exercise intensity and longevity in men

Exercise Intensity and Longevity in MenThe Harvard Alumni Health StudyI-Min Lee, MBBS, ScD; Chung-cheng Hsieh, ScD; Ralph S. Paffenbarger, Jr, MD, DrPH

Objective.\p=m-\Toexamine the independent associations of vigorous (\m=ge\6restingmetabolic rate [MET] score) and nonvigorous (<6 MET score) physical activity withlongevity.

Design.\p=m-\Prospectivecohort study, following up men from 1962 or 1966 through1988.

Setting/Participants.\p=m-\Subjectswere Harvard University alumni, without self\x=req-\reported, physician-diagnosed cardiovascular disease, cancer, or chronic obstruc-tive pulmonary disease (n=17 321). Men with a mean age of 46 years reported theirphysical activities on questionnaires at baseline.

Main Outcome Measure.\p=m-\All-causemortality (3728 deaths).Results.\p=m-\Totalenergy expenditure and energy expenditure from vigorous ac-

tivities, but not energy expenditure from nonvigorous activities, related inversely tomortality. After adjustment for potential confounders, the relative risks of dying as-sociated with increasing quintiles of total energy expenditure were 1.00 (referent),0.94, 0.95, 0.91 and 0.91, respectively (P [trend]<.05). The relative risks of dyingassociated with less than 630,630 to less than 1680,1680 to less than 3150,3150to less than 6300, and 6300 or more kJ/wk expended on vigorous activities were1.00 (referent), 0.88, 0.92, 0.87, and 0.87, respectively (P [trend]=.007). Corre-sponding relative risks for energy expended on nonvigorous activities were 1.00(referent), 0.89,1.00,0.98, and 0.92, respectively (P [trend]=.36). Analyses of vig-orous and nonvigorous activities were mutually adjusted. Among men who reportedonly vigorous activities (259 deaths), we observed decreasing age-standardizedmortality rates with increasing activity (P=.05); among men who reported only non-vigorous activities (380 deaths), no trend was apparent (P=.99).

Conclusions.\p=m-\Thesedata demonstrate a graded inverse relationship betweentotal physical activity and mortality. Furthermore, vigorous activities but not nonvig-orous activities were associated with longevity. These findings pertain only to all-cause mortality; nonvigorous exercise has been shown to benefit other aspects ofhealth.

(JAMA. 1995;273:1179-1184)

FEW PHYSICIANS, if any, would dis¬pute that physical activity enhanceshealth. Among other benefits, increasedactivity is associated with decreased in¬cidence of coronary heart disease,1 hy¬pertension,2 non-insulin-dependent dia-

From the Department of Epidemiology, HarvardSchool of Public Health, Boston, Mass (Drs Lee, Hsieh,and Paffenbarger); Division of Preventive Medicine,Department of Medicine, Brigham and Women's Hos-pital and Harvard Medical School, Boston (Dr Lee); andDivision of Epidemiology, Stanford (Calif) UniversitySchool of Medicine (Dr Paffenbarger).

Reprint requests to Department of Epidemiology,Harvard School of Public Health, 677 Huntington Ave,Boston, MA 02115 (Dr Lee).

betes mellitus,3 and colon cancer,4 andincreased longevity.6 What is uncertain,however, are the kinds and intensity ofphysical activity that should be pre¬scribed for health. In England, Morriset al6,7 found that to reduce coronaryheart disease risk, moderately vigorousexercise is necessary. Similarly, in a re¬cent study from Finland, Lakka et al8reported that only more intense, condi¬tioning physical activity reduces the riskof myocardial infarction; less intense,noneonditioning activities have no ef¬fect. Yet others maintain that as long astotal energy output—even if amassed

from light or moderate exercise—isincreased, risks of coronary heart di¬sease1,913 and premature mortality fromany cause5,9"11,13,14 are decreased. Forexample, in the Multiple Risk FactorIntervention Trial.9 an inverse rela¬tionship between leisure time physicalactivity and risk of coronary heart dis¬ease and total mortality was observedamong men engaged predominantly inlight to moderate activities. Dutch menand women who habitually carried outlight exercise (leisure-time walking, cy¬cling, and gardening) also enjoyed de¬creased coronary heart disease risk.15

Despite this lack ofconsensus and thefact that few studies have compared di¬rectly the relative merits of vigorousand nonvigorous exercise, a commonlyprescribed exercise regimen borrowsfrom recommendations for developingand maintaining cardiorespiratory fit¬ness. These regimens prescribe exer¬cise intense enough to produce sweat¬ing or hard breathing (60% to 90% ofmaximum heart rate), for at least 20minutes, three times per week.16,17 Re¬cently, the Centers for Disease Controland Prevention and the American Col¬lege of Sports Medicine issued a new,less stringent recommendation18: "Ev¬ery US adult should accumulate 30minutes or more of moderate-intensityphysical activity on most, preferably all,days of the week." This recommenda¬tion was meant to encourage more ex¬ercise among the almost 60% of USadults who engage in little or no leisure-time activity.19

Which is the more valid stance? Toprovide further information—at least, fortotal mortality—we investigated the rela¬tive merits of vigorous and nonvigorousexercise and their associations with pre¬mature mortality. Is, for example, 2100kJ (500 kcal) expended in vigorous exer¬cise associated with the same decrease inmortality risk as an equal amount ex¬

pended in nonvigorous activity?

at University of Colorado Denver HSL on August 17, 2009 www.jama.comDownloaded from

http://jama.ama-assn.org

SUBJECTS AND METHODSStudy Subjects

The Harvard Alumni Health Study isan ongoing cohort study that searchesfor predictors of chronic disease in menwho matriculated as undergraduates atHarvard University, Boston, Mass, be¬tween 1916 and 1950. The cohort wasestablished when 21582 alumni (68% re¬

sponse) returned a mailed questionnaireon medical history and health practicesin either 1962 or 1966. Eligible subjectsfor our study were men who reported no

physician-diagnosed cardiovascular dis¬ease, cancer, or chronic obstructive pul¬monary disease. Additionally, subjectshad to provide data on physical activity,body weight, height, cigarette habit, phy¬sician diagnosis ofhypertension and dia¬betes mellitus, vital status of both par¬ents, and, ifparents were deceased, ageat parental death. Alumni who satisfiedthese criteria numbered 17 321.

Assessment of Physical Activityand Other Predictors of Mortality

We assessed physical activity by ask¬ing alumni about flights ofstairs climbed,city blocks walked, types of sports orrecreational activities engaged in, andthe time (hours per week) spent on eachof these sports and recreational activi¬ties.20 Although we did not ask specifi¬callyabout occupational activity, alumniwere unlikely to have expended muchenergy on the job apart from walkingand climbing stairs. Physical activity as¬sessment was validated by comparisonof estimates of energy expenditure ob¬tained from the questionnaire againstthe following: (1) other widely acceptedand used physical activity question¬naires, (2) physiologic variables knownto be influenced by physical activity, (3)total energy intake, (4) physical activitydiaries, and (5) mechanical devices thatmeasure bodily movement. On the basisof these criteria, this physical activityquestionnaire has been shown to be re¬liable and valid.2127

Climbing up and down one flight ofstairs daily rated 118 kJ/wk, and walk¬ing 1 block daily, 235 kJ/wk. We assigneda multiple of resting metabolic rate(MET score) to every activity.28 Sinceresting metabolic rate is approximately4.2 kJ/kg of body weight per hour, we

estimated the energy expended on eachactivity by multiplying its MET score

by 4.2, body weight in kilograms andhours per week ofparticipation. We thensummed kilojoules per week from flightsclimbed, blocks walked, and activitiesperformed, to provide an index of totalenergy expenditure per week.

We were interested in two compo¬nents of total energy expenditure: that

derived from vigorous activities (requir¬ing >6 METs)29 and that from nonvig¬orous (ie, light and moderate) activities(requiring <6 METs).29 (Examples ofvigorous activities reported by alumniinclude walking briskly, running or jog¬ging, swimming laps, playing tennis, andshoveling snow.) Thus, for each alum¬nus, we further estimated energy ex¬

penditure from vigorous and nonvigor¬ous activities separately.

Using alumni self-reports, we obtainedinformation on other predictors of mor¬

tality: Quetelet's index (weight in kilo¬grams divided by the square ofheight inmeters), cigarette habit, physician di¬agnosis ofhypertension or diabetes mel-litus, and early parental death (definedas death occurring before age 65 years).Ascertainment of Mortality

The Harvard Alumni Office maintainsweekly rosters ofdeceased alumni. Theirrecords indicate mortality follow-up dataare unavailable for fewer than 1% ofalumni.30 Using information from theAlumni Office, we obtained copies ofdeath certificates. The end point of in¬terest for the present study was all-causemortality occurring after return of the1962 or 1966 questionnaire through 1988.

Statistical AnalysisWe first compared age-standardized

mortality rates, by means of the indi¬rect method,31 for increments of total,vigorous, and nonvigorous energy ex¬

penditure. To make statistical adjust¬ment for several potential confounderssimultaneously, we proceeded to use pro¬portional hazards regression to analyzetime to mortality or censoring.32 Cumu¬lative hazard plots disclosed no grossviolation ofproportional hazard assump-

tions. Mortality rate ratios (relativerisks) were modeled as a function ofphysical activity. For total energy ex¬

penditure, we categorized alumni intoquintiles (0 to <2524, 2524 to <4738,4738 to <8001, 8001 to <13142, and>13142 kJ/wk). When investigating theindependent associations ofvigorous andnonvigorous energy expenditure withlongevity, instead of using terms for to¬tal energy expenditure, we includedterms for its two components. We cat¬egorized alumni according to five groupseach of vigorous (<630, 630 to <1680,1680 to <3150,3150 to <6300, and >6300kJ/wk) and nonvigorous (same cutoffpoints) energy expenditure. Potentialconfounders included in regression mod¬els were age (single years), Quetelet'sindex (<22.5,22.5 to <23.5,23.5 to <24.5,24.5 to <26.0, or >26.0 kg/m2), cigarettehabit (never, former, or current smoker),and physician-diagnosed hypertensionor diabetes mellitus (no vs yes). Althoughearly parental death (neither, one, orboth parents dying early) was unlikelyto be related to physical activity andthus was not a confounder, it also wasincluded since we were interested in itsinfluence on mortality.

We tried to avoid observing an arti-factual association between physicalinactivity and increased mortality byexcluding alumni with cardiovascular dis¬ease, cancer, and chronic obstructive pul¬monary disease from the starting popu¬lation. To minimize further the potentialimpact of this bias, we conducted addi¬tional analyses that omitted the first 5years (arbitrarily chosen) of follow-up,because mortality among individuals withother illnesses that could limit physicalactivity, and also increase mortality risk,would likely occur early in follow-up.

Table 1.—Characteristics of Harvard Alumni in 1962 or 1966, According to Quintiles of Total EnergyExpenditure*

Characteristic

Quintile of Total Energy ExpenditureI

1

Mean age, y 46.8 46.1 45.8 45.5 46.4

Mean Quetelet's Index, kg/m2 24.65 24.31 24.14 24.19 24.47

Cigarette habit, %Never smokers 20.1 21.9 20.9 21.1 20.2Former smokers 32.1 30.4 32.0 31.9 33.5

Current smokers 47.8 47.7 47.1 47.0 46.3

Mean No./d of cigarettes smoked by current smokers 27 27 26 26 26

% reporting physician-diagnosed hypertension or diabetes mellltus 12.0 10.2 9.2 10.1 9.6

Early parental death (<65 y), %Neither parent dying early 66.9 65.4 64.6 66.3 67.61 parent dying early 29.0 30.7 30.6Both parents dying early 4.1 3.9 4.8 4.1 4.4

% of total energy expenditure expended in vigorous activltiest 41.7 32.1 32.2 38.9 43.5

*AII characteristics (except mean age) were age standardized. Energy expenditure was estimated in kilojoulesper week from climbing stairs, walking, and participating in sports or recreational activities. Quintile 1 represents thelowest energy expenditure.

tAmong alumni who did expend energy in climbing stairs, walking, or participating in sports or recreational activities(17 090 of 17 321 alumni). Vigorous activities were defined as those that required 6 or more units of resting metabolicrate (METs).



Ninety-five percent confidence inter¬vals were calculated for estimated rela¬tive risks, and all values were fromtwo-tailed tests. To assess whether vig¬orous and nonvigorous energy expen¬diture differed significantly in their in¬dependent associations with longevity,we used the covariance matrix to esti¬mate variance for the difference betweenthe two parameter estimates.33RESULTS

Table 1 describes alumni characteris¬tics at study entry by quintiles of totalenergy expenditure. The mean age ofalumni was 46 years, with little variationacross quintiles. The mean Quetelet's in¬dex decreased as energy expenditure in¬creased from the lowest to the third cat¬egory, then increased beyond that. Al¬most halfofthe alumni smoked cigarettesin 1962 or 1966; the proportion whosmoked declined steadily with increasingenergy expenditure. Approximately 10%of alumni declared a physician diagnosisofhypertensionordiabetes mellitus,whileapproximately 30% reported that one orboth parents had died early. Table 1 alsopresents, for alumni who declared someform of physical activity (n=17090), theproportion of total energy expenditurederived from vigorous activities.

We then examined age-standardizedmortality rates by level of energy ex¬penditure (Table 2). Between 1962 or1966 and 1988, a total of 3728 deathsoccurred in 384 681 person-years of ob¬servation. Mortality generally declinedwith increasing total energy expendi¬ture (P=.001). At about 14700 kJ/wk oftotal energy expenditure, mortality ap¬peared to stabilize. Mortality also de¬clined with higher levels of vigorous en¬

ergy expenditure, regardless ofthe levelof nonvigorous activity, up to 12 600 kJ/wk. Beyond this level of vigorous en¬

ergy expenditure, mortality increasedslightly. However, the overall inverseassociation was significant (P<.001). Onthe other hand, nonvigorous energy ex¬penditure, regardless of the level ofvig¬orous activity, was not associated withmortality (P=.87).

To consider other potential confound-ers, we proceeded to multivariate analy¬ses, adjustingadditionally forQuetelet'sindex, smoking, hypertension, diabetesmellitus, and early parental death. Totalenergy expenditure, in quintiles, con¬tinued to relate significantly and in¬versely to mortality. The adjusted rela¬tive risks were 1.00 (referent), 0.94 (95%confidence interval, 0.86 to 1.04), 0.95(0.86 to 1.05), 0.91 (0.83 to 1.01), and 0.91(0.82 to 1.00), respectively (P fortrend<.046).

With the findings in Table 2, we couldnot determine whether vigorous physi-

Table 2.—Age-Standardized Mortality Rates Among Harvarding to Energy Expended on All, Vigorous, and Nonvigorous

Alumni, 1962 or 1966 Through 1988, Accord-Activities in 1962 or 1966*

Physical Activityin 1962 or 1966,

kJ/wkNo. ofDeaths Person-Years

Age-StandardizedMortality Rate/10000

Energy expended on all activities<2100 700 65241 103.782100-<4200 850 82859 102.554200-<6300 553 55670 99.796300-<8400 360 41005 92.308400-<10 500 292 32 665 92.8810 500-<12600 222 26 347 88.3512600-<14700 182 20421 88.05==14700 569 60473 90.76 for trend <.001

Energy expended on vigorous activities<2100 2730 250181 102.252100-<4200 401 44373 95.904200-<6300 158 24590 74.266300-<8400 107 17098 76.948400-<10 500 104 14222 89.9710500-<12 600 50 9389 62.0012 600--C14700 43 6564 79.292:14700 135 18264 86.23 for trend <.001

Energy expended on nonvigorous activities<2100 1276 135 930 97.812100-<4200 855 94 574 94.924200-<6300 590 59 321 101.636300-< 8400 245 27100 90.408400-<10 500 203 19620 96.9810 500-<12 600 152 14779 90.3312600-<14 700 101 9842 89.43==14700 306 23515 102.22 for trend .87

*Energy expenditure was estimated from climbing stairs, walking, and participating in sports or recreationalactivities. Vigorous activities were defined as those that required 6 or more units of resting metabolic rate (METs);nonvigorous activities, those that required less than 6 METs.

cal activity, nonvigorous physical activ¬ity, or both was responsible for the in¬verse relationship with mortality. Wethus conducted multivariate analyses toadjust mutually for vigorous and non-vigorous energy expenditure, while si¬multaneously adjusting for potential con-founders (Table 3). Vigorous energy ex¬

penditure again was significantly andinversely related to mortality (P=.007),whereas the trend for nonvigorousenergy expenditure again was notsignificant (P=.36). However, relativerisk estimates for vigorous energyexpenditure did not differ significantlyfrom corresponding estimates for non-vigorous energy expenditure.

To minimize potential bias from illhealth in the starting population, we con¬ducted additional analyses that omittedthe first 5 years after physical activityassessment (see "StatisticalAnalysis" sec¬tion). We analyzed 3297 deaths (Table 4).Vigorous energy expenditure remainedsignificantly and inversely related to mor¬

tality (P=.007), but we again observed nosignificant trend with nonvigorous energyexpenditure (P=.32). Relative risk esti-

mates for the two kinds of energy ex¬

penditure differed significantly at 1680kJ/wk and higher. Trends for the twotypes ofenergy expenditure also differedsignificantly (P=.02).

We next tried to account for changesin physical activity over time. In 1977,we had sent another questionnaire tosurvivingalumni that requestedupdatedinformation on medical history andhealth habits, including physical activ¬ity. Seventy-six percent of survivingalumni responded. We conducted fur¬ther analyses that updated physical ac¬

tivity in 1977 for alumni who returned aquestionnaire that year and who con¬tinued to be free of self-reported, phy¬sician-diagnosed cardiovascular disease,cancer, or chronic obstructive pulmo¬nary disease. We also built in a 5-yearlag after either physical activity assess¬ment. Relative risks ofmortality for thesame five categories of vigorous energyexpenditure were 1.00 (referent), 0.89(95% confidence interval, 0.79 to 1.00),0.83 (0.71 to 0.97), 0.76 (0.63 to 0.91), and0.75 (0.64 to 0.89), respectively (P fortrend=.001). Corresponding relative



Table 3.—Relative Risks of All-Cause Mortality Among Harvard Alumni, 1962 or 1966 Through 1988,According to Vigorous and Nonvigorous Physical Activity in 1962 or 1966*

Kind of Activity

Energy Expenditure, kj/wk

<630 630-<16801680-<3150

3150-<6300 =6300

Vigorous activityNo. of deaths 1459 1123 428 279 439Relative risk (B1) 1.00 0.88 0.92 0.87 0.87

95% confidence Interval Referent 0.82-0.96 0.82-1.02 0.77-0.99 0.78-0.97

Nonvigorous activityNo. of deaths 481 652 784 804 1007Relative risk (B2) 1.00 0.89 1.00 0.98 0.92

95% confidence interval Referent 0.79-1.01 0.89-1.12 0.88-1.12 0.82-1.02

P(B1-B2) .29 ,1£ .53

Relative risks are adjusted for age, Quetelet's index, cigarette habit, physician-diagnosed hypertension or

diabetes mellitus, and early (<65 years) parental death. Relative risks for vigorous and nonvigorous energyexpenditure are mutually adjusted. Vigorous activities were defined as those that required 6 or more units of restingmetabolic rate (MET); nonvigorous activities, those that required less than 6 METs. of trend across categories ofvigorous energy expenditure is .007. of trend across categories of nonvigorous energy expenditure is .36.

Table 4.—Relative Risks of All-Cause Mortality Among Harvard Alumni, 1967 or 1971 Through 1988,According to Vigorous and Nonvigorous Physical Activity in 1962 or 1966*

Kind of Activity

Energy Expenditure, kJ/wk

<630 630--C16801680-<3150

3150-<6300 6300+

Vigorous activityNo. of deaths 1282 998 379 250 388

Relative risk (B1) 1.00 0.91 0.87

95% confidence interval Referent 0.81-0.96 0.81-1.02 0.76-1.00 0.76-0.96

Nonvigorous activityNo. of deaths 386 579 691 713 928

Relative risk (B2) 1.00 0.98 1.09 1.08 1.0595% confidence interval Referent 0.86-1.11 0.96-1.23 0.96-1.23 0.93-1.18

P(B1-B2) .19 <.044 .02 .02

Relative risks are adjusted for age, Quetelet's Index, cigarette habit, physician-diagnosed hypertension or

diabetes mellitus, and early (<65 years) parental death. Relative risks for vigorous and nonvigorous energyexpenditure are mutually adjusted. Vigorous activities were defined as those that required 6 or more units of restingmetabolic rate (METs); nonvigorous activities, those that required less than 6 METs. of trend across categoriesof vigorous energy expenditure is .001. of trend across categories of nonvigorous energy expenditure is .32.Analyses exclude first five years after physical activity assessment.

risks for nonvigorous energy expendi¬ture were 1.00 (referent), 0.94 (0.79 to1.12), 0.97 (0.82 to 1.15), 0.96 (0.81 to1.14), and 0.89 (0.75 to 1.04), respectively(P for trend=.18). The trend across cat¬egories of vigorous energy expenditurediffered significantly from that across

categories ofnonvigorous energy expen¬diture (P=.001).

We observed similar findings amongmen younger than 55 years and amongolder men. There also was no interac¬tion by Quetelet's index; findings didnot differ between leaner (<24.5 kg/m2)and heavier (>24.5) men.

In a final effort to disentangle theindependent associations of the twokinds of energy expenditure with mor¬

tality, we examined alumni who reportedonly one kind of activity in 1962 or 1966.Alumni who performed only vigorousactivities (and no nonvigorous exercise)numbered 919; of these, 259 died duringfollow-up. On thebasis ofthis small num¬

ber, we observed a marginally signifi¬cant trend (P=.05) of decreasing age-standardized mortality with increasingvigorous energy expenditure (data not

shown, but available from the authorson request). Among 1195 alumni whoreported only nonvigorous activity (andno vigorous exercise) in 1962 or 1966,380 died during follow-up. In age-stan¬dardized analysis, we found nonvigor¬ous energy expenditure and mortalityto be unrelated (P=.99) (data not shown,but available from the authors on re¬

quest).COMMENT

These prospective data demonstratea graded, inverse relationship betweenan index of total physical activity andmortality in middle-aged men, con¬

curring with most other investiga¬tions.9"11'13·14 Of the components of totalphysical activity, we found vigorous (ac¬tivities at a6 METs) but not nonvigor¬ous (<6 METs) exercise to be associ¬ated with decreased mortality. Men whoexpended 6300 kJ/wk or more in vigor¬ous exercise had 0.75 to 0.87 times therisk ofdying during follow-up, comparedwith those who expended less than 630kJ/wk. This difference in mortality riskis ofapproximately the same magnitude

as that between alumni 20% or more

overweight34 and those of ideal weight,6or that between alumni who smoked one

pack of cigarettes or less daily and non-

smokers.5These data also suggest that the de¬

crease in mortality associated withhigher levels ofenergy expenditure maytaper off after approximately 14 700 kJ/wk oftotal energy expenditure or 12 600kJ/wk of vigorous energy expenditure(Table 2). In a British study, investiga¬tors noted a similar finding amongyounger men, ie, declining rates of coro¬nary heart disease with increasing physi¬cal activity levels until the most activecategory ("very frequent sporting ex¬ercise or frequent sporting exercise plusother recreational activities"), in whichrates began to increase. Some investi¬gators have postulated that this phe¬nomenon may be explained by the in¬creased oxidative stress associated withprolonged physical exertion.35 However,when we classified alumni according totheir total energy expenditure in 1977(instead of 1962 or 1966, as in the pres¬ent analyses), we observed that mor¬tality continued to decline at 14700kJ/wk.5

It is unclear to us why vigorous, butnot nonvigorous, physical activity is as¬sociated with greater longevity. A re¬

cent report suggested that for favorablechanges in high-density lipoprotein cho¬lesterol and triglycéride leveis, a thresh¬old intensity of 5 to 6 METs of condi¬tioning exercise is needed.36 However,exercise intensity appears unrelated tothe magnitude ofdecrease in blood pres¬sure levels.2 Perhaps the inverse asso¬ciation between physical activity andmortality is related not so much to ex¬ercise itself, but to the improved car-

diorespiratory fitness that is induced.37·38Vigorous exercise is more effective thannonvigorous activity for cardiorespira-tory conditioning. The kind of vigorousactivity also may be relevant; for ex¬

ample, jogging, which is sustained anddynamic, is effective for such condition¬ing, whereasheavyyardwork is unlikelyto be as sustained and thus would beless effective in conditioning.

Several previous studies also have ex¬

amined the association of vigorous andnonvigorous exercise with health. Mor¬ris et al6·7 found an inverse relationshipbetween physical activity and coronaryheart disease incidence only among Brit¬ish men who reported vigorous sports.A similar observation was made amongFinnish men by Lakka et al.8 However,Shaper et al12 reported that even amongBritish men with little vigorous ("sport¬ing") exercise, myocardial infarctionrates apparently decreased (not formallytested) with increasing physical activ-



ity. Slattery et al11 also observed ap¬parent decreases in age-adjusted deathrates (not formally tested) with increas¬ing light to moderate activity among USrailroad workers who reported no vig¬orous activity. Among men who reportedsome vigorous activity, the trend withincreasing light to moderate activity wasunclear. In multivariate analyses adjust¬ing for age, smoking, blood pressure,and serum cholesterol level, the esti¬mated coefficient for vigorous activitywas significant, and that for light to mod¬erate activity, nonsignificant. However,investigators did not examine whetherthe coefficients for the two kinds of ac¬

tivity differed significantly.In investigating the independent, rela¬

tive merits ofvigorous and nonvigorousactivity, the issues involved are analo¬gous to those in epidemiologie studies ofdiet. We wished to investigate the in¬dependent associations of vigorous andnonvigorous physical activity with mor¬

tality, apart from their contributions tototal energy expenditure, and their rela¬tive merits. In studies ofdietary fat andcoronary heart disease, investigators areinterested in the effect of fat that isindependent of other nutrients, and oftotal energy intake.39 To separate thiseffect, various analytic strategies havebeen proposed.39-42 We have adapted oneof these strategies40 to achieve our end.As noted by Willett39 and restated re¬

cently by Wacholder et al,43 it is notadequate merely to note that the coef¬ficient of one kind of energy expendi¬ture (nutrient) is significant, and thecoefficient of the other, nonsignificant.The appropriate focus should be the dif¬ference in coefficients (ie, [B1-B2] inTables 3 and 4).

The most plausible alternate expla¬nation for our findings is that alumnireported light and moderate activitieswith greater imprecision than vigorousactivities, resulting in greater misclas-sification of the former. We had no datato test this hypothesis. In a separatevalidation study ofour physical activityquestionnaire, Ainsworth et al27 reportedthat in men aged 21 to 59 years, thecorrelation between energy expenditureestimated from the questionnaire andthat estimated from physical activity dia¬ries was .69 for activities of 6 METs ormore. For lower-intensity activities, thecorrelations were less than .35. How¬ever, among healthy alumni in the pres¬ent study, the correlation between non-

vigorous energy expenditure in 1962 or1966 and 1977 was comparable with thatfor vigorous energy expenditure (r=.35and .40, respectively). These low corre¬lations imply that during the long follow-up, patterns of physical activity hadchanged. In further analyses that did

account for changes in physical activityin 1977, we arrived at similar conclu¬sions.

We could not determine whether our

findings resulted from differences in diet,blood pressure levels, glucose tolerance,or serum lipid levels. We were unable tomake statistical adjustment for dietarydifferences because we did not have de¬tailed dietary information for alumni be¬fore 1988. According to the dietary dataof 1988, estimated total energy consumedincreased with increasing total, vigor¬ous, and nonvigorous energy expendi¬ture. However, the proportion of totalenergy consumed as fat or saturated fatdid not vary across activity categories.Thus, confounding by fat intake was un¬

likely, with the caveat that diet in 1988may not reflect earlier diet adequately.Some investigators argue that differ¬ences in blood pressure levels, glucosetolerance, and serum lipid levels shouldnot be controlled for. Physical traininglowers blood pressure levels,2 increasesinsulin sensitivity,44 and favorably in¬fluences lipid profiles.45 Therefore, thesevariables may represent some of themechanisms through which physical ac¬tivity modifies mortality risk. Ratherthan being true confounders, then, theymay represent events in the causal path¬way and thus should not be controlledfor.31 We did not have data on serumlipid levels, but we took into accountphysician-diagnosed hypertension anddiabetes mellitus. Self-reported physi¬cian-diagnosed disease among thesealumni is believed to be valid.4·20·30 Notadjusting for these diseases did not ma¬terially alter findings.

Although these observational datapreclude a conclusion of causality,46 sev¬eral highly plausible mechanisms existthat link increased physical activity todecreased mortality. In addition to thosedescribed previously, physical trainingalso improves cardiac mechanical andmetabolic function,47 reduces platelet ag¬gregation, and increases fibrinolytic ac¬

tivity.48 Our findings indicate that sed¬entary individuals should increase theiractivity level to enhance longevity. Spe¬cifically, vigorous activities were asso¬ciated with greater longevity. However,we strongly believe that even nonvig¬orous exercise is preferable to seden¬tariness. Our findings pertain only toall-cause mortality; meanwhile, evenmodest exercise has been shown to im¬prove, for example, lipid and glucoseprofiles.49·50

This study was supported by research grants HL34174 from the National Heart, Lung, and BloodInstitute and CA 44854 from the National CancerInstitute, Public Health Service, and in part by In¬stitutional National Research Service Award CA09001 from the National Cancer Institute, through

the Harvard University School of Public Health(Dr Lee).

This is report LIV in a series on chronic diseasein former college students.

We owe much to Jeremy N. Morris, MD, andWalter C. Willett, MD, DrPh, for their criticalcomments and invaluable help.

References1. Berlin JA, Colditz GA. A meta-analysis ofphysi-cal activity in the prevention of coronary heartdisease. Am J Epidemiol. 1990;132:612-628.2. Hagberg JM. Exercise, fitness, and hyperten-sion. In: Bouchard C, Shephard RJ, Stephens T,Sutton JR, McPherson BD, eds. Exercise, Fitness,and Health: A Consensus of Current Knowledge.Champaign, Ill: Human Kinetics Publishers Inc;1990:455-466.3. Helmrich SP, Ragland DR, Leung RW, Paffen-barger RS Jr. Physical activity and reduced occur-rence of non-insulin-dependent diabetes mellitus.N Engl J Med. 1991;325:147-152.4. Lee I-M, Paffenbarger RS Jr, Hsieh C-c. Physi-cal activity and risk ofdevelopingcolorectal cancer

among college alumni. J Natl Cancer Inst. 1991;83:1324-1329.5. Paffenbarger RS Jr, Hyde RT, Wing AL, LeeI-M, Jung DL, Kampert JB. The association ofchanges in physical-activity level and other life\x=req-\style characteristics with mortality among men.N Engl J Med. 1993;328:538-545.6. Morris JN, Everitt MG, Pollard R, Chave SPW.Vigorous exercise in leisure time: protection againstcoronary heart disease. Lancet. 1980;2:1207-1210.7. Morris JN, Clayton DG, Everitt MG, SemmenceAM, Burgess EH. Exercise in leisure time: coro-

nary attack and death rates. Br Heart J. 1990;63:325-334.8. Lakka TA, Venalainen JM, Rauramaa R, Sa-lonen R, Tuomilehto J, Salonen JT. Relation of lei-sure-time physical activity and cardiorespiratoryfitness to the risk of acute myocardial infarction inmen. N Engl J Med. 1994;330:1549-1554.9. Leon AS, Connett J, Jacobs DR, Rauramaa R.Leisure-time physical activity levels and risk ofcoronary heart disease and death: the Multiple RiskFactor Intervention Trial. JAMA. 1987;258:2388\x=req-\2395.10. Pekkanen J, Marti B, Nissinen A, TuomilehtoJ, Punsar S, Karvonen MJ. Reduction ofprematuremortality by high physical activity: a 20-year follow\x=req-\up ofmiddle-aged Finnish men. Lancet. 1987;1:1473\x=req-\1477.11. Slattery ML, Jacobs DR, Nichaman MZ. Lei-sure time physical activity and coronary heart dis-ease death: the US Railroad Study. Circulation.1989;79:304-311.12. Shaper AG, Wannamethee G. Physical activityand ischaemic heart disease in middle-aged Britishmen. Br Heart J. 1991;66:384-394.13. Lindsted KD, Tonstad S, Kuzma JW. Self-re-port of physical activity and patterns of mortalityin Seventh-Day Adventist men. J Clin Epidemiol.1991;44:355-364.14. Kaplan GA, Seeman TE, Cohen RD, KnudsenLP, Guralnik J. Mortality among the elderly in theAlameda County Study: behavioral and demographicrisk factors. Am J Public Health. 1987;77:307-312.15. Magnus K, Matroos A, Strackee J. Walking,cycling, or gardening, with or without seasonal in-terruption, in relation to acute coronary events.Am J Epidemiol. 1979;110:724-733.16. American College of Sports Medicine. Guide-lines for Graded Exercise Testing and ExercisePrescription. 3rd ed. Philadelphia, Pa: Lea & Fe-biger; 1985.17. Activity and Health Research. Allied DunbarNational Fitness Survey: A Report on ActivityPatterns and Fitness Levels: Main Findings. Lon-don, England: Sports Council and Health Educa-tion Authority; 1992.18. Pate RR, Pratt M, Blair SH, et al. Physicalactivity and public health: a recommendation formthe Centers for Disease Control and Prevention



and the American College of Sports Medicine.JAMA. 1995;273:402-407.19. Siegel PZ, Brackbill RM, Frazier EL, et al.Behavioral risk factor surveillance, 1986-1990.MMWR Morb Mortal Wkly Rep. 1991;40(SS-4):1\x=req-\23.20. Lee I-M, Paffenbarger RS Jr, Hsieh C-c. Physi-cal activity and risk of prostatic cancer among col-lege alumni. Am J Epidemiol. 1992;135:169-179.21. LaPorte RE, Cauley JA, Kinsey CM, et al. Theepidemiology of physical activity in children, col-lege students, middle-aged men, menopausal fe-males and monkeys. J Chronic Dis. 1982;35:787\x=req-\795.22. LaPorte RE, Black-Sandler RB, Cauley JA,Link M, Bayles C, Marks B. The assessment ofphysical activity in older women: analysis of theinterrelationship and reliability of activity moni-toring, activity surveys, and caloric intake/expen-diture. J Gerontol. 1983;30:394-397.23. Siconolfi SF, Lasater TM, Snow RCK, CarletonRA. Self-reported physical activity compared withmaximal oxygen uptake. Am J Epidemiol. 1985;122:101-105.24. CauleyJA, LaPorte RE, Sandler RB, SchrammMM, Kriska AM. Comparison of methods to mea-sure physical activity in postmenopausal women.

Am J Clin Nutr. 1987;45:14-22.25. Albanes D, Conway JM, Taylor PR, Moe PW,Judd J. Validation and comparison ofeight physicalactivity questionnaires. Epidemiology. 1990;1:65\x=req-\71.26. Washburn RA, Smith KW, Goldfield SRW,McKinlay JB. Reliability and physiologic correlatesof the Harvard Alumni Activity Survey in a generalpopulation. J Clin Epidemiol. 1991;44:1319-1326.27. Ainsworth BE, Leon AS, Richardson MT, Ja-cobs DR, Paffenbarger RS Jr. Accuracy of the Col-lege Alumnus physical activity questionnaire. J Clin-Epidemiol. 1993;46:1403-1411.28. Ainsworth BE, Jacobs DR Jr, Leon AS, et al.Compendium ofphysical activities: classification ofenergy costs of human physical activities. Med SciSports Exerc. 1993;25:71-80.

29. Taylor HL, Jacobs DR Jr, Schucker B, Knud-sen J, Leon AS, Debacker G. A questionnaire forthe assessment of leisure time physical activities.J Chronic Dis. 1978;31:741-755.30. Paffenbarger RS Jr, Wing AL, Hyde RT. Physi-cal activity as an index of heart attack risk in col-lege alumni. Am J Epidemiol. 1978;108:161-175.31. Rothman KJ. Modern Epidemiology. Boston,Mass: Little Brown & Co; 1986:41-49, 77-97.32. Cox DR. Regression models and life tables. JRStat Soc. 1972;34:187-220.33. Armitage P, Berry G. Statistical Methods inMedical Research. 2nd ed. Boston, Mass: BlackwellScientific Publications; 1987:91-92, 315-316.34. 1983 Metropolitan height and weight tables.Stat Bull Metrop Insur Co. 1983;64:2-9.35. Alessio HM. Exercise-induced oxidative stress.Med Sci Sports Exerc. 1993;25:218-224.36. Lakka TA, Salonen JT. Physical activity andserum lipids: a cross-sectional population study inEastern Finnish men. Am J Epidemiol. 1992;136:806-813.37. Ekelund L-G, Haskell WL, Johnson JL, Wha-ley FS, Criqui MH, Sheps DS. Physical fitness as

a predictor of cardiovascular mortality in asymp-tomatic North American men: the Lipid ResearchClinics Mortality Follow-up Study. N Engl J Med.1988;319:1379-1384.38. Blair SN, Kohl HW III, Paffenbarger RS Jr,Clark DG, Cooper KG, Gibbons LW. Physical fit-ness and all-cause mortality: a prospective study ofhealthy men and women. JAMA. 1989;262:2395\x=req-\2401.39. Willett WC. Implications of total energy intakefor epidemiologic analyses. In: Willett W, ed. Nu-tritional Epidemiology. New York, NY: OxfordUniversity Press; 1990:245-271.40. Howe GR, Miller AB, Jain M. Re: total energyintake: implications for epidemiologic analyses. AmJ Epidemiol. 1986;124:157-159.41. Shekelle RB, Nichaman MZ, Raynor WJJr. Re:total energy intake: implications for epidemiologicanalyses. Am J Epidemiol. 1987;126:980.42. Pike MC, Bernstein L, Peters RK. Re: total

energy intake: implications for epidemiologic analy-ses. Am J Epidemiol. 1989;129:1312-1313.43. Wacholder S, Schatzkin A, Freedman LS, Kip-nis V, Hartman A, Brown CC. Can energy adjust-ment separate the effects of energy from those ofspecific macronutrients? Am J Epidemiol. 1994;140:848-855.44. Newsholme EA. Effects of exercise on aspectsof carbohydrate, fat, and amino acid metabolism.In: Bouchard C, Shephard RJ, Stephens T, SuttonJR, McPherson BD, eds. Exercise, Fitness, andHealth:A Consensus ofCurrentKnowledge. Cham-paign, Ill: Human Kinetics Publishers Inc; 1990:293-308.45. Wood PD, Stefanick ML. Exercise, fitness, andatherosclerosis. In: Bouchard C, Shephard RJ, Ste-phens T, Sutton JR, McPherson BD, eds. Exercise,Fitness, and Health: A Consensus of CurrentKnowledge. Champaign, Ill: Human Kinetics Pub-lishers Inc; 1990:409-423.46. Smith GD, Phillips AN, Neaton JD. Smoking as

'independent' risk factor for suicide: illustration ofan artifact from observational epidemiology? Lan-cet. 1992;340:709-712.47. Froelicher VF. Exercise, fitness, and coronaryheart disease. In: Bouchard C, Shephard RJ, Ste-phens T, Sutton JR, McPherson BD, eds. Exercise,Fitness, and Health: A Consensus of CurrentKnowledge. Champaign, Ill: Human Kinetics Pub-lishers Inc; 1990:429-450.48. Oberman A. Rehabilitation ofpatients with coro-

nary artery disease. In: Braunwald E, ed. HeartDisease: A Textbook of Cardiovascular Medicine.3rd ed. Philadelphia, Pa: WB Saunders Co; 1988:1395-1409.49. Stefanick ML, Wood PD. Physical activity, lipidand lipoprotein metabolism, and lipid transport. In:Bouchard C, Shephard RJ, Stephens T, eds. Physi-cal Activity, Fitness, and Health: InternationalProceedings and Consensus Statement. Champaign,Ill: Human Kinetics Publishers Inc; 1994:417-431.50. Horton ES. Exercise and physical training: ef-fects on insulin sensitivity and glucose metabolism.Diabet Metab Rev. 1986;2:1-17.



exercise intensity and longevity in men

Documents