-search of the bibliographies of retrievedpapers and linked articles was also carried out
Data selection
Inclusion criteria were prospective studies in English ofadults (men andor women) with data on body weight andweight loss over more than 1 year Studies needed to presentRR of mortality and associated 95 CI for the group thatlost weight relative to a comparable reference group wholost minimal or no weight Drug treatment studies andstudies that measured weight loss following bariatricsurgery were excluded as the aim was to assess the effectof lifestyle interventions Twenty-six publications wereidentified that met the inclusion criteria Data on RR ofmortality and 95 confidence limits were extracted for allW
edic
ketal
(2002)
(56)
Calif
orn
ia
US
AW
eig
ht
sta
ble
(loss
10
lbs
or
gain
)
Weig
ht
loss
unspecifi
ed
71
years
atsta
rtof
12-y
ears
mort
alit
yfo
llow
-up
10
lbs
Age
curr
ent
and
past
sm
okin
g
exerc
ise
less
than
10
years
earlie
r
ndash
Healthy
men
628
26
1middot3
81middot0
61middot8
Healthy
wom
en
933
24
1middot7
61
middot33
2middot3
4D
iabetic
men
140
26
3middot6
62
middot15
6middot2
4D
iabetic
wom
en
90
25
1middot6
50middot7
3middot8
7In
tern
ational
weig
ht
loss
Tota
l
642
Healthy
men
1middot2
70
middot94
1middot7
1H
ealthy
Wom
en
1middot2
80middot9
5
1middot7
3
NH
AN
ES
N
ationalH
ealth
and
Nutr
itio
nE
xam
ination
Surv
ey
NID
DM
non-insulin
-dependent
dia
bete
sm
elli
tus
NH
IS
NationalH
ealth
Inte
rvie
wS
urv
ey
DO
M
Dia
gnostisch
Onderz
oek
Mam
macarc
inoom
(Dia
gnostic
Investigation
into
Bre
ast
Cancer)
IQ
R
inte
rquart
ilera
nge
1
lbfrac14
0middot4
536
kg
Weight loss and all-cause mortality 99
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Gregg et al(2003)(31)
Overweightor obesemen andwomen
Weight stable Overall(unspecified)
1931 Over 35years(mean54middot1)
30middot8kgm2
7 kg Age sex racesmoking educationBMI self-ratedhealth diabetesacute and chronicconditions functionallimitations dueto CVD or cancerhospital bed days
1middot09 0middot901middot32
BaselineBMI 25
kgm2
NHIS UScohort
Unintentional 188 ndash ndash 6middot9 kg 1middot31 1middot011middot70
Retrospectiveweight
change9-yearfollow-up
Intentional 827 ndash ndash 7middot1 kg 0middot76 0middot600middot97
Gregg et al(2004)(51)
Unhealthy menand
women(diabetics)from NHIS US
cohort(9-yearfollow-up)
Weight stable Overall(unspecified)
629 ndash 33middot0kgm2
6middot80 kg Age sex race BMIsmoking educationself-rated healthdiabetes medicationlength of diseasefunctional limitationhypertension strokeheart disease retinaldisease neuropathyhospital days doctorvisits
1middot19 0middot9 1middot47 BaselineBMI 25
kgm2
Unintentional 365 ndash ndash 6middot80 kg 1middot58 1middot082middot31
Intentional 34 ndash ndash 6middot80 kg 0middot83 0middot631middot08
Maru et al(2004)(52)
Healthy women Weight stableie 5
weightchange
Moderate weightloss
531 50ndash66years
MedianBMI25middot4 kgm2
5ndash9 weight
Age smoking BMI 1middot14 1middot1 1middot6 Medication forhypertensionCVD diabetesrestriction diet
DOM Dutchcohort
Severe weightloss
108 IQR 23middot3ndash 27middot8kgm2
10ndash14 weight
0middot9 0middot5 1middot4
Weight changein
1-year follow-up
Extreme weightloss
43 $15 weight 0middot8 0middot4 1middot8
Median follow-up
17 yearsDiaz et al(2005)(6)
Apparentlyhealthy
men andwomen
Weight stable Weight lossintentionunspecified
711 51middot6 30middot8 kgm2 5middot55 kgm2 Age sex race BMIsmoking healthstatus poor healthincapacitated
3middot36 2middot474middot55
Diabetes CVDcancer
NHANES I andfollow-up
Droslashyvoldet al
(2005)(7)
Apparentlyhealthy
men andwomen
Weight stableie change inBMI 0middot1per year
Weight lossintentionunspecified
Age BMI systolicblood pressureblood pressuremedication smokingalcohol intakephysicalactivity maritalstatuseducation
CVD diabetescancer
Nord-Troslashnde-lag Health
Study
Men 1319 54middot3 26middot9 kgm2 2middot2 kgm2 1middot6 1middot4 1middot8Women 1971 54middot0 27middot7 kgm2 2middot7 kgm2 1middot7 1middot5 2middot0
Weig
ht
loss
and
all-cause
mo
rtality9
7
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Elliott et al(2005)(53)
Apparentlyhealthy
women
Weight changefrom 21middot81to thorn1middot36 kg
Weight lossintentionunspecified
964 42ndash81years
ndash 116middot58ndash1middot81 kg
Social class BMIparity smokinghormonereplacementtherapy
0middot96 0middot651middot43
ndash
Oral Contra-ception
StudySoslashrensen
et al(2005)(8)
Apparentlyhealthy
men andwomen
Weight stable Intentional 398 41middot5 27middot4 kgm2 1middot21 kgm2 Age sex BMIhypertensionsmoking alcoholphysical activitylife satisfactionwork status drugs
1middot87 1middot222middot87
Angina myocar-dial
infarctiondiabetes CVDlung diseasehypertensionprescription
drugsunemployment
Finnish TwinCohort
Unintentional 728 42middot6 26middot72 kgm2 1middot09 kgm2 1middot17 0middot821middot66
Wanna-methee
et al(2005)(10)
Apparentlyhealthy
men
No weightchange
Unintentional 527 40ndash59 25middot6 kgm2 3middot91 kgm2 Age smoking socialclass physicalactivityalcohol intakeobesityperceived healthstatus CVD cancerhypertension stroke
1middot71 1middot332middot19
ndash
British RegionalHeart Study
Intentional 342 40ndash59 28middot0 kgm2 2middot37 kgm2 1middot00 0middot911middot10
Intentionalpersonalreason
178 40ndash59 26middot9 kgm2 2middot31 kgm2 0middot59 0middot341middot00
Intentionalphysicianrsquosadvice
164 40ndash59 28middot5 kgm2 2middot44 kgm2 1middot37 0middot961middot94
Breeze et al(2006)(54)
Apparentlyhealthy
men
Minimal weightchange ie
loss0ndash3 kg orgain 0ndash3 kg
Weight lossintentionunspecified
554 40ndash69 ndash $10 kg Age marital statusemploymentsmoking respiratorysymptoms heartdisease indicatorsdiastolic bloodpressure totalcholesterol
1middot88 1middot6 2middot2 ndash
WhitehallCohort
1190 40ndash69 ndash 4ndash9 kg 1middot26 1middot1 1middot5
Nilsson et al(2002)(55)
HealthySwedishmen (n 5194)
Weight stable(^0middot1 kgm2)
Weight lossunspecified(no directquestion)
464 47 years(38ndash52years)
22ndash25kgm2
ndash Age 1middot39 0middot981middot95
Cancer deathsdisease atbaseline
deathsin year 1 offollow-up
482 26 thornkgm2
1middot71 1middot182middot47
M
Harrin
gto
net
al
98
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
relative risk (RR) of all-cause mortality(1314) Methodo-logical problems have also been identified for example themethod by which the weight loss was achieved has usuallynot been reported (although dietary energy restriction islikely to have been a major factor) while weight changesbefore and after the recording periods have usually not beendetermined(12 ndash 16)
In light of the current obesity epidemic and the resultingfocus on encouraging those with BMI above 25 kgm2 tolose weight by changing their diet and lifestyle(17) it isimportant to establish whether the long-term effects ofweight loss benefit life expectancy The current advice fromthe UK Department of Healthrsquos Obesity Care Programme isfor those who are overweight or obese to reduce energyintake and increase physical activity as a method of lifestylemodification(18) Further treatment and advice may need toconsider a broad spectrum of evidence so as not to rule outpotential investigations that identify subgroups of patientsor certain conditions where weight loss may be detrimentalto health and increase mortality(19)
The aim of the present study was to examine the availableevidence of the impact of weight loss as a lifestyleintervention on the RR of all-cause mortality and toquantify this using meta-analysis Data were pooled in anumber of different ways in order to examine the influenceof a number of possible confounders Meta-analysis wasused to provide a more objective appraisal of the evidenceintegrating data from multiple prospective cohort studies toincrease the power and precision of estimates of effect andreducing the likelihood of false negative results(2021)
Methods
Search strategy
A literature search was carried out independently by twoinvestigators to identify prospective cohort studies thatevaluated the effect of weight loss as a lifestyle interventionon mortality risk A web search was undertaken onPubMedMedline and ScienceDirect databases Articlespublished between 1987 and 2008 and in the Englishlanguage were included Search terms included lsquoweightBMI loss change mortality intentional unintentionalrelative risk prospective and cohortrsquo Identified citationsand abstracts were obtained from journals libraries orauthors A hand-search of the bibliographies of retrievedpapers and linked articles was also carried out
Data selection
Inclusion criteria were prospective studies in English ofadults (men andor women) with data on body weight andweight loss over more than 1 year Studies needed to presentRR of mortality and associated 95 CI for the group thatlost weight relative to a comparable reference group wholost minimal or no weight Drug treatment studies andstudies that measured weight loss following bariatricsurgery were excluded as the aim was to assess the effectof lifestyle interventions Twenty-six publications wereidentified that met the inclusion criteria Data on RR ofmortality and 95 confidence limits were extracted for allW
edic
ketal
(2002)
(56)
Calif
orn
ia
US
AW
eig
ht
sta
ble
(loss
10
lbs
or
gain
)
Weig
ht
loss
unspecifi
ed
71
years
atsta
rtof
12-y
ears
mort
alit
yfo
llow
-up
10
lbs
Age
curr
ent
and
past
sm
okin
g
exerc
ise
less
than
10
years
earlie
r
ndash
Healthy
men
628
26
1middot3
81middot0
61middot8
Healthy
wom
en
933
24
1middot7
61
middot33
2middot3
4D
iabetic
men
140
26
3middot6
62
middot15
6middot2
4D
iabetic
wom
en
90
25
1middot6
50middot7
3middot8
7In
tern
ational
weig
ht
loss
Tota
l
642
Healthy
men
1middot2
70
middot94
1middot7
1H
ealthy
Wom
en
1middot2
80middot9
5
1middot7
3
NH
AN
ES
N
ationalH
ealth
and
Nutr
itio
nE
xam
ination
Surv
ey
NID
DM
non-insulin
-dependent
dia
bete
sm
elli
tus
NH
IS
NationalH
ealth
Inte
rvie
wS
urv
ey
DO
M
Dia
gnostisch
Onderz
oek
Mam
macarc
inoom
(Dia
gnostic
Investigation
into
Bre
ast
Cancer)
IQ
R
inte
rquart
ilera
nge
1
lbfrac14
0middot4
536
kg
Weight loss and all-cause mortality 99
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Table 1 Continued
Studyreference Population
Referentgroup
Weightloss
subgroupSamplesize (n)
Meanbaselineageandorrange(years)
Baselineweight(kg) orBMI(kgm2)
Weight(kg) orBMI loss(kgm2)
Dataadjustments
Relativerisk 95 CI Exclusions
Elliott et al(2005)(53)
Apparentlyhealthy
women
Weight changefrom 21middot81to thorn1middot36 kg
Weight lossintentionunspecified
964 42ndash81years
ndash 116middot58ndash1middot81 kg
Social class BMIparity smokinghormonereplacementtherapy
0middot96 0middot651middot43
ndash
Oral Contra-ception
StudySoslashrensen
et al(2005)(8)
Apparentlyhealthy
men andwomen
Weight stable Intentional 398 41middot5 27middot4 kgm2 1middot21 kgm2 Age sex BMIhypertensionsmoking alcoholphysical activitylife satisfactionwork status drugs
1middot87 1middot222middot87
Angina myocar-dial
infarctiondiabetes CVDlung diseasehypertensionprescription
drugsunemployment
Finnish TwinCohort
Unintentional 728 42middot6 26middot72 kgm2 1middot09 kgm2 1middot17 0middot821middot66
Wanna-methee
et al(2005)(10)
Apparentlyhealthy
men
No weightchange
Unintentional 527 40ndash59 25middot6 kgm2 3middot91 kgm2 Age smoking socialclass physicalactivityalcohol intakeobesityperceived healthstatus CVD cancerhypertension stroke
1middot71 1middot332middot19
ndash
British RegionalHeart Study
Intentional 342 40ndash59 28middot0 kgm2 2middot37 kgm2 1middot00 0middot911middot10
Intentionalpersonalreason
178 40ndash59 26middot9 kgm2 2middot31 kgm2 0middot59 0middot341middot00
Intentionalphysicianrsquosadvice
164 40ndash59 28middot5 kgm2 2middot44 kgm2 1middot37 0middot961middot94
Breeze et al(2006)(54)
Apparentlyhealthy
men
Minimal weightchange ie
loss0ndash3 kg orgain 0ndash3 kg
Weight lossintentionunspecified
554 40ndash69 ndash $10 kg Age marital statusemploymentsmoking respiratorysymptoms heartdisease indicatorsdiastolic bloodpressure totalcholesterol
1middot88 1middot6 2middot2 ndash
WhitehallCohort
1190 40ndash69 ndash 4ndash9 kg 1middot26 1middot1 1middot5
Nilsson et al(2002)(55)
HealthySwedishmen (n 5194)
Weight stable(^0middot1 kgm2)
Weight lossunspecified(no directquestion)
464 47 years(38ndash52years)
22ndash25kgm2
ndash Age 1middot39 0middot981middot95
Cancer deathsdisease atbaseline
deathsin year 1 offollow-up
482 26 thornkgm2
1middot71 1middot182middot47
M
Harrin
gto
net
al
98
Nutrition Research Reviews
httpsww
wcam
bridgeorgcoreterms httpsdoiorg101017S0954422409990035
Dow
nloaded from httpsw
ww
cambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cam
bridge Core terms of use available at
relative risk (RR) of all-cause mortality(1314) Methodo-logical problems have also been identified for example themethod by which the weight loss was achieved has usuallynot been reported (although dietary energy restriction islikely to have been a major factor) while weight changesbefore and after the recording periods have usually not beendetermined(12 ndash 16)
In light of the current obesity epidemic and the resultingfocus on encouraging those with BMI above 25 kgm2 tolose weight by changing their diet and lifestyle(17) it isimportant to establish whether the long-term effects ofweight loss benefit life expectancy The current advice fromthe UK Department of Healthrsquos Obesity Care Programme isfor those who are overweight or obese to reduce energyintake and increase physical activity as a method of lifestylemodification(18) Further treatment and advice may need toconsider a broad spectrum of evidence so as not to rule outpotential investigations that identify subgroups of patientsor certain conditions where weight loss may be detrimentalto health and increase mortality(19)
The aim of the present study was to examine the availableevidence of the impact of weight loss as a lifestyleintervention on the RR of all-cause mortality and toquantify this using meta-analysis Data were pooled in anumber of different ways in order to examine the influenceof a number of possible confounders Meta-analysis wasused to provide a more objective appraisal of the evidenceintegrating data from multiple prospective cohort studies toincrease the power and precision of estimates of effect andreducing the likelihood of false negative results(2021)
Methods
Search strategy
A literature search was carried out independently by twoinvestigators to identify prospective cohort studies thatevaluated the effect of weight loss as a lifestyle interventionon mortality risk A web search was undertaken onPubMedMedline and ScienceDirect databases Articlespublished between 1987 and 2008 and in the Englishlanguage were included Search terms included lsquoweightBMI loss change mortality intentional unintentionalrelative risk prospective and cohortrsquo Identified citationsand abstracts were obtained from journals libraries orauthors A hand-search of the bibliographies of retrievedpapers and linked articles was also carried out
Data selection
Inclusion criteria were prospective studies in English ofadults (men andor women) with data on body weight andweight loss over more than 1 year Studies needed to presentRR of mortality and associated 95 CI for the group thatlost weight relative to a comparable reference group wholost minimal or no weight Drug treatment studies andstudies that measured weight loss following bariatricsurgery were excluded as the aim was to assess the effectof lifestyle interventions Twenty-six publications wereidentified that met the inclusion criteria Data on RR ofmortality and 95 confidence limits were extracted for allW
edic
ketal
(2002)
(56)
Calif
orn
ia
US
AW
eig
ht
sta
ble
(loss
10
lbs
or
gain
)
Weig
ht
loss
unspecifi
ed
71
years
atsta
rtof
12-y
ears
mort
alit
yfo
llow
-up
10
lbs
Age
curr
ent
and
past
sm
okin
g
exerc
ise
less
than
10
years
earlie
r
ndash
Healthy
men
628
26
1middot3
81middot0
61middot8
Healthy
wom
en
933
24
1middot7
61
middot33
2middot3
4D
iabetic
men
140
26
3middot6
62
middot15
6middot2
4D
iabetic
wom
en
90
25
1middot6
50middot7
3middot8
7In
tern
ational
weig
ht
loss
Tota
l
642
Healthy
men
1middot2
70
middot94
1middot7
1H
ealthy
Wom
en
1middot2
80middot9
5
1middot7
3
NH
AN
ES
N
ationalH
ealth
and
Nutr
itio
nE
xam
ination
Surv
ey
NID
DM
non-insulin
-dependent
dia
bete
sm
elli
tus
NH
IS
NationalH
ealth
Inte
rvie
wS
urv
ey
DO
M
Dia
gnostisch
Onderz
oek
Mam
macarc
inoom
(Dia
gnostic
Investigation
into
Bre
ast
Cancer)
IQ
R
inte
rquart
ilera
nge
1
lbfrac14
0middot4
536
kg
Weight loss and all-cause mortality 99
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
relative risk (RR) of all-cause mortality(1314) Methodo-logical problems have also been identified for example themethod by which the weight loss was achieved has usuallynot been reported (although dietary energy restriction islikely to have been a major factor) while weight changesbefore and after the recording periods have usually not beendetermined(12 ndash 16)
In light of the current obesity epidemic and the resultingfocus on encouraging those with BMI above 25 kgm2 tolose weight by changing their diet and lifestyle(17) it isimportant to establish whether the long-term effects ofweight loss benefit life expectancy The current advice fromthe UK Department of Healthrsquos Obesity Care Programme isfor those who are overweight or obese to reduce energyintake and increase physical activity as a method of lifestylemodification(18) Further treatment and advice may need toconsider a broad spectrum of evidence so as not to rule outpotential investigations that identify subgroups of patientsor certain conditions where weight loss may be detrimentalto health and increase mortality(19)
The aim of the present study was to examine the availableevidence of the impact of weight loss as a lifestyleintervention on the RR of all-cause mortality and toquantify this using meta-analysis Data were pooled in anumber of different ways in order to examine the influenceof a number of possible confounders Meta-analysis wasused to provide a more objective appraisal of the evidenceintegrating data from multiple prospective cohort studies toincrease the power and precision of estimates of effect andreducing the likelihood of false negative results(2021)
Methods
Search strategy
A literature search was carried out independently by twoinvestigators to identify prospective cohort studies thatevaluated the effect of weight loss as a lifestyle interventionon mortality risk A web search was undertaken onPubMedMedline and ScienceDirect databases Articlespublished between 1987 and 2008 and in the Englishlanguage were included Search terms included lsquoweightBMI loss change mortality intentional unintentionalrelative risk prospective and cohortrsquo Identified citationsand abstracts were obtained from journals libraries orauthors A hand-search of the bibliographies of retrievedpapers and linked articles was also carried out
Data selection
Inclusion criteria were prospective studies in English ofadults (men andor women) with data on body weight andweight loss over more than 1 year Studies needed to presentRR of mortality and associated 95 CI for the group thatlost weight relative to a comparable reference group wholost minimal or no weight Drug treatment studies andstudies that measured weight loss following bariatricsurgery were excluded as the aim was to assess the effectof lifestyle interventions Twenty-six publications wereidentified that met the inclusion criteria Data on RR ofmortality and 95 confidence limits were extracted for allW
edic
ketal
(2002)
(56)
Calif
orn
ia
US
AW
eig
ht
sta
ble
(loss
10
lbs
or
gain
)
Weig
ht
loss
unspecifi
ed
71
years
atsta
rtof
12-y
ears
mort
alit
yfo
llow
-up
10
lbs
Age
curr
ent
and
past
sm
okin
g
exerc
ise
less
than
10
years
earlie
r
ndash
Healthy
men
628
26
1middot3
81middot0
61middot8
Healthy
wom
en
933
24
1middot7
61
middot33
2middot3
4D
iabetic
men
140
26
3middot6
62
middot15
6middot2
4D
iabetic
wom
en
90
25
1middot6
50middot7
3middot8
7In
tern
ational
weig
ht
loss
Tota
l
642
Healthy
men
1middot2
70
middot94
1middot7
1H
ealthy
Wom
en
1middot2
80middot9
5
1middot7
3
NH
AN
ES
N
ationalH
ealth
and
Nutr
itio
nE
xam
ination
Surv
ey
NID
DM
non-insulin
-dependent
dia
bete
sm
elli
tus
NH
IS
NationalH
ealth
Inte
rvie
wS
urv
ey
DO
M
Dia
gnostisch
Onderz
oek
Mam
macarc
inoom
(Dia
gnostic
Investigation
into
Bre
ast
Cancer)
IQ
R
inte
rquart
ilera
nge
1
lbfrac14
0middot4
536
kg
Weight loss and all-cause mortality 99
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
subgroups presented by the authors (for example men andwomen intentional v unintentional weight loss obese voverweight)
Data analysis
Meta-analysis was performed using Comprehensive Meta-analysis software (CMA version 2 Biostat Inc EnglewoodNJ USA) Moderator variables such as baseline BMI(normal overweight obese) reason for weight loss(intentional unintentional) baseline health status (healthyunhealthy) method used to estimate weight loss (measuredweight loss reported weight loss) and physical activityadjustment (adjusted data unadjusted data) were used toclassify subgroups for separate analysis For the subgroupanalysis based on baseline BMI the ranges used in papers
generally corresponded to those recommended by WHO(22)Analysis was carried out using adjusted data because papersgave insufficient data on CI for unadjusted data Althoughmultivariable adjustment of the data varied from study tostudy all adjusted for smoking Results are shown in the formof schematic plots (Forest plots) which illustrate the size anddirection of effect for each study and the weighted effect ofall studies combined with 95 (lower and upper) CI Meta-analysis uses a weighted average of the results in which thelarger and more precise studies have more influence than thesmaller ones Results are shown for the random effectsmodel which assumes the underlying effect may vary foreach population This is the most appropriate model whereheterogeneity is present(2021) Statistical significance of theoverall pooled effect was based on P 0middot05
Fig 1 Mortality risk for intentional weight loss according to health status RR relative risk
M Harrington et al100
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Results
Study characteristics
Table 1 shows a summary of the characteristics of thestudy populations and subgroups Sample sizes rangedfrom 34 to 5008 subjects and the majority of the data wascollected from white populations of US and UK originAll of the studies were designed to investigate RR ofmortality and weight change The stage of life duringwhich weight change occurred varied between adulthoodmiddle age and old age and the follow-up period rangedfrom 2 to 20 years
Quantitative data synthesis
Owing to the acknowledged importance of whetherweight loss is intended or not results are presented for(a) intentional (b) unintentional and (c) weight loss notspecified For the main category of interest ieintentional weight loss sub-analyses are given for
healthy v unhealthy subjects These have then beenfurther analysed to examine the influence of moderatorsand confounders
Intentional weight loss
Figure 1 shows the RR of all-cause mortality in relation tointentional weight loss Overall there was no significanteffect (RR 1middot01 (95 CI 0middot93 1middot09) P frac14 0middot89) Howeveramong healthy subjects RR was increased 11 by weightloss (RR 1middot11 (95 CI 1middot00 1middot22) P frac14 0middot05) whereas itwas reduced in unhealthy subjects by a similar amount (RR0middot87 (95 CI 0middot77 0middot99) P frac14 0middot028)
Unintentional weight loss
Unintentional weight loss was associated with highermortality (RR 1middot22 (95 CI 1middot09 1middot37) P frac14 0middot001) (Fig 2)as has been shown in other studies Unintentional weightloss is usually considered an indicator of pre-existing orsilent disease and this group was not considered further
Fig 2 Mortality risk for unintentional weight loss according to health status RR relative risk
Weight loss and all-cause mortality 101
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Fig 3 Mortality risk for weight loss (intention unknown) according to health status RR relative risk
M Harrington et al102
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
Unknown or unspecified cause of weight loss
Where the cause of weight loss was unspecified there wasalso excess mortality (RR 1middot39 (95 CI 1middot29 1middot51)P 0middot001) (Fig 3) Most of these studies were on lsquohealthyrsquosubjects but the subgroup who were unhealthy had evenhigher mortality associated with weight loss (RR 1middot75 (95 CI 1middot24 2middot46) P frac14 0middot001) Studies where weight lossintention was not explored may suffer from the sameproblem of confounding by illness as those in which weightloss was unintentional The remaining analyses were allperformed using studies of intentional weight loss only
Subgroup analyses of intentional weight loss
Relative weight at baseline Weight loss appeared tobenefit obese weight losers who were also classified asunhealthy at baseline (RR 0middot84 (95 CI 0middot73 0middot97)P frac14 0middot018) but had no benefit for healthy obese (RR 1middot02)Overall there was no change in risk for the obese group (RR0middot94 (95 CI 0middot86 1middot04) P frac14 0middot002) (Fig 4) Forintentional weight losers whose baseline BMI was withinthe normal to overweight range or for mixed-weightpopulations the RR of mortality was increased (RR 1middot09(95 CI 1middot02 1middot17) P frac14 0middot008) (Fig 5)
Method of assessing weight loss The majority of studygroups with data on intentional weight loss (fifteen out ofeighteen studies) relied on reported measurements of weightor weight loss Among these RR associated with weightloss was near unity However the three study groups withactual measurement had a net RR of 1middot28 (95 CI 1middot071middot53) (Fig 6)
Physical activity adjustment Adjustment for physicalactivity was made in most studies (fourteen out of eighteenstudies) but there was essentially no difference in the RRaccording to whether the models had adjusted for activity ornot (RR 0middot98 v 1middot01 where adjusted for physical activity)(Fig 7)
Discussion
Main findings
Meta-analysis was used to explore the effect of weight losson mortality using sensitivity and subgroup analysis toexplore some of the likely causes of heterogeneityespecially intentionality health and baseline BMI Whereasweight loss for unknown or unspecified reasons was clearlyassociated with excess mortality intentional weight loss
Fig 4 Mortality risk for intentional weight loss among obese adults RR relative risk
Weight loss and all-cause mortality 103
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
resulted in virtually no change in mortality overallImportantly we found opposing effects among healthyand unhealthy adults and between the obese and those withmore moderate degrees of overweight or from the generalpopulation The excess risk of weight loss in healthy adultswas estimated to be of the order of 11 This wascounterbalanced by a benefit of about 13 amongunhealthy adults (ie those with diabetes or obesity-relatedhealth conditions)
Other studies
The literature is equivocal on the risks and benefits ofweight loss(1516) Many prospective studies and reviewsappear to show an increased mortality associated withweight loss(12) which runs counter to conventional wisdomrelating to the adverse effects of obesity and the beneficialchanges in risk factors associated with weight loss(23) It hasbeen argued that methodological weaknesses explain muchof this paradox including failure to adjust for knownconfounders(24) In particular it has been claimed thatintentionality of weight loss is key(16) but many studies failto distinguish between intentional and unintentional weightloss the latter being a cardinal sign of ill health and apredictor of increased mortality in old age(2526)
Some clinical trials have demonstrated beneficial effectsof weight loss with regard to morbidity in individualssuffering from either diabetes obesity-related healthconditions cancer or other diseases(3) There are also anincreasing number of favourable reports from bariatricsurgery such as the lsquoSwedish obese subjectsrsquo (SOS) studywhich has shown that substantial long-term weight
reduction appreciably improves the cardiovascular riskprofile of morbidly obese subjects ultimately resulting in adecrease in overall mortality(12) Although such data may beencouraging their success cannot necessarily be extrapo-lated to the public health setting where the weight lossesnormally achieved by diet are modest and difficult tosustain and the subjects generally less severely obese andwith few co-morbidities Another study due to report in2015 will provide valuable additional data This is the LookAHEAD (Action For Health in Diabetes) clinical trialwhich is assessing the long-term effects (up to 11middot5 years) ofan intensive weight-loss programme delivered over 4 yearsin overweight and obese individuals with type 2 diabetes
Interpretation of present analysis
In the present review and meta-analysis intentional weightloss modestly reduced the risk of all-cause mortality onlyamong the subgroup of unhealthy adults (by approximately13 ) especially among those who were also obese (byapproximately 16 ) All these studies relied on reportedestimates of body weight Self-reporting of body weightmay be cheap and easily carried out however it is affectedby a number of biases Actual measurement of body weightusing appropriate devices is recommended for completeaccuracy and reliability of the data(27)
Our finding of a marginally increased risk of death amongoverweight but otherwise healthy adults who lost weightintentionally if true has important public health impli-cations This observation is consistent with recent findingsusing National Health and Nutrition Examination Survey
Fig 5 Mortality risk for intentional weight loss among overweight or mixed populations RR relative risk
M Harrington et al104
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
(NHANES) data that showed that the ideal weight forlongevity was the overweight category or BMI 23ndash30 kgm2(28 ndash 30)
Why should intentional weight loss have opposing effectsin different groups of individuals One possibility is thatobese individuals with risk factors may show a benefitbecause they are more motivated to make a series of changessuch as reducing fat intake or increasing exercise level andthese may lower RR of mortality by benefiting overall healthstatus(31) Unhealthy individuals are also more likely to berecipients of health care and medical interventions It ismore difficult to explain why intentional weight loss shouldhave an adverse effect among healthy but overweightindividuals More data on method of weight losspersistence of weight loss and body composition would behelpful in this regard Weight loss via energy restriction maydo little to alter the relative distribution of body fat and mayresult in decreased lean body mass A reanalysis of theFramingham Heart Study and the Tecumseh CommunityStudy suggests that weight loss as a result of a reduction inbody fat may reduce all-cause mortality while weight loss asa result of a reduction in lean body mass may increase it(32)Given the significance of fat distribution and the lean bodymassfat ratio in health prognosis(33) it is imperative thatfuture studies attempt to measure more than just weight orBMI Furthermore studies must adequately disentangle the
influence of physical activity andor fitness which mayinfluence both body weight and the morbidity and mortalityoutcomes under study Most studies did not includeassessment of physical activity and those that did usedquestionnaires rather than physical fitness which is astronger predictor of mortality(3435) The focus of newresearch may most usefully be directed to examiningsurvival among those population groups that might beexpected to benefit most from weight loss These includethose with diabetes those with obesity-related conditions(such as hypertension) and certain ethnic groups
Limitations
The present study inevitably has some limitations Theliterature search was carried out using only two databasesbut was complemented by thorough checking of cross-references and inclusion of new reviews published in2008 Limitations of the evidence base include the factthat none of the studies provided information on themethod of weight loss which is relevant because it is notclear if weight loss through energy restriction or increasedenergy expenditure differentially influences long-termoutcomes Second weight loss was usually assessedretrospectively and subjectively often at two time pointssome distance removed from the ultimate outcome ie
Fig 6 Mortality risk for intentional weight loss according to weight loss assessment method RR relative risk
Weight loss and all-cause mortality 105
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
death It is thus difficult to be sure that the weight lossestimate does not represent a transitory phase and that it isrepresentative of a reasonable period of adult life Thirdthe studies differed in the statistical treatment ofcovariates or confounders in adjusted models (forexample some excluded smokers others adjusted forsmoking) These problems are common to all attempts toreview and pool data from different studies and thepresent results are consistent with other recent reviewsthat have not used meta-analysis(1636) Furthermore usinga meta-analysis stratified by intentionality health andbaseline BMI we were able to quantify effect sizes indifferent groups The robustness of intentionality measureshas been questioned(16) because it depends on the questionasked and may change during the course of the follow-up(37) The study by Soslashrensen et al (8) was unusual inassessing intentionality prospectively and also reported thelargest effect size (RR 1middot87)(8) However as it was of highquality (as judged by Simonsen et al (16)) we did notconsider its exclusion justified in the main analysisInstead sensitivity analysis showed that the effect ofexcluding this paper would be to reduce the RR from 1middot11to 1middot09 On balance we think it unlikely that our estimatesof higher risk are inflated since most sources ofmisclassification and measurement error would tend toresult in underestimation of effect (for example self-reported body weight)
Conclusion
Recently a great emphasis has been placed on weight lossby lifestyle change for everyone who is even slightlyoverweight However a review of the available literaturecomplemented by meta-analysis suggests that at-riskindividuals may benefit but for healthy overweightindividuals intentional weight loss does not decreasemortality and may even increase it Appropriatelydesigned intervention studies in subgroups differing byage sex and ethnic group as well as by risk status areurgently needed Until more reliable data are available todemonstrate consistent improvements in survival thequestion remains as to whether the correction of obesityper se should have such emphasis as a clinical and publichealth target
Acknowledgements
We wish to thank Orlaith McDaid for her contribution to thepreparation of this paper
The present review was funded by The World SugarResearch Organisation The findings and conclusions arethose of the authors and do not necessarily represent theviews of the funding organisation
The authors have no conflicts of interest to declare
Fig 7 Mortality risk for intentional weight loss according to adjustment for physical activity RR relative risk
M Harrington et al106
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
References
1 Lee IM amp Paffenbarger RS Jr (1992) Change in body weightand longevity JAMA 268 2045ndash2049
2 Van Gaal LF Wauters MA amp De Leeuw IH (1997) Thebeneficial effects of modest weight loss on cardiovascularrisk factors Int J Obes Relat Metab Disord 21 Suppl 1S5ndashS9
3 Goldstein DJ (1992) Beneficial health effects of modestweight loss Int J Obes Relat Metab Disord 16 397ndash415
4 Williamson DF Thompson TJ Thun M et al (2000)Intentional weight loss and mortality among overweightindividuals with diabetes Diabetes Care 23 1499ndash1504
5 Sjostrom L Narbro K Sjostrom CD et al (2007) Effects ofbariatric surgery on mortality in Swedish obese subjectsN Engl J Med 357 741ndash752
6 Diaz VA Mainous AG III amp Everett CJ (2005) Theassociation between weight fluctuation and mortality resultsfrom a population-based cohort study J Community Health30 153ndash165
7 Droslashyvold WB Lund Nilsen TI Lydersen S et al (2005)Weight change and mortality the Nord-Troslashndelag HealthStudy J Intern Med 257 338ndash345
8 Soslashrensen TI Rissanen A Korkeila M et al (2005)Intention to lose weight weight changes and 18-ymortality in overweight individuals without co-morbiditiesPLoS Med 2 e171
9 Sauvaget C Ramadas K Thomas G et al (2008) Body massindex weight change and mortality risk in a prospectivestudy in India Int J Epidemiol 37 990ndash1004
10 Wannamethee SG Shaper AG amp Lennon L (2005) Reasonsfor intentional weight loss unintentional weight loss andmortality in older men Arch Intern Med 165 1035ndash1040
11 Eilat-Adar S Goldbourt U Resnick HE et al (2005)Intentional weight loss blood lipids and coronary morbidityand mortality Curr Opin Lipidol 16 5ndash9
12 Nilsson PM (2008) Is weight loss beneficial for reduction ofmorbidity and mortality What is the controversy aboutDiabetes Care 31 Suppl 2 S278ndashS283
13 Soslashrensen TI (2003) Weight loss causes increased mortalitypros Obes Rev 4 3ndash7
14 Yang D Fontaine KR Wang C et al (2003) Weight losscauses increased mortality cons Obes Rev 4 9ndash16
15 Poobalan AS Aucott LS Smith WC et al (2007) Long-termweight loss effects on all cause mortality in overweightobesepopulations Obes Rev 8 503ndash513
16 Simonsen MK Hundrup YA Obel EB et al (2008)Intentional weight loss and mortality among initially healthymen and women Nutr Rev 66 375ndash386
17 World Health Organization amp Food and AgricultureOrganization (2003) Diet Nutrition and the Prevention ofChronic Diseases Geneva WHO
18 Department of Health (2006) Your Weight Your HealthLondon Department of Health Central Office of InformationhttpwwwdhgovukenPublicationsandstatisticsPublica-tionsPublicationsPolicyAndGuidanceDH_4134408
19 Astrup A (2003) Weight loss and increased mortalityepidemiologists blinded by observations Obes Rev 4 1ndash2
20 Egger M amp Smith GD (1997) Meta-analysis potentials andpromise BMJ 315 1371ndash1374
21 Egger M Smith GD amp Phillips AN (1997) Meta-analysisprinciples and procedures BMJ 315 1533ndash1537
22 World Health Organization (2000) Obesity Preventing andManaging the Global Epidemic Geneva WHO
23 Heitmann BL Svendsen OL Martinussen T et al (1997)Significance of intentional weight loss on health (article inDanish) Ugeskr Laeger 159 4099ndash4104
24 Stampfer M (2005) Weight loss and mortality what does theevidence show PLoS Med 2 e181
25 Ryan C Bryant E Eleazer P et al (1995) Unintentionalweight loss in long-term care predictor of mortality in theelderly South Med J 88 721ndash724
26 Shahar A Shahar D Kahar Y et al (2005) Low-weight andweight loss as predictors of morbidity and mortality in oldage (article in Hebrew) Harefuah 144 443ndash448 452
27 John U Hanke M Grothues J et al (2006) Validity ofoverweight and obesity in a nation based on self-report versusmeasurement device data Eur J Clin Nutr 60 372ndash377
28 Fontaine KR Redden DT Wang C et al (2003) Years of lifelost due to obesity JAMA 289 187ndash193
29 Flegal KM Graubard BI Williamson DF et al (2005)Excess deaths associated with underweight overweight andobesity JAMA 293 1861ndash1867
30 Flegal KM Graubard BI Williamson DF et al (2007)Cause-specific excess deaths associated with underweightoverweight and obesity JAMA 298 2028ndash2037
31 Gregg EW Gerzoff RB Thompson TJ et al (2003)Intentional weight loss and death in overweight and obeseUS adults 35 years of age and older Ann Intern Med 138383ndash389
32 Allison DB Zannolli R Faith MS et al (1999) Weight lossincreases and fat loss decreases all-cause mortality rateresults from two independent cohort studies Int J Obes RelatMetab Disord 23 603ndash611
33 Berentzen T amp Soslashrensen TI (2006) Effects of intended weightloss on morbidity and mortality possible explanations ofcontroversial results Nutr Rev 64 502ndash507
34 Warburton DE Nicol CW amp Bredin SS (2006) Healthbenefits of physical activity the evidence CMAJ 174801ndash809
35 Myers J Kaykha A George S et al (2004) Fitness versusphysical activity patterns in predicting mortality in menAm J Med 117 912ndash918
36 Fontaine KR amp Allison DB (2001) Does intentional weightloss affect mortality rate Eat Behav 2 87ndash95
37 Coffey CS Gadbury GL Fontaine KR et al (2005) Theeffects of intentional weight loss as a latent variable problemStat Med 24 941ndash954
38 Harris T Cook EF Garrison R et al (1988) Body mass indexand mortality among nonsmoking older persons TheFramingham Heart Study JAMA 259 1520ndash1524
39 Pamuk ER Williamson DF Madans J et al (1992) Weightloss and mortality in a national cohort of adults 1971ndash1987Am J Epidemiol 136 686ndash697
40 Higgins M DrsquoAgostino R Kannel W et al (1993) Benefitsand adverse effects of weight loss Observations from theFramingham Study Ann Intern Med 119 758ndash763
41 Chaturvedi N amp Fuller JH (1995) Mortality risk by bodyweight and weight change in people with NIDDM The WHOMultinational Study of Vascular Disease in DiabetesDiabetes Care 18 766ndash774
42 Iribarren C Sharp DS Burchfiel CM et al (1995)Association of weight loss and weight fluctuation withmortality among Japanese American men N Engl J Med333 686ndash692
43 Manson JE Willett WC Stampfer MJ et al (1995) Bodyweight and mortality among women N Engl J Med 333677ndash685
44 Wallace JI Schwartz RS LaCroix AZ et al (1995)Involuntary weight loss in older outpatients incidence andclinical significance J Am Geriatr Soc 43 329ndash337
45 Williamson DF Pamuk E Thun M et al (1995)Prospective study of intentional weight loss and mortality
Weight loss and all-cause mortality 107
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at
in never-smoking overweight US white women aged 40ndash64years Am J Epidemiol 141 1128ndash1141
46 Yaari S amp Goldbourt U (1998) Voluntary and involuntaryweight loss associations with long term mortality in 9228middle-aged and elderly men Am J Epidemiol 148546ndash555
47 French SA Folsom AR Jeffery RW et al (1999) Prospectivestudy of intentionality of weight loss and mortality in olderwomen the Iowa Womenrsquos Health Study Am J Epidemiol149 504ndash514
48 Williamson DF Pamuk E Thun M et al (1999) Prospectivestudy of intentional weight loss and mortality in overweightwhite men aged 40ndash64 years Am J Epidemiol 149 491ndash503
49 Newman AB Yanez D Harris T et al (2001) Weight changein old age and its association with mortality J Am Geriatr Soc49 1309ndash1318
50 Wannamethee SG Shaper AG amp Walker M (2002) Weightchange weight fluctuation and mortality Arch Intern Med162 2575ndash2580
51 Gregg EW Gerzoff RB Thompson TJ et al (2004) Trying tolose weight losing weight and 9-year mortality in
overweight US adults with diabetes Diabetes Care 27657ndash662
52 Maru S van der Schouw YT Gimbrere CH et al (2004)Body mass index and short-term weight change in relation tomortality in Dutch women after age 50 y Am J Clin Nutr 80231ndash236
53 Elliott AM Aucott LS Hannaford PC et al (2005) Weightchange in adult life and health outcomes Obes Res 131784ndash1792
54 Breeze E Clarke R Shipley MJ et al (2006) Cause-specificmortality in old age in relation to body mass index in middleage and in old age follow-up of the Whitehall cohort of malecivil servants Int J Epidemiol 35 169ndash178
55 Nilsson PM Nilsson JA Hedblad B et al (2002) The enigmaof increased non-cancer mortality after weight loss in healthymen who are overweight or obese J Intern Med 252 70ndash78
56 Wedick NM Barrett-Connor E Knoke JD et al (2002) Therelationship between weight loss and all-cause mortality inolder men and women with and without diabetes mellitus theRancho Bernado study J Am Geriatr Soc 50 1810ndash1815
M Harrington et al108
NutritionResearchReviews
httpswwwcambridgeorgcoreterms httpsdoiorg101017S0954422409990035Downloaded from httpswwwcambridgeorgcore IP address 541914080 on 15 Apr 2017 at 214438 subject to the Cambridge Core terms of use available at