does patient blood glucose monitoring improve diabetes

Patient Blood Glucose Monitoring

991

McAndrew et al

Does Patient Blood GlucoseMonitoring ImproveDiabetes Control?A Systematic Review of the Literature

Objective

The purpose of this systematic review was 2-fold: first, toperform a comprehensive review of relevant studies onthe impact of self-monitoring of blood glucose (SMBG)on HbA1c levels for patients with type 2 diabetes melli-tus and, second, to explore mediators and moderatorswithin a self-regulation framework.

Data Sources

Five databases—Medline, PsychInfo, Cochrane Databaseof Systematic Reviews, Cochrane Central Register ofControlled Trials, and Cumulative Index to Nursing &Allied Health Literature (CINAHL)— were searched.

Study Selection

Cross-sectional, longitudinal, and randomized controltrials from 1990 to 2006, which included patients withtype 2 diabetes not on insulin, were reviewed. In total,6769 studies were screened for inclusion, 89 were retrievedfor detailed analysis, and 29 met criteria for inclusion inthe review.

Data Extraction

Data on the impact of SMBG on HbA1c, potential medi-ators and moderators, study design and participants, andlimitations of each study were retrieved.

Lisa McAndrew, MS

Stephen H. Schneider, MD

Edith Burns, MD

Howard Leventhal, PhD

From Rutgers University, Department of Psychology,Institute for Health & Behavior, New Brunswick,New Jersey (Ms McAndrew, Dr Leventhal); RobertWood Johnson Medical School, Division ofEndocrinology, Diabetes, Metabolism, New Brunswick,New Jersey (Dr Schneider); and Department ofMedicine, Medical College of Wisconsin, Milwaukee(Dr Burns). This article was funded through funds fromthe National Institute of Health, grant R24 AG023958,Center for the Study of Health Beliefs and Behaviors.

Correspondence to Howard Leventhal, PhD, RutgersUniversity, Department of Psychology, Institute forHealth & Behavior, 30 College Ave, New Brunswick,NJ 08901; e-mail: [email protected].

DOI: 10.1177/0145721707309807

TDE309807.qxd 11/7/2007 11:17 AM Page 991

Data Synthesis

Twenty-nine studies were included in this review: 9cross-sectional studies, 9 longitudinal studies, and 11randomized controlled trials. Evidence from the cross-sectional and longitudinal studies was inconclusive.Evidence from randomized controlled trials suggests thatSMBG may lead to improvements in glucose control.Very few studies examined potential mediators or mod-erators of SMBG on HbA1c levels.

Conclusions

SMBG may be effective in controlling blood glucose forpatients with type 2 diabetes. There is a need for studiesthat implement all the components of the process forself-regulation of SMBG to assess whether patient use ofSMBG will improve HbA1c levels.

This review summarizes current studies on therelationship between patient self-monitoringof blood glucose (SMBG) and the control oftype 2 diabetes. Although there is some evi-dence to suggest that SMBG is effective in

improving blood glucose levels among patients with type2 diabetes not on insulin, the evidence is still unclear andcontested. In this article, assessment of these studiesfocuses on the overall efficacy of monitoring on glucosecontrol and, more important, examines what the studiesdo and do not tell us as to how, and for whom, SMBGmay be effectively incorporated into patients’ voluntary,self-regulatory behavior to achieve effective blood glu-cose control. The authors use the commonsense model ofself-regulation1,2 as the theoretical framework for apprais-ing whether the studies address the “for whom” and“how” questions. The use of the model is consistent withcalls to make theory more explicit in research on diabeteseducation3 and to state hypotheses a priori.4

The current debate is focused on whether SMBG isefficacious among patients with type 2 diabetes who arenot on insulin.5,6 The American Diabetes Association(ADA), in its Clinical Practice Recommendations 2007,has found sufficient evidence to recommend SMBG forpatients prescribed multiple daily insulin injections(ADA level of evidence = A); the recommendation for

the use of SMBG among patients not on insulin, in addi-tion to recommendations for when and how patientsshould use SMBG, is based on less sufficient evidence(ADA level of evidence = E).7 The ambiguity regardingthe use of SMBG for patients not on insulin arises asprior studies and reviews have focused on the overallimpact of SMBG on HbA1c and have not assessedwhether the implementation was sufficiently comprehen-sive to address each of the factors needed to create acomplete feedback system for effective, patient-centeredself-management.6,8-11 The central issue for both theoryand practice is to identify the conditions under whichself-monitoring, a patient-centered “executive controlprocess,” can control the biomarkers that are ordinarilyregulated by complex, automatic, involuntary processesin patients without diabetes. Teaching volitional manage-ment to patients with type 2 diabetes involves more thaninstructing them in the use of a meter. SMBG can only beeffective if patients recognize when a reading is high orlow; make changes in oral medication, diet, or exercise inresponse to the reading; and evaluate the efficacy of theseself-management behaviors with a subsequent reading.The patients also must learn that these objective readingsand not their subjective feelings (fatigue, dizziness, etc)are valid indicators of elevated blood glucose and thebasis for guiding self-management. Learning to useobjective readings rather than subjective feelings asthe target for self-regulation means that patients mustignore the commonsense knowledge of their personalhistories and culture that define symptoms and dysfunc-tion as valid signs of illness.12,13

The logic underlying this study’s view of patient-centered self-management (ie, the commonsense model ofself-regulation1-3) led to the expectation that the efficacyof SMBG would depend on whether the interventionscreated a patient-centered behavioral control system thatwould address the patient’s skills in (1) taking a bloodglucose reading; (2) interpreting the reading as a targetfor action; (3) perceiving linkages between specific behav-iors (diet, exercise) and the reading (ie, which behaviorslower an above-target reading and which raise a below-target reading); (4) implementing action plans (ie, behav-ioral and treatment adjustments) in response to SMBG;(5) giving less weight to subjective symptoms that are thebasis for commonsense decisions that one is sick or well,as these cues are invalid guides for the regulation of bloodglucose levels; (6) incorporating the behavioral systeminto the patient’s ongoing daily behavioral patterns to

The Diabetes EDUCATOR

992

Volume 33, Number 6, November/December 2007

TDE309807.qxd 11/7/2007 11:17 AM Page 992

eventually become automatic; and (7) viewing difficul-ties in achieving control as issues of adjusting the behav-ioral treatment, not deficits in personal motivation orcompetence for self-management. The goals of this sys-tematic review are (1) to report on the overall impact ofSMBG on HbA1c for patients with type 2 diabetes and(2) to examine whether existent studies have identifiedand assessed the mediators and moderators comprisingthe processes underlying the relationship of SMBG toHbA1c.

Method

Searching

Five databases were searched: Medline, PsychInfo,Cochrane Database of Systematic Reviews (first quarter2006), Cochrane Central Register of Controlled Trials(first quarter 2006), and Cumulative Index to Nursing &Allied Health Literature (CINAHL). Search terms includedwords for (1) diabetes mellitus/diabetes, (2) monitoring/self-monitoring, and (3) glucose/blood glucose/bloodglucose in each database.

Selection

All citations were entered into endnotes and dupli-cates removed. Reviewer LD screened the remainingentries’ titles and abstracts for articles that met the inclu-sion and exclusion criteria. The articles were thenretrieved and read for final inclusion.

The inclusion criteria were empirical studies pub-lished since 1990 that included patients with type 2 dia-betes who were not using insulin and that examined theeffect of SMBG on HbA1c levels, the primary outcomemeasure, and the currently accepted standard for moni-toring long-term glucose control.14-18 Because a majorobjective of this review is to identify conditions thatmediate the efficacy of SMBG, this study retained cross-sectional, longitudinal, and randomized control trials(RCTs), as well as studies that included both patientswho were on and not on insulin and studies with patientswith both type 2 and type 1 diabetes.

Studies that focused on special subsets of patients,such as adolescents with type 2 diabetes and patientswith gestational diabetes and patients dealing with blind-ness or renal failure, were excluded, as these patients arenot representative of the broader population of patients

with type 2 diabetes. It is unclear how findings in spe-cialized populations would illuminate the efficacy ofmonitoring in the larger population. Studies were alsoexcluded if monitoring was performed by someone otherthan the patient (eg, a health professional) or used fruc-tosamine monitoring. Although this discussion isinformed by findings from relevant qualitative studies, acomprehensive review of these studies was not con-ducted as they do not provide quantitative results of theeffect of SMBG on HbA1c. Studies in which SMBG wasembedded in complex interventions that used multiplefactors unrelated to SMBG monitoring were alsoexcluded. The decision to exclude studies publishedbefore 1990 was based on substantial improvements inease and accuracy of monitoring from that time to thepresent. As the efficacy of monitoring for patients oninsulin is accepted,7 studies assessing efficacy of moni-toring only for patients on insulin were excluded. Finally,only studies available in English were reviewed.

Validity Assessment

Studies were reviewed for the ADA evidence grad-ing criteria by LD, and questions regarding gradingwere referred to and resolved by the medical membersof the team (SHS, EB). Only studies that met ADA evi-dence grading criteria A, B, or C were included in thereview.

Data Abstraction

The data were collected and organized by LD, and atrained research assistant reviewed the data for accuracy.Both the results of the data abstraction and questionsabout the data were then reviewed and discussed in detailwith all authors.

Study Characteristics

Data were collected on the primary outcome (HbA1c),study population (number of participants, gender, ethnic-ity, current medication used by patients, and type of dia-betes), study design, mediators and moderators, andlimitations.

Data Synthesis

Data are summarized in Table 1. Missing data weremarked “not reported” (NR) within the table.


993

McAndrew et al

(text continues on p. 1003)

TDE309807.qxd 11/7/2007 11:17 AM Page 993


994


Tab

le 1

Stud

y Ch

arac

teris

tics

Chan

et a

l

(200

0)20

Fran

cios

i et a

l

(200

1)21

Hann

inen

et a

l

(200

1)22

Harr

is e

t al

(200

1)23

N =

562

,typ

e 2,

no m

edic

atio

n

limits

repo

rted,

mal

e =

43%

,

ethn

icity

= N

R,m

ean

age

=

53,s

tudy

com

plet

ed in

Chin

a.

N =

285

5,ty

pe 2

,no

med

icat

ion

limits

,eth

nici

ty =

NR,m

ean

age

= 6

3,st

udy

com

plet

ed in

Ital

y.

N =

260

,typ

e 2,

no m

edic

atio

n

limits

,mal

es =

54%

,

ethn

icity

= N

R,m

ean

age

= 5

6

(pat

ient

s’ d

escr

iptio

ns w

ere

repo

rted

in a

pre

viou

s st

udy)

,

stud

y co

mpl

eted

in F

inla

nd.

N =

148

0,ty

pe 2

,no

med

icat

ion

limits

repo

rted,

mal

e =

44%

,eth

nici

ty =

27%

iden

tifie

d as

min

ority

,

mea

n ag

e =

63.

Cros

s-se

ctio

nal.

Recr

uite

d

cons

ecut

ive

patie

nts

at

teac

hing

hos

pita

l.

Cros

s-se

ctio

nal.

Patie

nts

with

type

2 d

iabe

tes

wer

e

recr

uite

d fro

m d

iabe

tes

clin

ics

and

GPs

in It

aly.

Cros

s-se

ctio

nal.

Appr

oach

ed

381

patie

nts

to p

artic

ipat

e;

68%

(N =

260

) par

ticip

ated

.

Cros

s-se

ctio

nal.

Anal

ysis

from

NHAN

ES II

I.

SMBG

was

ass

ocia

ted

with

low

er H

bA1c

(SM

BG =

7.9

%

vs n

ot S

MBG

= 8

.6%

).

This

rela

tions

hip

was

not

sign

ifica

nt in

line

ar

regr

essi

on a

naly

sis,

whi

ch

look

ed a

t man

y pr

edic

tors

.

SMBG

was

ass

ocia

ted

with

high

er H

bA1c

.

SMBG

was

ass

ocia

ted

with

high

er H

bA1c

(SM

BG =

9%

vs n

ot S

MBG

= 8

.5%

).

SMBG

was

ass

ocia

ted

with

high

er H

bA1c

.

NR For p

atie

nts

not o

n in

sulin

,

SMBG

was

ass

ocia

ted

with

high

er H

bA1c

.

For p

atie

nts

on in

sulin

,SM

BG w

as

not a

ssoc

iate

d w

ith H

bA1c

.

Abili

ty to

adj

ust i

nsul

in w

as

asso

ciat

ed lo

wer

HbA

1c

amon

g pa

tient

s w

ho h

ad

SMBG

>1

a da

y.

NR With

in e

ach

treat

men

t (in

sulin

vs o

ral m

edic

atio

n vs

.die

t),

ther

e w

as li

ttle

asso

ciat

ion

betw

een

SMBG

and

HbA

1c.

Patie

nts

on in

sulin

had

hig

her

HbA1

c an

d w

ere

mor

e lik

ely

to m

onito

r.

Stud

ySa

mpl

eDe

sign

Resu

lts-H

bA1c

Mod

erat

ors

and

Med

iato

rs

TDE309807.qxd 11/7/2007 11:17 AM Page 994


995

McAndrew et al

Jaw

orsk

a et

al

(200

4)26

Mitc

hell

et a

l

(200

4)25

Oki e

t al (

1997

)24

Patri

ck e

t al

(199

4)27

Rost

et a

l

(199

0)19

N =

218

,typ

e 2,

no m

edic

atio

n

limits

repo

rted,

mal

es =

31%

,eth

nici

ty =

NR,

mea

n

age

= 6

2%,s

tudy

com

plet

ed

in P

olan

d.

N =

434

,typ

e 2,

no in

sulin

,

mal

e =

52%

,eth

nici

ty =

NR,

mea

n ag

e =

64,

stud

y

com

plet

ed in

Can

ada.

N =

98,

type

2,n

o m

edic

atio

n

limits

,mal

e =

29%

,eth

nici

ty:

AA =

69%

,C=

27%

,H=

2%,

O =

2%

,mea

n ag

e =

56.

N =

200

,typ

e 2,

no in

sulin

,

mal

e =

52%

,eth

nici

ty =

NR,

mea

n ag

e =

65.

N =

84,

type

2,n

o m

edic

atio

n

limita

tions

,mea

n ag

e =

56,

32%

= m

ale,

ethn

icity

= N

R.

Cros

s-se

ctio

nal.

Recr

uite

d

patie

nts

who

pre

sent

ed fo

r

outp

atie

nt c

are.

Cros

s-se

ctio

nal.

This

stu

dy

repo

rted

base

line

data

from

a

rand

omiz

ed c

ontro

l tria

l.

Patie

nts

wer

e gi

ven

a

ques

tionn

aire

and

sen

t to

have

HbA

1c te

sted

.

Cros

s-se

ctio

nal.

A to

tal o

f 98

cons

ecut

ive

patie

nts

who

pres

ente

d fo

r rou

tine

care

.

Cros

s-se

ctio

nal.

Com

pare

d

patie

nts

who

mon

itore

d

(blo

od,u

rine,

or b

oth)

to

thos

e w

ho d

id n

ot m

onito

r.

Cros

s-se

ctio

nal.

Surv

eyed

patie

nts

with

type

2 d

iabe

tes

hosp

italiz

ed fo

r ele

ctiv

e

adm

issi

on.

SMBG

was

not

ass

ocia

ted

with

HbA1

c.

SMBG

was

not

ass

ocia

ted

with

HbA1

c.

SMBG

was

not

ass

ocia

ted

with

HbA1

c.

SMBG

was

not

ass

ocia

ted

with

HbA1

c.

SMBG

was

ass

ocia

ted

with

low

er H

bA1c

.

With

in e

ach

treat

men

t gro

up

(insu

lin v

s or

al m

edic

atio

n),

SMBG

was

not

rela

ted

to

HbA1

c.Am

ong

patie

nts

who

repo

rted

the

abili

ty to

cha

nge

insu

lin d

ose

base

d on

SM

BG,

SMBG

was

not

rela

ted

to

HbA1

c.

NR No d

iffer

ence

in H

bA1c

betw

een

patie

nts

who

had

SMBG

<7/

wk

and

patie

nts

who

had

SM

BG >

7/w

k.

SMBG

not

ass

ocia

ted

with

HbA1

c w

hen

anal

yzed

bas

ed

on tr

eatm

ent g

roup

(ins

ulin

vs d

iet/o

ral).

No d

iffer

ence

bet

wee

n pa

tient

s

who

repo

rted

that

they

wou

ld

act o

n m

onito

ring

resu

lts a

nd

thos

e w

ho re

porte

d th

at th

ey

wou

ld n

ot a

ct.

Not t

este

d di

rect

ly,bu

t the

rela

tions

hip

of S

MBG

to H

bA1c

was

inde

pend

ent

of o

ther

sel

f-ca

re

beha

vior

s.

(con

tinue

d)

TDE309807.qxd 11/7/2007 11:17 AM Page 995


996


Blon

de e

t al

(200

2)34

Karte

r et a

l

(200

1)33

Klei

n et

al

(199

3)35

Mei

er e

t al

(200

2)32

N =

228

,typ

e 2,

no m

edic

atio

n

limits

repo

rted,

mal

e =

NR,

ethn

icity

= N

R,m

ean

age

=

NR (a

ge ra

nge,

35-6

5).

N =

23

412,

type

s 1

& 2,

no

med

icat

ion

limits

,mal

e =

52%

,eth

nici

ty:C

= 6

0%,

AA =

11%

,H =

8%

,A/P

I =

12%

,NA

= 1

%,O

= <

1%,

M =

7%

,mea

n ag

e =

60.

N =

229

,typ

e 2,

no m

edic

atio

n

limits

,mal

e =

97%

,eth

nici

ty:

AA =

11%

,C =

68%

,H =

20%

,mea

n ag

e =

62,

recr

uite

d fro

m V

A.

N =

146

7,ty

pe 2

,no

insu

lin,

mal

e =

98%

,eth

nici

ty =

NR,

mea

n ag

e =

64,

patie

nts

recr

uite

d fro

m V

A.

Long

itudi

nal.

Char

t rev

iew

com

pare

d co

nsis

tent

repo

rting

of S

MBG

vs

inco

nsis

tent

repo

rting

of S

MBG

vs

no

repo

rting

of S

MBG

.

Long

itudi

nal.

Revi

ewed

HM

O

data

base

for a

dher

ence

to g

ener

al S

MBG

reco

mm

enda

tions

.

Long

itudi

nal.

Revi

ewed

cha

rts

at V

A m

edic

al c

ente

r for

patie

nts

with

type

2 d

iabe

tes

who

rece

ived

blo

od o

r urin

e

mon

itorin

g su

pplie

s.

Long

itudi

nal.

Char

t rev

iew

of

patie

nts

who

fille

d sc

ript f

or

mon

itorin

g st

rips.

Initi

ated

a

VA p

olic

y ch

ange

that

redu

ced

the

num

ber o

f

disp

ense

d st

rips

for p

atie

nts

with

type

2 d

iabe

tes

not o

n

insu

lin to

50

per 9

0 da

ys.

Seve

nty

perc

ent o

f pat

ient

s w

ith

cons

iste

nt re

porti

ng o

f SM

BG

had

95%

of r

epor

ted

HbA1

c

belo

w 8

% v

s 18

% o

f pat

ient

s

with

inco

nsis

tent

repo

rting

of

SMBG

vs

22%

of p

atie

nts

with

no

repo

rting

of S

MBG

.

Follo

win

g AD

A

reco

mm

enda

tions

for S

MBG

,

base

d on

pat

ient

dia

bete

s

type

and

med

icat

ion,

was

asso

ciat

ed w

ith lo

wer

HbA

1c.

Ther

e w

ere

no s

igni

fican

t

diffe

renc

es b

etw

een

SMBG

and

urin

e m

onito

ring.

SMBG

was

ass

ocia

ted

with

low

er H

bA1c

.Cha

nges

in

frequ

ency

of S

MBG

had

no

effe

ct o

n Hb

A1c.

NR Grea

ter f

requ

ency

of S

MBG

asso

ciat

ed w

ith lo

wer

HbA

1c.

HbA1

c di

d no

t diff

er b

ased

on

leng

th o

f tim

e m

onito

red

or

num

ber o

f stri

ps d

ispe

nsed

.

Seve

n pa

tient

s re

porte

d

mak

ing

insu

lin c

hang

es in

resp

onse

to m

onito

ring;

this

was

not

ass

ocia

ted

with

HbA1

c.

NR

Tab

le 1

(c

on

tinu

ed

)

Stud

ySa

mpl

eDe

sign

Resu

lts-H

bA1c

Mod

erat

ors

and

Med

iato

rs

TDE309807.qxd 11/7/2007 11:17 AM Page 996


997

McAndrew et al

New

man

et a

l

(199

0)31

Rind

one

et a

l

(199

7)28

Soum

erai

et a

l

(200

4)36

Wen

et a

l

(200

4)30

Wie

land

et a

l

(199

7)29

N =

38,

type

s 1

& 2,

no

med

icat

ion

limits

,mal

e =

NR,e

thni

city

= N

R,m

ean

age

= 6

0,re

crui

ted

from

VA.

N =

115

,typ

e 2,

on o

ral

med

icat

ions

onl

y,m

ale

= N

R,

ethn

icity

= N

R,m

ean

age

= 6

8,pa

tient

s re

crui

ted

from

VA.

N =

321

9,on

insu

lin o

r ora

l

diab

etes

med

icat

ion,

type

1

or 2

,mea

n ag

e =

54,

mal

e =

54%

,eth

nici

ty:C

=

46%

,AA

= 1

9%,O

= 3

%,

M =

32%

.

N =

976

,mal

e =

97%

,

pres

crib

ed o

ral m

edic

atio

n,

mea

n ag

e =

63,

ethn

icity

:

AA =

8%

,C =

38%

,H=

48%

,

O =

6%

,rec

ruite

d fro

m V

A.

N =

216

,typ

e 2,

pres

crib

ed

glyb

urid

e,m

ale

= 1

00%

,

ethn

icity

= N

R,m

ean

age

=

NR (a

ge ra

nge,

39-8

9),

recr

uite

d fro

m V

A.

Long

itudi

nal.

Rand

omly

revi

ewed

cha

rts fr

om p

atie

nts

who

mon

itore

d an

d pa

tient

s

who

did

not

(3 p

atie

nts

who

mon

itore

d ur

ine

wer

e

cons

ider

ed n

onm

onito

rs).

All p

atie

nts

had

2 Hb

A1c

reco

rded

a y

ear f

or 3

yea

rs.

Long

itudi

nal.

Com

pare

d pa

tient

s

who

wer

e pr

escr

ibed

mon

itorin

g st

rips

to p

atie

nts

who

wer

e no

t.Da

ta w

ere

colle

cted

ove

r 2 y

ears

.If

mul

tiple

dat

a po

ints

wer

e

reco

rded

,mea

ns w

ere

used

.

Long

itudi

nal.

Inte

rrup

ted

time-

serie

s an

alys

is o

f a d

atab

ase

of a

n HM

O th

at b

egan

offe

ring

free

SMBG

mon

itors

to p

atie

nts

with

dia

bete

s.

Prev

ious

yea

r’s d

ata

wer

e

cont

rolle

d fo

r.

Long

itudi

nal.

Revi

ewed

3 y

ears

of m

edic

al c

harts

for p

atie

nts

pres

crib

ed o

ral m

edic

atio

n fo

r

diab

etes

.Com

pare

d Hb

A1c

of

patie

nts

who

did

not

SM

BG v

s

SMBG

yea

r 3 v

s SM

BG y

ears

2-3

vs S

MBG

yea

rs 1

-3.

Long

itudi

nal.

SMBG

was

oper

atio

naliz

ed a

s re

fill o

f

SMBG

stri

ps.C

ompa

red

HbA1

c

of p

atie

nts

who

did

not

mon

itor,

SMBG

onc

e a

day,

and

SMBG

2 o

r mor

e tim

es a

day

.

SMBG

was

not

ass

ocia

ted

with

HbA1

c.

Rece

ivin

g a

scrip

t for

SM

BG

strip

s w

as n

ot a

ssoc

iate

d

with

HbA

1c.

Star

ting

SMBG

was

not

asso

ciat

ed w

ith im

prov

ed

HbA1

c fo

r pat

ient

s w

ith

“goo

d or

ade

quat

e co

ntro

l”

but w

as a

ssoc

iate

d w

ith

impr

oved

HbA

1c a

mon

g

thos

e w

ith p

oor c

ontro

l.

SMBG

was

not

ass

ocia

ted

with

HbA1

c.

Freq

uenc

y of

SM

BG w

as n

ot

asso

ciat

ed w

ith H

bA1c

.

NR Rece

ivin

g m

ore

bottl

es o

f SM

BG

strip

s w

as n

ot a

ssoc

iate

d

with

HbA

1c.

Effe

ct o

f SM

BG w

as m

oder

ated

by in

itial

HbA

1c a

nd

pote

ntia

lly m

edia

ted

by

med

icat

ion

adhe

renc

e.

NR NR

(con

tinue

d)

TDE309807.qxd 11/7/2007 11:17 AM Page 997


998


Alle

n et

al

(199

0)46

Davi

dson

et a

l

(200

5)43

Guer

ci e

t al

(200

3)37

N =

54,

type

2,n

o in

sulin

,

mal

e =

100

%,e

thni

city

:

C =

63%

,mea

n ag

e =

58,

recr

uite

d fro

m V

A.

N =

88,

type

2,n

o in

sulin

,

mea

n ag

e =

50,

ethn

icity

:AA

= 2

2%,H

= 7

5%,O

= 3

%.

N =

689

,typ

e 2,

oral

med

icat

ion

only,

mal

es =

55%

,eth

nici

ty =

NR,

mea

n

age

= 6

2,st

udy

com

plet

ed in

Fran

ce.

RCT.

Com

pare

d ur

ine

mon

itorin

g to

SM

BG.

SMBG

:SM

BG 3

6/m

onth

.

Urin

e:M

onito

red

urin

e

36/m

onth

.

Both

:Rec

eive

d di

et

coun

selin

g an

d m

onth

ly v

isits

with

trea

tmen

t tea

m,w

hich

used

trea

tmen

t alg

orith

m

(incl

udin

g m

edic

atio

n

chan

ges)

in re

spon

se to

mon

itorin

g.

RCT SM

BG =

inst

ruct

ed to

mon

itor B

G be

fore

and

1/2

hour

s af

ter m

eals

6 ti

mes

a

wee

k.

C =

not

inst

ruct

ed to

SM

BG.

Both

= m

eet w

ith d

ietit

ian

5

times

; nur

ses

blin

ded

to

treat

men

t use

d a

treat

men

t

algo

rithm

that

incl

uded

med

icat

ion

chan

ges.

RCT SM

BG =

SM

BG tr

aine

d by

GP; p

atie

nts

inst

ruct

ed to

mon

itor 6

/wee

k.

Both

= F

ollo

wed

eve

ry 6

wee

ks fo

r 24

wee

ks,g

iven

gene

ral d

ieta

ry a

nd d

iabe

tes

reco

mm

enda

tions

,HbA

1c

SMBG

:Bas

elin

e =

12.

4%,6

mon

ths

= 1

0.4%

(P <

.001

).

Urin

e:ba

selin

e =

11.

7%,6

mon

ths

= 9

.7%

(P <

.001

).

No s

igni

fican

t bet

wee

n-gr

oup

diffe

renc

es.

SMBG

= a

sig

nific

ant

redu

ctio

n of

0.8

% in

HbA

1c.

C =

a s

igni

fican

t red

uctio

n

of 0

.6%

.

No s

igni

fican

t bet

wee

n-gr

oup

diffe

renc

es.

SMBG

= a

sig

nific

ant

redu

ctio

n of

0.8

8% in

HbA

1c.

Cont

rol =

a s

igni

fican

t

redu

ctio

n of

0.6

0%.

SMBG

sho

wed

a s

tatis

tical

ly

sign

ifica

nt im

prov

emen

t as

com

pare

d to

con

trol.

Tren

d to

war

d yo

unge

r and

bette

r edu

cate

d im

prov

ing

mor

e.Ba

selin

e Hb

A1c,

base

line

wei

ght,

base

line

fast

ing

plas

ma

gluc

ose,

use

of o

ral m

edic

atio

ns,d

urat

ion

of d

iabe

tes,

and

race

did

not

pred

ict i

mpr

oved

glu

cose

cont

rol.

NR Patie

nts

who

did

SM

BG w

ere

mor

e lik

ely

to re

port

follo

win

g di

etar

y ad

vice

.

Pred

icto

rs o

f im

prov

emen

t

incl

uded

SM

BG g

roup

,hig

her

initi

al H

bA1c

,low

er d

urat

ion

of d

iabe

tes,

and

low

er b

ody

mas

s in

dex.

Tab

le 1

(c

on

tinu

ed

)

Stud

ySa

mpl

eDe

sign

Resu

lts-H

bA1c

Mod

erat

ors

and

Med

iato

rs

TDE309807.qxd 11/7/2007 11:17 AM Page 998


999

McAndrew et al

Kibr

iya

et a

l

(199

9)38

Kwon

et a

l

(200

4)39

N =

64,

type

2,o

n or

al

med

icat

ion

or in

sulin

,mal

e =

55%

,mea

n ag

e =

50,

ethn

icity

= N

R,st

udy

com

plet

ed in

Ban

glad

esh.

N =

110

,typ

e 2,

no m

edic

atio

n

limits

repo

rted,

mal

e =

61%

,

ethn

icity

= N

R,m

ean

age

=

54,s

tudy

com

plet

ed in

Kore

a.

take

n ev

ery

3 m

onth

s,an

d

GP a

llow

ed to

cha

nge

treat

men

t bas

ed o

n Hb

A1c

at

3-m

onth

vis

it.

RCT SM

BG =

2-d

ay c

lass

on

SMBG

,ins

truct

ed to

mon

itor

fast

ing

and

2 ho

urs

afte

r

brea

kfas

t and

/or 2

hou

rs

afte

r lun

ch a

nd to

adj

ust

med

icat

ions

.Tau

ght t

o m

ake

treat

men

t cha

nges

in

resp

onse

to S

MBG

.Vis

ited

doct

or e

very

3 m

onth

s.

Cont

rol =

Mon

thly

vis

its;

doct

or c

ould

mak

e m

ed

chan

ges.

Both

:Rec

eive

d ed

ucat

ion.

RCT In

tern

et S

MBG

= 1

2-w

eek

Inte

rnet

reco

rdin

g of

SM

BG;

patie

nt re

ceiv

ed

reco

mm

enda

tions

via

the

Inte

rnet

,inc

ludi

ng s

ome

med

icat

ion

chan

ges,

no

outp

atie

nt v

isit,

SMBG

pre

-

and

post

mea

ls.

Cont

rol =

regu

lar m

onth

ly

outp

atie

nt v

isits

.

Both

= re

com

men

ded

to

SMBG

3 o

r mor

e da

ys p

er

wee

k 1

to 3

tim

es a

day

,

incl

udin

g af

ter m

eals

.

SMBG

= H

bA1c

dec

reas

ed

1.37

% (P

= .0

22).

Cont

rol =

HbA

1c d

ecre

ased

0.38

% (P

= .2

9).

Did

not r

epor

t if t

here

was

a

stat

istic

ally

sig

nific

ant

diffe

renc

e be

twee

n th

e

2 gr

oups

.

Inte

rnet

SM

BG =

HbA

1c

redu

ced

0.54

% (P

< .0

01).

Cont

rol =

HbA

1c in

crea

sed

0.33

% (N

S).

Ther

e w

as a

sta

tistic

ally

sign

ifica

nt d

iffer

ence

in th

e

2 gr

oups

.

NR The

diffe

renc

es b

etw

een

the

2 gr

oups

wer

e m

ore

pron

ounc

ed a

mon

g pa

tient

s

with

an

initi

al H

bA1c

of 7

or

high

er.

(con

tinue

d)

TDE309807.qxd 11/7/2007 11:17 AM 99


1000


Mile

s et

al

(199

7)47

Mor

elan

d et

al

(200

6)45

Muc

hmor

e et

al

(199

4)44

N =

150

,typ

e 2,

no m

edic

atio

n

limits

repo

rted,

mal

e =

60%

,

mea

n ag

e =

65,

ethn

icity

=

NR,s

tudy

com

plet

ed in

Unite

d Ki

ngdo

m.

N =

199

,typ

es 1

& 2

,no

med

icat

ion

limits

,mal

e =

58%

,eth

nici

ty =

NR,

mea

n

age

= 4

9.

N =

23,

type

2,n

o in

sulin

,

mal

e =

39%

,eth

nici

ty =

NR,

mea

n ag

e =

59,

all p

atie

nts

over

wei

ght.

RCT.

Patie

nts

aske

d to

eith

er

SMBG

or u

rine

mon

itor.

Both

= P

atie

nts

aske

d to

mon

itor e

ither

bef

ore,

2 ho

urs

afte

r mea

ls,o

r bed

time

once

daily

and

to a

ttend

4

educ

atio

n se

ssio

ns.A

t 3

mon

ths,

patie

nts

switc

hed

to

an a

ltern

ativ

e m

onito

ring

met

hod.

At 6

mon

ths,

patie

nts

wer

e ab

le to

cho

ose

pref

erre

d

met

hod

for m

onito

ring.

RCT Co

ntro

l = d

iabe

tes

educ

atio

n.

“MT”

(atte

ntio

nal c

ontro

l) =

mon

itorin

g ed

ucat

ion

sess

ion,

supp

ort f

rom

edu

cato

r.

“BGT

”(in

terv

entio

n) =

blo

od

gluc

ose

mon

itorin

g bo

okle

t,

mon

itorin

g ed

ucat

ion

sess

ion,

supp

ort f

rom

edu

cato

r.

RCT SM

BG =

edu

catio

n on

car

b

coun

ting

and

SMBG

,give

n

mon

itor a

nd s

trips

to m

onito

r 6

times

a d

ay (b

efor

e an

d 2

hour

s af

ter m

eals)

for 4

wee

ks,

then

redu

ced

to tw

ice

a da

y fo

r

4 m

onth

s,re

sults

cha

rted.

HbA1

c si

gnifi

cant

ly d

eclin

ed in

both

gro

ups

at 1

2 m

onth

s

(10.

3% to

7.5

% in

bot

h

grou

ps).

Ther

e w

ere

no

betw

een-

grou

p di

ffere

nces

.

No s

igni

fican

ce,i

mpr

ovem

ent i

n

HbA1

c in

any

gro

up.W

hen

indi

vidu

als

who

impr

oved

wer

e co

mpa

red

to in

divi

dual

s

who

did

not

impr

ove,

a

grea

ter p

erce

ntag

e of

impr

over

s w

ere

in th

e “B

GT”

grou

p.

SMBG

= H

bA1c

redu

ced

1.54

% (P

< .0

5).

Cont

rol =

HbA

1c re

duce

d

0.84

% (P

> .3

).

No s

igni

fican

t diff

eren

ces

betw

een

grou

ps a

t stu

dy e

nd.

NR NR Dura

tion

of d

iabe

tes,

initi

al

HbA1

c,an

d nu

mbe

r of S

MBG

wer

e no

t rel

ated

to H

bA1c

.

Tab

le 1

(c

on

tinu

ed

)

Stud

ySa

mpl

eDe

sign

Resu

lts-H

bA1c

Mod

erat

ors

and

Med

iato

rs

TDE309807.qxd 11/7/2007 11:17 AM 00


1001

McAndrew et al

Rutte

n et

al

(199

0)44

Schw

edes

et a

l

(200

2)41

N =

149

,typ

e 2,

no in

sulin

,

mal

e =

35%

,eth

nici

ty =

NR,

mea

n ag

e =

63.

N =

223

,typ

e 2,

no in

sulin

,

mal

e =

52%

,eth

nici

ty =

NR,

mea

n ag

e =

60.

Cont

rol =

gen

eral

dia

bete

s

educ

atio

n.

Both

= 2

8-w

eek

beha

vior

al

wei

ght c

ontro

l pro

gram

,

follo

wed

by

GP w

ho c

ould

mak

e m

edic

atio

n ch

ange

s.

RCT.

Eigh

t pra

ctic

es m

atch

ed

and

rand

omiz

ed.

Inte

rven

tion

= 3

3 pa

tient

s

rece

ived

2 to

5 e

duca

tion

sess

ions

on

SMBG

,and

patie

nts

cont

acte

d di

abet

es

nurs

e m

onth

ly to

repo

rt

fast

ing

bloo

d gl

ucos

e.If

BG

high

,mad

e ap

poin

tmen

t and

also

met

with

GP

afte

r 6

mon

ths.

Thirt

y-th

ree

patie

nts

did

not S

MBG

and

met

with

GP a

t lea

st 4

tim

es a

yea

r.Al

l

used

med

icat

ion

algo

rithm

to

treat

pat

ient

s.

Cont

rol =

pat

ient

s no

t

inst

ruct

ed in

SM

BG,d

id n

ot

use

fixed

app

oint

men

ts.

RCT SM

BG =

SM

BG b

efor

e an

d 1

hour

afte

r 3 m

eals

,tw

ice

a

wee

k,an

d re

cord

the

mon

itorin

g; p

atie

nts

wer

e

seen

eve

ry 4

wee

ks b

y a

nurs

e w

ho u

sed

a SM

BG

coun

selin

g al

gorit

hm.

Cont

rol =

non

stan

dard

ized

coun

selin

g ev

ery

4 w

eeks

.

Inte

rven

tion

= m

ean

grou

p

HbA1

c de

crea

sed

0.5%

(P <

.05)

.

Cont

rol =

mea

n gr

oup

HbA1

c

incr

ease

d in

0.5

% (P

< .0

01).

SMBG

= re

duct

ion

of 1

%.

Cont

rol =

redu

ctio

n of

0.5

4%.

Ther

e w

as a

sta

tistic

ally

sign

ifica

nt d

iffer

ence

bet

wee

n

the

grou

ps.

Patie

nts

with

initi

al H

bA1c

>10

in th

e in

terv

entio

n co

nditi

on

wer

e m

ore

likel

y to

impr

ove

than

thos

e w

ith in

itial

HbA

1c

>1

in th

e co

ntro

l con

ditio

n.

In th

e SM

BG g

roup

,not

mak

ing

beha

vior

al c

hang

es in

resp

onse

to S

MBG

led

to

failu

re.

Long

er d

iabe

tes

dura

tion

and

high

er b

asel

ine

HbA1

c

pred

icte

d a

dela

y in

resp

onse

.

(con

tinue

d)

TDE309807.qxd 11/7/2007 11:17 AM Page 1001


1002


Seat

on (1

996)

42N

= 1

0,ty

pe 2

,onl

y on

ora

l

med

icat

ions

,mea

n

age

= N

R,m

ale

= N

R,

ethn

icity

= N

R.

RCT.

Repo

rted

in a

bstra

ct.

Both

= s

tand

ardi

zed

treat

men

t alg

orith

m.

SMBG

= S

MBG

.

Cont

rol =

no

SMBG

.

SMBG

= s

tatis

tical

ly

sign

ifica

nt re

duct

ion

of 0

.8%

in H

bA1c

.

Cont

rol =

no

chan

ge.

Did

not r

epor

t if t

here

was

a

stat

istic

ally

signi

fican

t

diffe

renc

e be

twee

n th

e 2

grou

ps.

NR

Tab

le 1

(c

on

tinu

ed

)

Stud

ySa

mpl

eDe

sign

Resu

lts-H

bA1c

Mod

erat

ors

and

Med

iato

rs

RCT,

rand

omiz

ed c

ontro

l tria

l; SM

BG,s

elf-

mon

itorin

g of

blo

od g

luco

se; N

R,no

t rep

orte

d; G

P,ge

nera

l pra

ctiti

oner

; AA,

Afric

an A

mer

ican

; H,H

ispa

nic;

C,C

auca

sian

; O,o

ther

; A/P

I,As

ian/

Paci

fic Is

land

er; N

A,Na

tive

Amer

ican

; M,m

ixed

race

or e

thni

city

; ADA

,Am

eric

an D

iabe

tes

Asso

ciat

ion;

HM

O,he

alth

mai

nten

ance

org

aniz

atio

n; V

A,Ve

tera

ns A

dmin

istra

tion;

BG,

bloo

d gl

ucos

e; N

HANE

S III

,Thi

rd N

atio

nal H

ealth

and

Nut

ritio

nEx

amin

atio

n Su

rvey

.

TDE309807.qxd 11/7/2007 11:17 AM Page 1002

Results

Trial Flow and Study Characteristics

Twenty-nine studies met criteria for review: 9 cross-sectional studies, 9 longitudinal studies, and 11 RCTs(see Figure 1 for trial flow). Study details are shown inTable 1.

Data Synthesis

Data are presented by the 3 types of study design:cross-sectional, longitudinal, and RCT. Each section firstdescribes the results of SMBG on HbA1c, reports nexton study limitations, and then describes factors thatmediate or moderate the relationship between SMBGand HbA1c levels.

Cross-sectional studies. Results of the 9 cross-sectional studies were inconclusive. Two showed thatmonitoring of glucose was correlated with lower HbA1clevels,19,20 and 3 found that glucose monitoring correlatedwith higher HbA1c levels.21-23 Among the latter 3 studies,

that by Franciosi and colleagues21 found monitoringassociated with worse control only for patients not oninsulin; there was no association of SMBG with HbA1cfor patients on insulin. The remaining 4 studies reportedno significant association between self-monitoring andHbA1c24-27; 1 of the 4 compared “home glucose monitor-ing” of urine (74 patients) and blood glucose (19patients) to patients not monitoring (103) and found nosignificant differences in HbA1c among the groups.27

As monitoring and glucose control are assessed at thesame time point in cross-sectional studies, it is impossibleto tell whether monitoring affected glucose control or, con-versely, whether patients in poor control were motivated tomonitor. As it is unlikely that monitoring would worsencontrol, it seems likely that the 3 studies reporting a nega-tive relationship of monitoring to HbA1c reflect high levelsof self-monitoring among patients with elevated HbA1c.This interpretation is consistent with Harris’s23 finding thatpatients with higher levels of HbA1c were both more likelyto monitor and more likely to be on insulin.

With respect to mediators and moderators ofSMBG, only 3 cross-sectional studies assessed if and how


1003

McAndrew et al

Articles retrieved for detail appraisal (N=89)

Articles included within the review (N=29)

Articles excluded (N=60)Intervention too comprehensive (N=8). Did not examine the association of HbA1c and SMBG (N=26)

Only included pt on insulin

Articles screened for inclusion from databases and references of relevant

Articles eliminated because were not relevant to current review or duplicates

Figure 1. Trial flow diagram.

TDE309807.qxd 11/7/2007 11:17 AM Page 1003

monitoring affected patient behavior. For example,Franciosi and colleagues21 reported better control amongpatients prescribed insulin who self-administered inresponse to SMBG readings in comparison to patientswho did not use insulin in response to these readings.Self-management procedures were not assessed for non-insulin users. In a study by Patrick and colleagues,27 only38% (37 of 97) of patients with type 2 diabetes not treatedwith insulin reported they would change a treatment inresponse to self-monitoring feedback. Glucose control forthese patients was not significantly better than that of the62% of patients who did not report making changes; how-ever, it is unclear how often these patients actuallychanged their management practices in response to mon-itoring information. Finally, Jaworska and colleagues26

found that patients who reported being able to makeinsulin changes in response to readings did not have lowerHbA1c levels. In summary, the cross-sectional studies areinconclusive as to the result of the effect of SMBG on glu-cose control. Most important, the directionality of any ofthe reported effects is indeterminate.

Longitudinal studies. Nine studies examined the effectof SMBG on HbA1c using longitudinal data. Four28-31 ofthe 9 found no association of SMBG with HbA1c. Twoother studies found lower HbA1c levels among patientswho did SMBG in comparison to those who did not mon-itor32 and to those who did not monitor at a recommendedfrequency.33 A brief commentary in a seventh studystated that patients whose medical charts documentedthat they were doing SMBG were significantly morelikely to have HbA1c levels below 8%.34 An eighth studyby Klein and colleagues35 compared those who moni-tored urine to those who did SMBG. There was no con-trol group that did not monitor, and there was nodifference in HbA1c between the 2 groups that did.

The ninth study by Soumerai and colleagues36 exam-ined the effects of providing free monitors to patients ina New England health maintenance organization (onlymonitoring strips had been previously provided).Monitoring behavior, defined as test strip distribution,was examined for 19 months prior to the policy change,5 months during the transition, and 17 months after thepolicy change. An interrupted time-series analysis oftrend line for monitoring behavior, pre- and postintro-duction of free monitors, showed an increase in the initi-ation of SMBG and in the use of test strips amongpatients with type 2 diabetes on oral medication. Patients

who previously had been inconsistent in refilling theiroral medication and now increased their use of test stripsalso increased medication adherence, and patients whowere previously in poor control and began monitoringshowed a reduction in HbA1c levels. The analyses bySoumerai and colleagues36 suggest that the initial level ofHbA1c may moderate one’s ability to detect the efficacyof monitoring and that the effects of monitoring onHbA1c may be mediated by increases in medicationadherence.

The Methods sections of these studies point to limita-tions in the implementation and assessment of mediatorsand moderators similar to those found in the cross-sectional studies. It is unclear if patients were instructedto monitor because they were in poor control or if self-monitoring did not lead to control. It was also unclearwhether the intervention addressed the components ofself-regulation essential for mediating the benefits ofmonitoring. The specific instructions patients were givenabout the procedure and the validity of self-monitoring(ie, when to monitor and what to do in response to read-ings) were rarely described, and most studies did notassess these mediators. One study31 reported that whenclinicians became aware “that SMBG level may be unre-liable, these levels were rarely used alone as criteria forinsulin change or life-style adjustment suggestions.” Ifthis was explicitly or implicitly communicated topatients, it would likely undermine motivation to self-monitor and the use of SMBG to guide behavior.

Five of the 9 studies reported mixed results on the rela-tionship between frequency of monitoring and HbA1clevels. Karter and colleagues33 found that more frequentmonitoring was correlated with better control amongpatients with type 2 diabetes. Although Meier and col-leagues32 found that SMBG was associated with bettercontrol, they also found no change in glucose controlfrom a policy change that reduced the availability of teststrips and lowered the average frequency of monitoringfrom 1.37 to 0.74 times a day had. Finally, 3 of the 9 thatassessed frequency of monitoring28,29,35 found no evidenceof a relationship of frequency of SMBG to HbA1c, and 2reported no relationship between length of time monitor-ing and HbA1c levels.30,35 In summary, the evidence issuggestive at best that monitoring might help patients bet-ter manage blood glucose levels and lower HbA1c levels.

Randomized control trials. Eleven RCTs tested theeffects of monitoring on HbA1c levels. Six of the 11 trials


1004


TDE309807.qxd 11/7/2007 11:17 AM Page 1004

comparing monitoring to a control condition reportedimproved glucose control in the monitoring condition.37-42

One38 of the 6 reported a significant decrease of HbA1c(1.37%) in the intervention arm and a nonsignificantreduction (0.38%) in the control arm but did not, how-ever, report whether the magnitude of HbA1c reductionin the monitoring arm was statistically superior to that inthe control.

Two of the 5 remaining trials failed to detect an advan-tage of a monitoring intervention on glucose control.That by Davidson and colleagues43 reported a reductionof HbA1c in the monitoring condition (0.8%) not signif-icantly different from the control (0.6%), and Muchmoreand colleagues44 reported that the comparison betweenthe 12 patients in the monitoring group (HbA1c loweredby 1.54%) was not significantly greater than that for the11 patients in the control condition (HbA1c reduced0.84%); the study was likely underpowered. The third ofthe 5 studies by Moreland and colleagues45 used a 3-armdesign that compared an intervention, which combinedmonitoring with a booklet designed to help patients formmore reasonable “expectations and responses” to moni-toring, to monitoring alone and a control arm. Althoughthere were no significant differences in the magnitude ofthe change in HbA1c across the 3 conditions, Morelandand colleagues45 found that the number of patients whoshowed an improvement in HbA1c was significantlygreater in the monitoring intervention with a booklet.

The final 246,47 of the 5 nonsignificant RCTs are lessrelevant to the main objective of this review (ie, doespatient monitoring lead to improvements in HbA1c?) asthey compared SMBG to urine monitoring and did notuse a nonmonitoring control. Both studies46,47 reportedsignificant reductions in HbA1c in both trial arms and nodifferences in HbA1c between patients who monitoredurine levels and those monitoring blood glucose.

A number of methodological factors complicate theinterpretation of the results of these trials. In 3 tri-als,39,41,42 patients in the monitoring arm received addi-tional resources (ie, education, medication, or support)not given control patients. Additions that told patientshow to respond to high or low SMBG readings woulddefine a key mediator for the formation of a coherentself-regulatory system. Additions that facilitated reduc-tions in HbA1c but were unrelated to the behavioralmonitoring system (eg, additional social support) wouldbe defined as independent influences or trial confounds.Studies also combined data from patients with type 1 and

type 2 diabetes, as well as data from patients on insulin,with patients controlled by oral medication or dietalone.38,39 Perhaps less problematic, patients were ran-domized by group rather than by person, and newpatients were added to replace dropouts.38,44,46 Davidsonand colleagues43 conducted the only trial that blinded cli-nicians to the patients’ treatment group. Finally, the trialby Seaton42 should be given little weight as only 10patients were randomized, the author did not report if thebetween-group difference was statistically significant,and the results were reported only in an abstract.42

In examining potential moderators of response, 3studies37,39,40 found that patients in poor control weremore likely to benefit from SMBG. This supports theprior suggestion36 that the initial level of glucose controlwill moderate the effects of SMBG on HbA1c. Guerciand colleagues37 found that lower duration of diabetespredicted improved HbA1c, whereas Schwedes and col-leagues41 found that longer duration of diabetes may leadto a delayed response. Muchmore and colleagues44 andAllen and colleagues,46 however, did not find that eitherinitial HbA1c or duration of illness was related to theeffect of SMBG on HbA1c. Finally, Allen and colleaguesreported a trend for greater improvement in HbA1c fromSMBG among younger and more educated patients.

The RCTs differed on the instruction given to patientsconcerning monitoring. Supplemental findings from 2studies37,41 are consistent with the view that SMBG shouldbe viewed as a component of a larger behavioral system.Schwedes and colleagues41 instructed patients to monitor12 times a week before and after meals and to changetheir regulatory behaviors (diet, exercise, medication) inresponse to their SMBG readings; patients who changedbehaviors in response to SMBG readings had signifi-cantly lower HbA1c levels than those who did not.Although Guerci and colleagues37 did not give explicitinstruction to patients concerning changing behaviors inresponse to monitoring, they reported that patients whomonitored were more likely to follow dietary recommen-dations and had better controlled HbA1c.

Discussion

Effect of SMBG on HbA1c

Neither the cross-sectional nor longitudinal studiessupport the hypothesis that SMBG can be effective forimproving the control of blood glucose levels among


1005

McAndrew et al

TDE309807.qxd 11/7/2007 11:17 AM Page 1005

patients with type 2 diabetes not using insulin. Six stud-ies found that monitoring was related to lower levels ofHbA1c (2 cross-sectional19,20 and 4 longitudinal33-36), 8found no association between monitoring and HbA1c(4 cross-sectional24-27 and 4 longitudinal28-31), and 3 cross-sectional studies21-23 reported that SMBG was associatedwith worse control. Regardless of their findings, thecross-sectional studies are basically uninterpretable asone cannot conclude that SMBG preceded or was acci-dentally correlated with good control, or was initiated bypatients who were in poor control. The longitudinal stud-ies fail to clarify the picture as they are evenly dividedbetween positive effects and null effects.

A more favorable image of SMGB for controllingblood glucose levels emerges from the results of theRCTs and Soumerai and colleagues’36 well-designedinterrupted time-series analyses. Of the 9 RCTs37-45 thatcompared a monitoring intervention to a control, 637-42

reported lower levels of HbA1c for patients in the moni-toring condition. In no trial did the monitoring conditionlead to worsening of HbA1c, although for 343-45 of the 9,improvements in the monitoring condition were not sta-tistically greater than the control condition. Soumeraiand colleagues’36 time-series analysis finding of lowerHbA1c levels following the initiation of monitoring thanduring the nonmonitoring time period is consistent withthe results of the RCTs.

This study’s positive view of the benefit of SMBG incontrolling HbA1c for patients with type 2 diabetes isconsistent with previous reviews by Welschen and col-leagues in Diabetes Care6 and the Cochrane Database.10

Sarol and colleagues11 also concluded that SMBGimproved glucose control for patients with type 2 dia-betes. All 3 reviews examined only RCTs whose partici-pants were not on insulin. Less favorable judgments wereoffered by Coster and colleagues9 and Faas and col-leagues,8 both of whom stated that the evidence is insuf-ficient to support the hypothesis that self-monitoring iseffective in improving glycemic control, although bothcalled for further studies. Coster and colleagues9 andFaas and colleagues8 also included a study in which someparticipants were on insulin.

This study broadened the criteria for inclusion andreviewed cross-sectional studies, longitudinal studies,and RCTs using a variety of treatment strategies, includ-ing studies with some patients on insulin, to obtain asmuch evidence as possible respecting the effects ofSMBG on HbA1c and the moderators and mediators of

treatment outcomes.48-50 This study balanced the broaderinclusiveness in the type of studies reviewed by narrow-ing the focus to studies of patients with type 2 diabetes,some of whom were not on insulin and others who were.By excluding studies of patients with type 1 diabetes andstudies that only included patients on insulin, this studyavoided duplicating prior recommendations that have ledto acceptance of SMBG for patients using multipleinsulin injections a day (ADA level of evidence = A).7

This allowed the authors to conduct a thorough review ofstudies for which the ADA recommendations have theleast evidence (ADA level of evidence = E).7 It is impor-tant to note, however, that the favorable evidence forSMBG for patients using insulin is consistent with thisstudy’s theoretical model. Insulin is relatively fast acting,and SMBG provides a direct indication of its effects. Thecombination of insulin and SMBG provides patients withclear evidence of the utility of the combination for con-trolling blood sugar levels and facilitates the acquisitionof the volitional patient-centered feedback loop essentialfor effective self-management.

Mediators and Moderators

The optimistic message that monitoring can be aneffective tool in patient-centered care for patients withtype 2 diabetes (ie, for self-regulation of blood glucoseand lower levels of HbA1c over time) assumes thatSMBG is embedded in a multicomponent system foreffective, volitional control. This study suggests thatconscious or voluntary management requires the imple-mentation of a complete control system in which thepatient must (1) know how to take a reading; (2) under-stand when the reading is above (or below) target values;(3) see the connection between deviant readings andprior behavior (eating, exercise, and under- or overuse ofmedication); (4) have and implement an action plan tocontrol glucose levels; (5) rely more heavily on SMBGreadings and give less weight to subjective feelings ofwell-being and possibly false signs of hyperglycemicdistress (eg, feeling dizzy or shaky); (6) create simpleaction plans that will allow the patient to integrate theminto his or her ongoing life patterns, the use of SMBG,and the behaviors needed for effective blood glucosemanagement; and (7) evaluate glucose readings in anonjudgmental framework. Meeting blood glucose tar-gets needs to be viewed as an ongoing problem-solvingprocess and target misses as problems with the intensity


1006


TDE309807.qxd 11/7/2007 11:17 AM Page 1006

or timing of specific control behaviors and not as fail-ures of the self.

Of these 6 hypotheses, only 1 hypothesis was examinedby the studies reviewed. The RCT conducted by Schwedesand colleagues41 found that patients who changed self-management behaviors in response to SMBG showedimprovement in HbA1c, and this finding is consistent withthis study’s hypothesis that SMBG is effective when it isused by the patient to make treatment adjustments.Similarly, Guerci and colleagues37 found that patients whomonitored were more likely to follow dietary advice andhad better glucose control. Of the 3 cross-sectional stud-ies21,26,27 and 1 longitudinal study35 that examined theimpact of treatment adjustment in response to SMBG onHbA1c, only 1 found that patients who made treatmentchanges in response to SMBG were in better control.21 Thequality of these studies, however, prevents interpretation(all used self-report, 2 studies asked only about the abil-ity26 or if the patient “would ever”27 adjust therapy, and in1 study, only 7 patients reported making changes35).Although this study had no specific hypotheses regardingmoderators, evidence from the review of RCTs suggestedthat greater initial HbA1c37,39,40 and shorter duration ofdiabetes37,41 may be associated with improved responseand lower HbA1c. Other RCTs, however, found no asso-ciation of initial HbA1c44,46 or duration of diabetes44,46 withimproved response.

This study’s finding of an inadequate assessment ofmoderators and mediators by the studies reviewed mayreflect a particular bias in the way medical investigatorsview the RCTs. Randomized controlled trials testingSMBG and lifestyle changes are not experiments inwhich the independent variable is a single factor, as seenin classical behavioral studies of perception and learning.The RCT for glucose monitoring implements a multi-component system; SMBG is not equivalent to a medicalRCT comparing an active pill to a placebo (with biasessuch as differences in adherence, etc, removed).Interventions that introduce monitoring without creatingthe perceived linkages between specific behaviors andmeaningful and valid blood glucose readings do littlemore than provide the patient a number, which may bediffer from target but be of little utility for control. Thecall for the assessment of moderators and mediators isnot new: it has been advocated by clinical trial investiga-tors.51 Assessment of the specific components of an inter-vention allows statistical analyses to identify those thatare active.52

Implications

This systematic review suggests that SMBG may leadto improvements in HbA1c for patients with type 2 dia-betes not on insulin. Few studies, however, implementedand/or examined the components of the behavioral mon-itoring system that self-regulation and behavior theorywould suggest as critical for patients to manage diabeteson their own. Thus, this study found little data to supportor refute hypotheses regarding mediators. For example,the studies reviewed did not report on the specifics ofwhat patients were instructed to do in response to theirmonitoring results. Some instructed patients to use fixedtime monitoring schedules, which does little to helppatients see how self-management by medication, diet,and exercise relates to glucose levels. And in 1 instance,46

patients were instructed to monitor blood glucose levelsbefore meals. It is not clear that patients would learn howspecific meals affected blood glucose levels by premealand/or fasting blood glucose monitoring. When SMBG isembedded in a volitional control system, it providesfeedback as to how specific behaviors increase ordecrease blood glucose.

In addition to using monitoring within a system thatconnects behaviors to blood glucose levels, interventionsneed to ensure that the objective data provided by SMBGare used in place of the commonsense subjective cues thatare highly available and easy for patients to use to man-age diabetes. Patients and clinicians believe that fluctua-tions in blood glucose produce subjective changes.Seventy-seven percent of patients have answered yes tothe question “Can you tell, just by how you feel, whenyour glucose is too high?”53 The subjective cues that com-mon sense selects as indicators of blood glucose may beambiguous, confounded by affective experiences unre-lated to variation in blood glucose and therefore poorlycorrelated with actual blood glucose levels. Not surpris-ing, patients generally underestimate their blood glucoselevels.53 Moving patients from commonsense cues toobjective blood glucose monitoring may be an importantstep for improving monitoring interventions and is con-sistent with Peel and colleagues’54 conclusion from theirdetailed qualitative findings that SMBG helped patientsmake “their otherwise invisible and imperceptible illnessvisible.” None of the studies reviewed assessed or dis-cussed the transition from subjective cues and feelings toobjective monitoring as patients shaped their self-regulatorybehaviors to control blood glucose.


1007

McAndrew et al

TDE309807.qxd 11/7/2007 11:17 AM Page 1007

Finally, success in self-regulation requires thatpatients create and implement action plans that fit withintheir daily lives. Studies of patients with type 1 diabeteshave taught strategies and specific skills that allowpatients to adjust insulin in response to glucose levelsand diet and develop a high level of autonomy for effec-tive management that allows them the “life they wouldlive without diabetes.”55,56 There is no indication that theinterventions reviewed here considered the issuesinvolved in incorporating SMBG into the daily life ofpatients with type 2 diabetes. This is not a trivial issue astime for self-management has been reported to take anaverage of 2 hours a day.57 Measures of quality of lifeare a potential source of evidence for effective andautonomous self-management processes that fit within apatient’s life pattern; quality of life was assessed in only421,22,41,44 of the studies reviewed. Only 1 study foundthat SMBG was associated with higher depression,21

whereas the remaining 322,41,44 found that SMBG was notassociated with impaired quality of life. Of these, 1 RCTfound that SMBG led to improved quality of life forpatients in the monitoring arm. Both qualitative54 andquantitative58 research on quality of life and SMBG isconsistent with the assumption that quality of life will bepositively associated with SMBG when it is imple-mented effectively (ie, the patient views the informationnonjudgmentally and takes responsibility for his or hermanagement).

Given the scarcity of the empirical evidence, cautionis needed respecting the conclusions that have beendrawn about specific mediators and moderators of theeffectiveness of SMBG. Caution does not, however, pre-clude emphasizing the importance of a theoretical analy-sis of the behavioral processes that underlie a volitionalaction such as SMBG. Effective blood glucose controlthrough blood glucose monitoring, either in a RCT or inclinical practice, requires careful and thorough imple-mentation. Medication and lifestyle changes that areunused or used inappropriately will not improve thefunction of a complex, in vivo physiological system; thisis true for the control of blood glucose, blood pressure,and other biological markers for risk. Successful treat-ment, whether it be medication, diet, or exercise in com-bination with SMBG, requires that practitioners addressboth the physiological and the behavioral control sys-tems regulating blood glucose levels. Failure of bloodglucose control in personalized medicine represents afailure of the medical care system, not a failure ofthe patient.

References

1. Leventhal H, Brissette I, Leventhal EA. The common sense modelsof self-regulation of health & illness. In: Cameron LD, Leventhal H,eds. The Self-Regulation of Health and Illness Behavior. London:Routledge Taylor & Francis Group; 2003:42-61.

2. Leventhal H, Diefenbach M, Leventhal EA. Illness cognition:using common sense to understand treatment adherence andaffect cognition interactions. Cogn Ther Res. 1992;16:143-163.

3. Anderson R, Funnell M. Theory is the care, vision is the horse:reflections on research in diabetes patient education. DiabetesEduc. 1999;25:43-51.

4. Cook D, Sackett D, Spitzer W. Methodological guidelines for sys-tematic reviews of randomized control trials in health care fromthe Potsdam consultation on meta-analysis. J Clin Epidemiol.1995;48:167-171.

5. Davidson MB. Counterpoint: self-monitoring of blood glucose intype 2 diabetic patients not receiving insulin: a waste of money.Diabetes Care. 2005;28:1531-1533.

6. Welschen LM, Bloemendal E, Nijpels G, et al. Self-monitoring ofblood glucose in patients with type 2 diabetes who are not usinginsulin: a systematic review [see comment]. Diabetes Care. 2005;28:1510-1517.

7. American Diabetes Association. Standards of medical care in dia-betes [position statement]. Diabetes Care. 2007;30:S14-S41.

8. Faas A, Schellevis FG, Van Eijk JT. The efficacy of self-monitoringof blood glucose in NIDDM subjects: a criteria-based literaturereview. Diabetes Care. 1997;20:1482-1486.

9. Coster S, Gulliford MC, Seed PT, Powrie JK, Swaminathan R.Monitoring blood glucose control in diabetes mellitus: a system-atic review. Health Technol Assess. 2000;4:1-93.

10. Welschen LMC, Bloemendal E, Nijpels G, et al. Self-monitoringof blood glucose in patients with type 2 diabetes mellitus who arenot using insulin. Cochrane Database Syst Rev. 2006;1:1.

11. Sarol JN Jr, Nicodemus NA Jr, Tan KM, Grava MB. Self-monitoring of blood glucose as part of a multi-component therapyamong non-insulin requiring type 2 diabetes patients: a meta-analysis (1966-2004). Curr Med Res Opin. 2005;21:173-184.

12. Meyer D, Leventhal H, Gutmann M. Common-sense models of ill-ness: the example of hypertension. Health Psychol. 1985;4:115-135.

13. Halm E, Mora P, Leventhal H. No symptoms, no asthma: the acuteepisodic disease belief is associated with poor self-managementamong inner city adults with persistent asthma. Chest. 2006;129:573-580.

14. American Diabetes Association. Standards of medical care in dia-betes [position statement]. Diabetes Care. 2004;27:S15-S35.

15. Vijan S, Hofer TP, Hayward RA. Estimated benefits of glycemiccontrol in microvascular complications in type 2 diabetes. AnnIntern Med. 1997;127:788-795.

16. UK Prospective Diabetes Study Group. Intensive blood-glucosecontrol with sulphonylureas or insulin compared with conven-tional treatment and risk of complications in patients with T2DM.Lancet. 1998;352:837-853.

17. Gray A, Raikou M, McGuire A, et al. Cost-effectiveness of an inten-sive blood glucose control policy in patients with type 2 diabetes:economic analysis alongside randomized controlled (UKPKS 41).Br Med J. 2000;320:1373-1378.

18. Carter JS, Pugh JA, Monterrosa A. Non-insulin-dependent dia-betes mellitus in minorities in the United States. Ann Intern Med.1996;125:221-232.


1008


TDE309807.qxd 11/7/2007 11:17 AM Page 1008

19. Rost KM, Flavin KS, Schmidt LE, McGill JB. Self-care predic-tors of metabolic control in NIDDM patients. Diabetes Care.1990;13:1111-1113.

20. Chan WB, Chan JC, Chow CC, et al. Glycaemic control in type 2diabetes: the impact of body weight, beta-cell function and patienteducation. QJM. 2000;93:183-190.

21. Franciosi M, Pellegrini F, De Berardis G, et al. The impact ofblood glucose self-monitoring on metabolic control and quality oflife in type 2 diabetic patients: an urgent need for better educa-tional strategies. Diabetes Care. 2001;24:1870-1877.

22. Hanninen J, Takala J, Keinanen-Kiukaanniemi S. Good continu-ity of care may improve quality of life in type 2 diabetes. DiabetesRes Clin Pract. 2001;51:21-27.

23. Harris MI. Frequency of blood glucose monitoring in relation toglycemic control in patients with type 2 diabetes [see comment].Diabetes Care. 2001;24:979-982.

24. Oki JC, Flora DL, Isley WL. Frequency and impact of SMBG onglycemic control in patients with NIDDM in an urban teachinghospital clinic. Diabetes Educator. 1997;23:419-424.

25. Mitchell CG, Bowker SL, Majumdar SR, Toth EL, Johnson JA.Lack of correlation between patient-reported outcomes andglycemic control in type 2 diabetes not managed by insulin. CanJ Diabetes. 2004;28:362-368.

26. Jaworska J, Dziemidok P, Kulik TB, Rudnicka-Drozak E.Frequency of self-monitoring and its effect on metabolic controlin patients with type 2 diabetes. Ann Univ Mariae CurieSklodowska [Med]. 2004;59:310-316.

27. Patrick AW, Gill GV, MacFarlane IA, Cullen A, Power E,Wallymahmed M. Home glucose monitoring in type 2 diabetes: isit a waste of time? Diabetes Med. 1994;11:62-65.

28. Rindone JP, Austin M, Luchesi J. Effect of home blood glucosemonitoring on the management of patients with non-insulindependent diabetes mellitus in the primary care setting. Am JManag Care. 1997;3:1335-1338.

29. Wieland LD, Vigil JM, Hoffman RM, Janis LW. Relationshipbetween home glucose testing and hemoglobin Alc in type II dia-betes patients. Am J Health Syst Pharm. 1997;54:1062-1065.

30. Wen L, Parchman ML, Linn WD, Lee S. Association betweenself-monitoring of blood glucose and glycemic control in patientswith type 2 diabetes mellitus. Am J Health Syst Pharm. 2004;61:2401-2405.

31. Newman WP, Laqua D, Engelbrecht D. Impact of glucose self-monitoring on glycohemoglobin values in a veteran population.Arch Intern Med. 1990;150:107-110.

32. Meier JL, Swislocki AL, Lopez JR, Noth RH, Bartlebaugh P,Siegel D. Reduction in self-monitoring of blood glucose in per-sons with type 2 diabetes results in cost savings and no change inglycemic control. Am J Manag Care. 2002;8:557-565.

33. Karter AJ, Ackerson LM, Darbinian JA, et al. Self-monitoring ofblood glucose levels and glycemic control: the NorthernCalifornia Kaiser Permanente Diabetes registry. Am J Med. 2001;111:1-9.

34. Blonde L, Ginsberg B, Horn S. Frequency of blood glucose mon-itoring in relation to glycemic control in patients with type 2 dia-betes. Diabetes Care. 2002;25:245-255.

35. Klein CE, Oboler SK, Prochazka A, et al. Home blood glucosemonitoring: effectiveness in a general population of patients whohave non-insulin-dependent diabetes mellitus. J Gen Intern Med.1993;8:597-601.

36. Soumerai S, Mah C, Zhang F, et al. Effects of health maintenanceorganization coverage of self-monitoring devices on diabetes self-care and glycemic control. Arch Intern Med. 2004;164:645-652.

37. Guerci B, Drouin P, Grange V, et al. Self-monitoring of blood glu-cose significantly improves metabolic control in patients withtype 2 diabetes mellitus: the Auto-Surveillance InterventionActive (ASIA) study. Diabetes Metab. 2003;29:587-594.

38. Kibriya MG, Ali L, Banik NG, Azad Khan AZ. Home monitoringof blood glucose (HMBG) in type-2 diabetes mellitus in a devel-oping country. Diabetes Res Clin Pract. 1999;46:253-257.

39. Kwon HS, Cho JH, Kim HS, et al. Establishment of blood glucosemonitoring system using the Internet. Diabetes Care. 2004;27:478-483.

40. Rutten G, van Eijk J, de Nobel E, Beek M, van der Velden H.Feasibility and effects of a diabetes type II protocol with bloodglucose self-monitoring in general practice. Fam Pract. 1990;7:273-278.

41. Schwedes U, Siebolds M, Mertes G, Group SS. Meal-relatedstructured self-monitoring of blood glucose: effect on diabetescontrol in non-insulin-treated type 2 diabetic patients. DiabetesCare. 2002;25:1928-1932.

42. Seaton TL. Benefit of self-monitoring of blood glucose in patientswith NIDDM receiving oral sulfonylureas [abstract]. Pharmaco-therapy. 1996;16:498.

43. Davidson MB, Castellanos M, Kain D, Duran P. The effect of selfmonitoring of blood glucose concentrations on glycated hemo-globin levels in diabetic patients not taking insulin: a blinded, ran-domized trial. Am J Med. 2005;118:422-425.

44. Muchmore DB, Springer J, Miller M. Self-monitoring of bloodglucose in overweight type 2 diabetic patients. Acta Diabetol.1994;31:215-219.

45. Moreland E, Volkening L, Lawlor M, Chalmers K, Anderson B,Laffel L. Use of a blood glucose monitoring manual to enhancemonitoring adherence in adults with diabetes. Arch Intern Med.2006;166:689-695.

46. Allen BT, DeLong ER, Feussner JR. Impact of glucose self-monitoring on non-insulin-treated patients with type II diabetesmellitus: randomized controlled trial comparing blood and urinetesting. Diabetes Care. 1990;13:1044-1050.

47. Miles P, Everett J, Murphy J, Kerr D. Comparison of blood orurine testing by patients with newly diagnosed non-insulindependent diabetes: patient survey after randomised crossovertrial. BMJ. 1997;315:348-349.

48. Cook D, Sackett D, Spitzer W. Methodological guidelines for sys-tematic reviews of randomized control trials in health care fromthe Potsdam consultation on meta-analysis. J Clin Epidemiol.1995;48:167-171.

49. Montori V, Swiontkowski M, Cook D. Methodological issues insystematic reviews and meta-analysis. Clin Orthop. 2002;412:43-54.

50. Higgins JPT, Green S, eds. Cochrane Handbook for SystematicReviews of Interventions 4.2.6 [updated September 2006]. In: TheCochrane Library, Issue 4. Chichester, UK: John Wiley; 2006.

51. Kraemer H, Wilson GT, Fairburn CG, Agras WS. Mediators andmoderators of treatment effects in randomized clinical trials. ArchGen Psychiatry. 2002;59:877-883.

52. Peyrot M, Rubin RR. Modeling the effect of diabetes educationon glycemic control. Diabetes Educator. 1994;20:143-148.

53. Diamond E, Massey K, Covey D. Symptom awareness and bloodglucose estimation in diabetic adults. Health Psychol. 1989;8:15-26.


1009

McAndrew et al

TDE309807.qxd 11/7/2007 11:17 AM Page 1009

54. Peel E, Parry O, Douglas M, Lawton J. Blood glucose self-monitoring in non-insulin treated type 2 diabetes: a qualitativestudy of patients’ perspectives. Br J Gen Pract. 2004;54:183-188.

55. Reichard P. To be a teacher, a tutor and a friend: the physician’srole according to the Stockholm Diabetes Intervention Study(SDIS). Patient Educ Couns. 1996;29:231-235.

56. DAFNE Study Group. Training in flexible, intensive insulin man-agement to enable dietary freedom in people with type 1 diabetes:Dose Adjustment for Normal Eating (DAFNE) randomized con-trolled trial. BMJ. 2002;325:746-751.

57. Russell L, Suh D, Safford M. Time requirements for diabetes self-management: too much for too many? J Fam Pract. 2005;54:1.

58. Hanninen J, Takala J, Keinanen-Kiukaanniemi S. Quality of lifein NIDDM patients assessed with the SF-20 questionnaire.Diabetes Res Clin Pract. 1998;42:17-27.

Addendum

Since the completion of this systematic review of self-monitoring of blood glucose (SMBG) for patients withtype 2 diabetes, one of the largest randomized control tri-als of SMBG was published in the British MedicalJournal.1 This study provides an important contribution tothe literature in that it is a well-designed trial thataddresses many of the limitations of previous SMBGinterventions, including sufficient power to detect clini-cally meaningful differences in HbA1c levels. In addition,it examined not only SMBG but also SMBG within abehavioral framework using psychological theory. Farmerand colleagues1 randomized 453 patients to 1 of 3 arms: acontrol where patients and their doctors received HbA1creadings prior to an appointment; an SMBG interventionthat asked patients to monitor blood glucose 3 times a day,twice a week and talk to their doctor for interpretation;and an intense SMBG intervention that asked patients tomonitor their glucose levels at the same rate and taughtthem skills for personal interpretation. All patientsreceived a behavioral goal-setting intervention based onLeventhal’s commonsense model of self-regulation.Although the intense self-monitoring group showed a0.17% reduction in HbA1c, the between-group differencefor HbA1c did not reach statistical significance.

Although Farmer and colleagues1 should be com-mended for their well-thought-out and timely study, ofconcern is the discrepancy between HbA1c improvementfound in this trial and the HbA1c improvement found inother trials. Of the 11 randomized control trials (RCTs)examined in this systematic review, only 1 found lessimprovement in HbA1c levels for the SMBG arm,2 withthe majority showing at least 3 times the level of improve-ment. The results presented here suggest that patients

who are in poor control are the ones who may benefit themost from SMBG. The mean baseline HbA1c level of7.5% of the patients recruited by Farmer and colleagues1

may have limited the ability to detect improvement frommonitoring. Consistent with this possibility, Farmer andcolleagues’ reporting of subgroups suggests that patientsin their study who were in poorer control benefited themost from the SMBG intervention, as patients on oralmedication had a greater reduction in HbA1c levels thanthose treated with diet alone (this effect was not statisti-cally tested).

Further analysis of mediating and moderating vari-ables may provide clues as to why intensive interventionpatients did not show significant improvements inHbA1c levels. Specifically, patients need to substituteobjective readings (SMBG) for subjective cues to assessblood sugar levels. Participants in Farmer and col-leagues’ intensive intervention, however, were reportingsignificantly more subjective symptoms used to identifyhypoglycemia episodes.1 As patients concerned abouthypoglycemia have elevated glucose, intensive partici-pants may have experienced conflict between subjectivecues and objective indicators, accepted higher readings,and had the greatest difficulty with self-managementbased on objective readings. This is consistent with theirless frequent use of meters. In addition, patients appearmore likely to succeed in using SMBG to regulate dietand exercise if they monitor both before and 2 hours aftera meal.3 In the paper by Farmer and colleagues, patientswere described as monitoring either before or after ameal.4 It is unclear how patients could develop a behav-ioral regulatory system without specific feedback abouttheir diet and exercise activities.

Although the conclusion that SMBG may be effectivein controlling HbA1c for patients with type 2 diabetesnot treated with insulin is maintained here, the well-designed and adequately powered study by Farmer andcolleagues1 highlights the difficulties both for conductingtrials and for clinical practice. Attention to how (pre-and postaction), when (initially diagnosed), and forwhom (poorly controlled) can maximize the possibilitythat SMBG can help the patients achieve effective bloodsugar control. As patients develop a steady state of gooddiabetes control, they may use SMBG only as an occa-sional check. But as the patient’s physiology changes,causing a deterioration of control, the individual shoulduse SMBG to readjust his or her self-regulation.Recommendations to not use self-monitoring may have


1010


TDE309807.qxd 11/7/2007 11:17 AM Page 1010

small ramifications in the short term, but as patients con-tinue in maladaptive regulatory behaviors, the effectsmay be magnified. As such, trials should continue toexamine the effect of SMBG on patients with type 2 dia-betes not on insulin.

References

1. Farmer A, Wade A, Goyder E, et al. Impact of self monitoring ofblood glucose in the management of patients with non-insulintreated diabetes: open parallel group randomised trial. BMJ.2007;335:132.

2. Moreland E, Volkening L, Lawlor M, Chalmers K, Anderson B,Laffel L. Use of a blood glucose monitoring manual to enhancemonitoring adherence in adults with diabetes. Arch Intern Med.2006;166:689-695.

3. Schwedes U, Siebolds M, Mertes G, Group SS. Meal-relatedstructured self-monitoring of blood glucose: effect on diabetescontrol in non-insulin-treated type 2 diabetic patients. DiabetesCare. 2002;25:1928-1932.

4. Farmer A, Wade A, French D, Goyder E, Kinmonth A, Neil A.The DIGEM trial protocol: a randomised controlled trial to deter-mine the effect on glycaemic control of different strategies ofblood glucose self-monitoring in people with type 2 diabetes.BMC Fam Pract. 2005;6:25.


1011

McAndrew et al

TDE309807.qxd 11/7/2007 11:17 AM Page 1011

does patient blood glucose monitoring improve diabetes

Documents