long-term glycemic variability and risk of adverse ... · group or whether variability was...

Long-term Glycemic Variabilityand Risk of Adverse Outcomes:A Systematic Review andMeta-analysisDiabetes Care 2015;38:2354–2369 | DOI: 10.2337/dc15-1188

OBJECTIVE

Glycemic variability is emerging as a measure of glycemic control, which may be areliable predictor of complications. This systematic review and meta-analysisevaluates the association between HbA1c variability and micro- and macrovascularcomplications and mortality in type 1 and type 2 diabetes.

RESEARCH DESIGN AND METHODS

Medline and Embase were searched (2004–2015) for studies describing associa-tions between HbA1c variability and adverse outcomes in patients with type 1 andtype 2 diabetes. Data extraction was performed independently by two reviewers.Random-effects meta-analysis was performed with stratification according to themeasure of HbA1c variability, method of analysis, and diabetes type.

RESULTS

Seven studies evaluated HbA1c variability among patients with type 1 diabetesand showed an association of HbA1c variability with renal disease (risk ratio 1.56[95% CI 1.08–2.25], two studies), cardiovascular events (1.98 [1.39–2.82]), andretinopathy (2.11 [1.54–2.89]). Thirteen studies evaluated HbA1c variabilityamong patients with type 2 diabetes. Higher HbA1c variability was associated withhigher risk of renal disease (1.34 [1.15–1.57], two studies), macrovascular events(1.21 [1.06–1.38]), ulceration/gangrene (1.50 [1.06–2.12]), cardiovascular disease(1.27 [1.15–1.40]), and mortality (1.34 [1.18–1.53]). Most studies were retrospectivewith lack of adjustment for potential confounders, and inconsistency existed in thedefinition of HbA1c variability.

CONCLUSIONS

HbA1c variability was positively associated with micro- and macrovascular com-plications and mortality independently of the HbA1c level and might play a futurerole in clinical risk assessment.

Current management of type 1 and type 2 diabetes uses the average glycemiameasure HbA1c to monitor control. This rationale is based on trial and observationalevidence that lowering HbA1c reduces the risk of the micro- and macrovascularcomplications of diabetes (1–4). Whether an average glycemic measure is mostappropriate to assess the risk for complications is currently under debate. Forexample, one analysis of the Diabetes Control and Complications Trial indicatedhigher rates of retinopathy in the conventional treatment group than in the

1Institute of Population Health, Centre for Pri-mary Care, University of Manchester, Manches-ter, U.K.2Royal Stoke University Hospital, University Hos-pitals of North Midlands, Stoke-on-Trent, U.K.3Keele Cardiovascular Research Group, Instituteof Science and Technology in Medicine and In-stitute of Primary Care and Health Science, Uni-versity of Keele, Stoke-on-Trent, U.K.4Central University Hospitals NHS FoundationTrust, Manchester Academic Health Science Cen-tre, Manchester, U.K.5Farr Institute, University of Manchester, Man-chester, U.K.6Epidemiology Group, Institute of Applied HealthSciences, School of Medicine, Medical Sciences,and Nutrition, University of Aberdeen, Aberdeen,Scotland, U.K.7University East Anglia, Norwich, Norfolk, U.K.8Manchester Diabetes Centre, Central Manches-ter University Hospitals NHS Foundation Trust,Manchester Academic Health Science Centre,Manchester, U.K.9Centre for Endocrinology andDiabetes, Instituteof Human Development, University of Manches-ter, Manchester, U.K.

Corresponding author: CatherineGorst, [email protected].

Received 13 June 2015 and accepted 22 Septem-ber 2015.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc15-1188/-/DC1.

© 2015 by the American Diabetes Association.Readersmayuse this article as longas thework isproperly cited, the use is educational and not forprofit, and the work is not altered.

Catherine Gorst,1 Chun Shing Kwok,2,3

Saadia Aslam,4 Iain Buchan,5

Evangelos Kontopantelis,1 Phyo K. Myint,6

Grant Heatlie,2 Yoon Loke,7

Martin K. Rutter,8,9 and

Mamas A. Mamas2,3,5

2354 Diabetes Care Volume 38, December 2015

META-ANALYSIS

http://crossmark.crossref.org/dialog/?doi=10.2337/dc15-1188&domain=pdf&date_stamp=2015-11-11

mailto:[email protected]

mailto:[email protected]

http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc15-1188/-/DC1


intensive treatment group over time inpatients with similar average HbA1c val-ues in the two groups (5), suggestingthat additional factors other thanmean HbA1c may be responsible forthis increased retinopathy risk (5–7).Glycemic variability is now emergingas a possible additional measure of gly-cemic control, which may be a betterpredictor of complications than averageglycemic measures.Glycemic variability relates to fluctu-

ations in glycemia. Short-term glycemicvariability refers to within- or between-day fluctuations in an individual and in-cludes multiple methods of assessment.Long-term glycemic variability refers tofluctuations over several weeks ormonths and is most commonly assessedby HbA1c variability. However, neitherhave a standardized method of mea-surement or definition (8). A recentmeta-analysis concluded that HbA1c var-iability, assessed by the SD, is associatedwith renal disease in type 1 and type 2diabetes (9). However, no systematic re-views or meta-analyses have evaluatedthe relationship between long-term gly-cemic variability and other complica-tions in diabetes, despite contradictoryliterature providing evidence in sup-port (6,10–15) and against (16–20) such arelationship.Long-term glycemic variability is im-

portant for several reasons. First, unlikeshort-term glycemic variability, long-term glycemic variability may predictcomplications in both type 1 and type 2diabetes (6,10–15,21–29). Second, HbA1cis routinely recorded in primary care forboth types of diabetes, whereas mea-sures of short-term variability are not(30,31). Finally, it could be a potentiallymodifiable risk factor. Through a system-atic review and meta-analysis, we evalu-ated the evidence for the association ofHbA1c variability withmortality and com-plications in type 1 and type 2 diabetesto gain insight into the clinical utility ofthis relationship to predict adverseoutcomes.

RESEARCH DESIGN AND METHODS

Data Sources and SearchesWe searched Medline and Embase forarticles published between 2004 andSeptember 2014, using the search termsshown in the Supplementary Data. Weupdated the search in July 2015. C.S.K.conducted the initial search, and C.G.

duplicated it. All resulting articles, in-cluding conference abstracts, were re-viewed. Broad search criteria includeddiabetes terms, outcomes of interestterms, and exposure terms (HbA1c

variability).

Study SelectionWe included studies of patients with di-abetes that evaluated HbA1c variabilityand adverse outcomes published withinthe past 10 years. No restrictions wereplaced on participant age or definitionof HbA1c variability used. The main ad-verse outcomes of interest were renaldisease (diabetic nephropathy, microal-buminuria, macroalbuminuria, renalfailure, chronic kidney disease), diabeticretinopathy, diabetic neuropathy, cardio-vascularmacrovascular events (myocardialinfarction, ischemic heart disease, heartfailure, stroke, peripheral vascular dis-ease), and death. We excluded reviews,editorials, and case reports and searchedthe bibliographies of included studiesand relevant reviews for additional stud-ies. Study titles and abstracts were ini-tially screened independently by tworeviewers (C.G. and S.A.), and full articleson potentially relevant studies weredownloaded and reviewed for inclusion.Five reviewers discussed and decided onthe final inclusion of studies for this re-view andmeta-analysis (C.G., C.S.K., S.A.,E.K., M.A.M.) (Fig. 1).

Data Extraction and QualityAssessmentData extraction was performed inde-pendently by two reviewers (C.G. andS.A.). Data collected were study design,participant characteristics, quality ofstudy assessment, definitions of HbA1cvariability, outcomes evaluated, and re-sults. Discrepancies in extractions werediscussed with two other reviewers (C.S.K.and Y.L.).

Data Synthesis and AnalysisWe conducted a random-effects meta-analysis of the adjusted risk estimates(where available) with use of the inversevariance method in RevMan 5.3 (NordicCochrane Centre). Analysis was stratifiedaccording to the definition of HbA1c vari-ability used, the method of analysis used,and the type of diabetes. In terms ofHbA1c variability, studies were dividedinto those that reported a coefficient ofvariation (CV) and those that reportedan SD. Within the two groups, the

analysis was further divided accordingto whether the highest variability groupwas comparedwith the lowest variabilitygroup or whether variability was mea-sured per incremental increase in CV orSD. Where possible, we chose to analyzeresults for the group with the greatestHbA1c variability against that of the onewith lowest variability. If there were sev-eral groups with differing levels of vari-ability, we conducted the meta-analysisbased on the group with the greatestvariability compared with the one withthe least variability.

Both SD and CV are measures of var-iability. SD measures how much valuesdiffer from the group mean. CV is theratio of SD to the mean, so it is a mea-sure independent of the mean. The CVmay be appropriate for parameters suchas HbA1c where the variability is likelyto increase as the mean increases.However, there is no standardizedmethod of measuring HbA1c variability(8).

Where there were insufficient studiesfor pooling or significant heterogeneitythat could not be explained, we per-formed narrative synthesis. We as-sumed similarity between risk ratios(RRs) and odds ratios (ORs) because ad-verse events are rare (32).

Statistical heterogeneity was as-sessed with the I2 statistic (33), wherevalues of 30–60% represent a moderatelevel of heterogeneity. Six sensitivityanalyses were performed. These in-cluded prospective studies; studieswith a follow-up of .5 years; and stud-ies that adjusted for duration of diabe-tes, number of HbA1c measurements,comorbidities, and baseline medica-tions. Publication bias was assessed us-ing funnel plots if there were .10studies and no evidence of statisticalheterogeneity in a particular meta-analysis (34).

RESULTS

Studies Included and ParticipantCharacteristicsFigure 1 shows a flow diagram of thestudy selection process. Twenty studiesin 87,641 participants met the inclusioncriteria. Ten studies included participantsfrom Europe (10,11,14,16–19,26,29,35,36), eight from Asia (12,13,15,18,20,24,27,28), four from North America (6,18,23,25), and one from Australasia (18). Thenumber of participants in each study

care.diabetesjournals.org Gorst and Associates 2355


http://care.diabetesjournals.org

ranged from 234 to 35,891. Details ofthe study design and participants areshown in Table 1.

Type 1 Diabetes

Seven studies included 44,021 participantswith type 1 diabetes (6,10,11,14,25,29,35).These comprised three retrospectivecohort studies (11,14,25), two pro-spective cohort studies (10,29), onepost hoc analysis of a randomized con-trolled trial (6), and one cross-sectionalstudy (35). Most studies used datafrom secondary care apart from two(10,11) that used primary and second-ary care data.

Type 2 Diabetes

Thirteen studies included 43,620 partic-ipants with type 2 diabetes (12,13,15–20,23,24,26–28,36). These comprisedsix retrospective cohort studies (13,15,20,23,28,36), five prospective cohortstudies (12,16,17,19,26,27), and twopost hoc analyses of randomized con-trolled trials (18,24). All studies used sec-ondary care data apart from one ofprimary and secondary care data (19)

and one of solely U.S. primary caredata (23).

Quality Assessment of IncludedStudiesThe quality assessment of includedstudies is shown in Table 2. For bothtype 1 and type 2 diabetes, the out-come assessment varied from bloodand urine tests for diabetic nephropa-thy, to fundoscopy for retinopathy, to for-mal follow-up for cardiovascular eventsand death. The frequency of outcomeevaluation differed depending on thestudy. All studies adjusted for meanHbA1c.

Type 1 Diabetes

The shortest follow-up was a mean of 5.2years (11) and the longest, 23 years (14).The number of HbA1c measurements perpatient ranged from amedian of 4 (29) to13 (10). Data from all studies were un-clear about loss to follow-up. All studiesused some form of adjustment for base-line covariates; however, five did not ad-just for baseline diabetes medications

(10,11,14,25,29), and none adjusted forbaseline hypertension medication.

Type 2 Diabetes

The shortest follow-up was 2 years (20)and the longest, a median of 15.9 years(13). The number ofHbA1cmeasurementsper patient ranged from 3 (19) to a me-dian of 79 (13). In six studies, loss tofollow-up was unclear; six studies had,10% of participants lost to follow-up,whereas one had lost 27.5% to follow-up (13). All the studies used some formof adjustment for baseline covariates;however, six did not adjust for baselinediabetes medication (15,19,20,23,24,27),and four did not adjust for baselinehypertension medication (13,16,17,20,27). Of the seven studies that did in-clude hypertension medication (12,15,18,23,24,26,28), two adjusted for ACEinhibitor/angiotensin receptor blockeruse (23,24). The definition of glycemicvariability, outcome evaluated, andstudy follow-up and results are shownin Table 3.

Adverse Outcomes

Type 1 Diabetes

Three studies evaluated adverse out-comes by considering the impact ofHbA1c CV (Supplementary Fig. 2)(11,14,35). There was no significantassociation between HbA1c CV and reti-nopathy (RR [95% CI] 1.34 [0.89–2.04], two studies) or microalbuminuria(1.04 [1.00–1.08], one study). The studyby Hermann et al. (14), however,reported that HbA1c variability basedon CVwas associatedwith a 3.5% higherrisk of diabetic retinopathy per 1-unitincrease in HbA1c CV at 10 years diseaseduration.

Four studiesevaluatedadverseoutcomesassociated with HbA1c SD (SupplementaryFig. 2) (6,10,25,29). All showed a signif-icant association of HbA1c SD and ad-verse outcomes. Highest to lowestvariation SD group was associatedwith an increased risk of nephropathy(RR [95% CI] 1.92 [1.49–2.47]) andcardiovascular events (1.98 [1.39–2.82]). Incremental increases in SDwere also associated with increasedrisk of nephropathy (1.86 [1.41–2.46]), microalbuminuria (1.56 [1.08–2.25], two studies), and retinopathy(2.11 [1.54–2.89]).

No studies evaluated HbA1c variabilityin type 1 diabetes and mortality. Sensi-tivity analyses for study type and studies

Figure 1—Flow diagram of study selection.

2356 HbA1c Variability and Adverse Outcomes Diabetes Care Volume 38, December 2015




Table

1—Designandparticipantch

aracteristicsofstudiesthat

eva

luatedglyce

mic

variability

StudyID

Studydesign;year;country

Samplesize

Age

Male(%

)Inclusioncriteria

Studiesofparticipan

tswithtype1

diabetes

Hermann20

14(14)

Retrospective

cohortstudy;19

90–March

2013

;GermanyandAustria

35,891

Med

ian16

years

52Participan

tsin

theGerman/Austrian

Diabetes

Prospective

Documen

tation

Initiative

Hietala20

13(11)

Retrospective

cohortstudy;19

97–January

2012

;Finland

2,01

9Nopriorlasertreatm

entgroup:

35years

Proliferativediabeticretinopathy

group:39

years

49Adultparticipants

withtype1diabetes

who

tookpartin

FinnDiane

Kilpatrick

2008

(6)

PosthocanalysisofRCT;19

83–19

93;U

.S.and

Can

ada

1,44

127

years

53Participan

tsin

theDiabetes

Controland

ComplicationsTriald

ataset

Marcovecchio

2011(29)

Prospective

cohortstudy;19

86–19

96and

2000

–20

05;U.K.

1,23

2Med

ianat

diagn

osis9years

55Participan

tsin

theOxford

Regional

Prospective

Study,whichincluded

children

withtype1diabetes

aged

,16

years,and

theNep

hropathyFamily

Study,which

included

adolescen

tswithtype1diabetes

aged

10–16

years

Nazim

2014

(35)

Cross-sectionalstudy;19

85–20

04;Po

land

438

Meanat

diagnosis9years

55Childrenandadolescen

tswithnew

lydiagnosedtype1diabetes

under

thecare

oftheEndocrinology

Dep

artm

entof

University

Children’sHospital

Ram

an20

11(25)

Retrospective

cohortstudy;19

93–20

09;U

.S.

893

Meanat

diagnosis8years

47Pediatricpatients

withtype1diabetes

ina

singlelargetertiary

care

referralcenter

Wad

en20

09(10)

Prospective

cohortstudy;November

1997–

January20

09;Finland

2,10

736

years

53Participan

tswithtype1diabetes

diagnosed

atage,35

yearsinFinnDiane,withinsulin

treatm

entinitiatedwithin

1year

of

diagnosis

Studiesofparticipan

tswithtype2

diabetes

Lin20

13(28)

Retrospective

cohortstudy;Augu

st20

02–

Augu

st20

08;Taiwan

3,22

057

years

51Participan

tswithtype2diabetes

treatedat

theChinaMed

icalUniversity

Hospital

Cummings

2011

(23)

Retrospective

cohortstudy;19

98–20

08;U

.S.

791

54years

32Participan

tswithtype2diabetes

aged

.18

yearsseen

inoneoftheprimarycare

practices

(fam

ilymed

icine,

internal

med

icine)

inthesoutheasternU.S.

Foo20

14(20)

Retrospective

cohortstudy;notstated

;Singapore

234

Notstated

Notstated

Participan

tsattendingatertiary

eyehospital

inSingapore

withserialHbA1cmonitoring

for.2years

Hirakaw

a20

14(18)

PosthocanalysisofRCT;

November

2001–

2007

;Asia,Australasia,Europe,

North

America

4,39

966

years

43Participan

ts.55

yearsofagewithmajor

macrovascularormicrovasculardisease

or

atleastonevasculardisease

risk

factor

from

1of21

5collaboratingcentersofthe

ADVANCEtrial

Con

tinuedon

p.23

58



Table

1—Continued

StudyID

Studydesign;year;country

Samplesize

Age

Male(%

)Inclusioncriteria

Hsu

2012

(24)

PosthocanalysisofRCT;20

03–20

10;Taiwan

821

Atonsetofdiabetes

51years

46Participan

tswithtype2diabetes

enrolledin

theDiabetes

Managem

entThrough

anIntegrated

DeliverySystem

project

Lang20

15(36)

Retrospective

cohortstudy;notstated

;Scotland

1,70

1Med

ian74

years

60Participan

tswithtype2diabetes

andinciden

tchronicheartfailure

Luk20

13(12)

Prospective

cohortstudy;July19

94–20

09;

HongKo

ng

8,43

9NoCKD

atbaseline:

7,18

4NoCVDat

baseline:

6,98

3

58years

47Participan

tsin

theHongKo

ngDiabetes

Registry

Patien

tswithbaselineCKD

wereexcluded

intheanalysisoftherenalen

dpoint,and

patientswithbaselineCVDwereexcluded

intheanalysisofcardiovascularen

dpoints

Ma20

12(15)

Retrospective

cohortstudy;20

03–20

10;

Taiwan

881

60years

48Participan

tsin

theDiabetes

Shared

Care

Program

attheCardinalTien

Hospitaland

attendingclinicapproximatelyevery

3months

Penno20

13(16,17

)Prospective

cohortstudy;20

07–20

08;

Italy

8,26

0Med

ian68

years

57Participan

tsincluded

intheRIACEItalian

Multicen

treStudywithHbA1cvalues

of

3–5%

measuredseriallyina2-year

period

Rodrıguez-Segade20

12(26)

Prospective

cohortstudy,March

1994–

March

2009

;Spain

2,10

359

years

48Participan

tswithdiabetes

attending

outpatientclinicso

ftheUniversityHospital

ComplexofSantiagodeCompostela

Skriver20

15(19)

Prospective

cohortstudy;19

70–20

10;

Den

mark

11,205

Med

ian64

years

52Participan

tswithtype2diabetes

with

registered

public

datafilesin

Aarhus

County,D

enmark,whosubsequen

tlyhad

atleastthreeHbA1cmeasuremen

tsSugawara20

12(27)

Prospective

cohortstudy;20

00–20

07;

Japan

812

55years

69Participan

tswithtype2diabetes

inthe

TsukubaKawaiDiabetes

Registrydatabase

Takao20

14(13)

Retrospective

cohortstudy,19

95–20

12;

Japan

754

54years

82Participan

tswithtype2diabetes

attending

anoutpatientclinicandfollowed

upfor

2yearswithat

leastfourHbA1clevels

CKD

,chronickidney

disease;CVD,cardiovasculardisease;FinnDiane,

FinnishDiabeticNep

hropathyStudy;RCT,randomized

controlledtrial;RIACE,Ren

alInsufficien

cyAndCardiovascularEven

ts.


Table

2—Riskofbiasamongstudiesthateva

luatedglyce

mic

variab

ilityandadve

rseoutcomes

StudyID

Timefram

eandnumber

ofsamplesusedto

defi

neHbA1cvariability

Casedefi

nition,ascertainmen

t,and

assessmen

tfreq

uen

cy,10

%loss

tofollow-up

Adjustmen

tsforpotentialconfounders

Studiesofparticipan

tswithtype1

diabetes

Hermann20

14(14)

Between19

90andMarch

2013

Med

iannumber

ofHbA1cvalues

per

patient

during1year:4.3

Diabeticretinopathyaccordingto

trained

ophthalmologistdirectfundoscopyin

mydriasisto

grad

econditionbased

on

modified

Airlie

House

Classification/ETD

RS

stan

dards

Freq

uen

cyofevaluationunclear

Unclear

Age

atdiabetes

diagn

osis,sex,and

med

ianHbA1c

Hietala20

13(11)

Average

follow-up:5.2years

10(IQR3–

18)HbA1cmeasuremen

tsper

patient

Proliferativeretinopathyfrom

fundus

photographsand/orrecordsofdilatedslit

lampfundusexam

inationperform

edbyan

ophthalmologist

Photograph

stakenforamedianof3(IQR1–5)

occasion

sperpatient

Proliferative

retino

pathydefinedas

$61

onETDRSgradingscale

Unclear

Ren

alstatus,diabetes

duration,m

ean

HbA1c,bloodpressure,sex,and

number

ofHbA1cmeasuremen

ts

Kilpatrick

2008

(6)

Average

follow-up:6.5years

HbA1cwas

measuredquarterlybutnumber

of

HbA1cmeasuremen

tsper

patientunclear

Developmen

tandprogressionofdiabetic

retinopathydefi

ned

asachan

gefrom

baselineof$3unitsontheETDRSinterim

score

onanytw

osuccessive

annual

evaluations

Nep

hropathydefi

ned

asan

increase

inAER

$40

mg/24

honanyannualevaluation

provided

that

baselineAER

was

,40

mg/dL

Unclear

Age,sex,disease

duration,ran

domization

treatm

ent,preventioncohort,and

baselineHbA1c

Marcovecchio

2011(29)

Between19

86and19

96andbetween20

00and

2005

Med

iannumber

ofHbA1cassessmen

ts:

4(2–16

)

Microalbuminuriawas

defi

ned

asACR3.5–35

mg/mmolinmen

and4.0–

47mg/mmolin

women

intw

oofthreeconsecutive

early

morningurinesamplesmeasuredannually

Unclear

Sex,ageat

diagnosis,chronologicage,

andmeanHbA1c

Nazim

2014

(35)

Follow-up:9.2years

Number

ofHbA1cmeasuremen

tsunclear

Microalbuminuriadefi

ned

asAER

$20

mg/min

and,20

0mg/min

inat

leasttw

osamplesobtained

within

twoormore

samplesobtained

within

aperiodof3–6

months

Freq

uen

cyofurinetestingunclear

Unclear

Age

atonsetofdiabetes,p

resence

of

arterialhypertensionatbaseline,mean

HbA1c,andmeaninsulin

dailydose

Ram

an20

11(25)

Average

follow-up:7years

Number

ofHbA1cmeasuremen

tsunclear

Microalbuminuria(AER

$20

mg/min

or

microalbumin:creatinineratio$30

mg/g)

Freq

uen


Unclear

Age,sex,race,andmeanHbA1c

Con

tinuedon

p.23

60



Table

2—Continued

StudyID

Timefram

eandnumber

ofsamplesusedto

defi

neHbA1cvariability

Casedefi

nition,ascertainmen

t,and

assessmen

tfreq

uen

cy,10

%loss

tofollow-up

Adjustmen


Wad

en20

09(10)

Med

ianfollow-up:5.7years

Med

iannumber

ofHbA1cmeasuremen

tsper

patient:13

(IQR7–20

),2.13measuremen

tsper

patientper

year

Ren

alstatusp

rospectivelyassessed

byreview

ofallrecorded

values

ofurineAER

and

med

icalrecords

Progressionofrenaldisease

defi

ned

asa

shiftto

ahigher

albuminurialevelinany

two(ofthree)

consecutive

urine

collectionsoren

d-stage

renalfailure

Cardiovasculareven

ts(m

yocardial

infarction,coronaryartery

procedure,

stroke,lim

bam

putationdueto

ischem

ia,

peripheralartery

procedure)based

on

med

icalrecordsat

baselineandfollow-up

Freq

uen


Unclear

Durationofdiabetes,sex,bloodpressure,

totalcholesterol,sm

oking,

intrapersonalmeanofserialHbA1c

measuremen

ts,n

umber

ofHbA1c

measuremen

ts,diabeticnep

hropathy,

andbaselinecardiovasculareven

ts

Studiesofparticipan

tswithtype2

diabetes

Lin20

13(28)

Average

follow-up:4.40years

Patien

tshad

tohavemore

than

twoHbA1c

measuremen

tseach

year

Although

notreported

,patientslikelyhad

more

than

eigh

tHbA1cmeasuremen

ts

Diabeticnep

hropathydefi

ned

aseG

FR,60

mL/min/1.73m

2andpatientsfollowed

up

regu

larlyevery3–6months

Yes

Age,sex,lifestylefactors,comorbidities,

myocardialinfarction,m

eanfasting

plasm

aglucose,m

eanHbA1c,anddrug

treatm

ents

Cummings

2011

(23)

Average

follow-up:7.6years

Patien

tshad

tohaveat

leastfive

HbA1c

measuremen

ts

Increase

ofo

neormore

CKD

stages

based

on

baselineandmostrecentfollow-upvisit

Unclear

Age,race,sex,durationofdiabetes,blood

pressure,d

rugtreatm

ents,initial

HbA1candnumber

ofHbA1cvalues,

andfastingbloodglucose

CV

Foo20

14(20)

Follow-upof2

yearswithserial3-monthlyHbA1c

(ran

ge3–6)

values

per

patient

Moderatediabeticretinopathyorworse

assessed

usingretinalphotograp

hsofboth

eyes

withan

ETDRSlevel$

43after2years

ofHbA1cmeasuremen

t

Unclear

Age,sex,e

thnicity,durationofdiabetes,

hypertension,h

yperlipidem

ia,

smoking,microalbuminuriaand

cardiovasculareven

ts,andmeanand

SDofHbA1c

Hirakaw

a20

14(18)

Med

ianstudyfollow-up:3years

Five

HbA1cmeasuremen

tsper

patient

Compositeoutcomes

ofm

ajorm

acrovascular

even

ts(death

from

cardiovascularcause,

nonfatalm

yocardialinfarction,ornonfatal

stroke),majormicrovasculareven

ts(new

orworsen

ingnep

hropathyorretinopathy)

andall-cause

mortality

Patien

tsfollowed

upforfirst24

months

Freq

uen


Yes

Age,sex,randomized

bloodpressure

lowering,region,durationofdiabetes,

smokingstatus,alcoholintake,systolic

bloodpressure,totalcholesterol,log-

tran

sform

edtriglycerides,B

MI,

med

ications,andmeanHbA1cor

fastingglucose

inthefirst24

months

Con

tinuedon

p.23

61


Table

2—Continued

StudyID

Timefram

eandnumber

ofsamplesusedto

defi

neHbA1cvariability

Casedefi

nition,ascertainmen

t,and

assessmen

tfreq

uen

cy,10

%loss

tofollow-up

Adjustmen


Hsu

2012

(24)

Average

follow-up:6.2years

Bloodcollected

every6monthsbutnumber

of

HbA1cmeasuremen

tsper

patientunclear

Microalbuminuriadefi

ned

asACR$3.4

mg/mmolintw

oconsecutive

urinetests

Freq

uen


Yes

Age

atdiabetes

onset,sex,ed

ucation,

diabetes

duration,sm

okingstatus,

waistcircumference,serum

lipids,

meanHbA1c,bloodpressure,andACE

inhibitororangiotensinreceptor

blocker

use

Lang20

15(36)

Med


Freq

uen

cyofevaluationornumber

ofHbA1c

measuremen

tsunclear

Methodofmortalityascertainmen

tunclear

Unclear

Significantcovariates,includingchronic

heartfailure

durationandcurren

tdrug

exposure

Luk20

13(12)

Med


Med

ianHbA1cmeasuremen

ts:10

(IQR5–

17)

Inciden

tCKD

(eGFR

,60

mL/min/1.73m

2)

andinciden

tCVD(m

yocardialinfarction,

ischem

icheartdisease,p

eripheral

vasculardisease,h

eartfailure,ischem

icstroke)anden

d-stage

renaldisease

obtained

from

HospitalAuthority

discharge

diagnoses

Unclear

Age,sex,sm

okinghistory,diabetes

duration,BMI,waistcircumference,

bloodpressure,serum

lipids,logurine

ACR,eG

FR,h

emoglobin,and

med

icationuse

Ma20

12(15)

Average

follow-up:4.7years

Average

numberofH

bA1cmeasuremen

ts:126

7Mortalityandcause

ofdeath

obtained

from

computerizeddeath

certificates

maintained

bytheDep

artm

entofHealth,

Executive

Yuan

,Taiwan

Yes

Age,sex,BMI,durationofdiabetes,blood

pressure,use

ofantihypertensivesand

statins,meanLD

Lcholesterol,sm

oking

status,CKD

,andmeanHbA1cvalues

Penno20

13(16,17

)Follow-upunclear

Average

number

ofHbA1cmeasuremen

ts:

4.526

0.76

Patien

tshad

tohavethreeto

five

HbA1c

measuremen

ts

Diabeticnep

hropathybyalbuminuriaand

eGFR

withunclearfreq

uen

cyofevaluation

Diabeticretinopathyassessed

atbaselineby

dilatedfundoscopy;follow-upevaluation

unclear

CVD(acute

myocardialinfarction;stroke;

footulcer

organgren

e;am

putation;

coronary,carotid,andlower-lim

brevascularization;andsurgeryforaortic

aneu

rysm

)assessed

from

med

icalrecords

andadjudicated

based

onhospital

discharge

recordsofspecialistvisitbyan

adhoccommitteein

each

center

Yes

Age,B

MI,sex,knowndisease

duration,

smokinghabits,triglycerides,H

DL

cholesterol,hypertension,

dyslipidem

ia,p

reviousmajorCVD

even

ts,specificdiabetes

treatm

ents,

andeG

FRandalbuminuriacategories

ifdiabeticretinopathywas

dep

enden

tvariable

ordiabeticretinopathy

categories

ifrenalparam

eterswere

dep

enden

tvariable

Rodrıguez-Segade20

12(26)

Average

follow-up:6.6years

Med

iannumber

ofHbA1cmeasuremen

tsper

patient:10

(IQR6–14

)

Progressionofdiabeticnep

hropathyifAER

$10

0mg/24

handhad

been,40

mg/24h

aten

tryorifAER

$30

0mg/24handhad

been,20

0mg/24

hat

entry

Freq

uen


Unclear

Age,d

urationofdiabetes,u

seofinsulin,

baselineHbA1c,BMI,retinopathy

status,use

ofantihypertensive

agen

ts,

smokingstatus,lipid

status,sex,

cohort,n

umber

ofHbA1c

measuremen

ts,andupdated

mean

Con

tinuedon

p.23

62



that adjusted for duration of diabetes,number of HbA1c measurements, co-morbidities, and baseline medicationsproduced similar results to those re-corded with inclusion of all studies(Supplementary Table 1).

Type 2 Diabetes

Studies reporting all-cause mortality asan outcomewere not pooled due to highlevels of heterogeneity, which was be-lieved to be a result of differing follow-up durations and loss to follow-up. Theoutcome, therefore, was split accordingto short follow-up (,5 years) and longfollow-up ($5 years).

Six studies evaluated adverse out-comes by considering the impact ofHbA1c CV (13,15,18,26,28,36), and ninestudies considered HbA1c SD (12,13,15–18,20,24,26,27). Increase in HbA1c vari-ability defined by high versus low CVgroups was associated with increasedrisk of diabetic nephropathy (RR [95%CI] 1.58 [1.19–2.10]) and all-cause mor-tality in studies with$5 years of follow-up (2.89 [1.45–5.74]) and in those with,5 years follow-up (1.06 [1.01–1.11])(Supplementary Fig. 3A). Incrementalincreases in CV were also associatedwith a significantly increased risk of ne-phropathy (1.03 [1.01–1.05]), macro/microvascular events (1.11 [1.02–1.21]), macrovascular events (1.18[1.04–1.33]), and mortality with $5years of follow-up (1.10 [1.03–1.16])and ,5 years of follow-up (1.31 [1.16–1.48]). No significant association wasfound between incremental increase inCV and microvascular events (1.07[0.96–1.20]) (Supplementary Fig. 3B).

Considering HbA1c variability with SD,high versus low SD groupwas associatedwith increased risk of nephropathy (RR[95% CI] 1.24 [1.02–1.51]), all-causemortality (2.34 [1.48–3.71], two stud-ies), microalbuminuria (1.34 [1.15–1.57], two studies), macroalbuminuria(1.41 [1.03–1.93]), ulceration/gangrene(1.50 [1.06–2.12]), and mortality instudies with $5 years of follow-up(3.09 [1.45–6.58]) and in those with,5 years of follow-up (1.99 [1.11–3.55]) (Supplementary Fig. 4A). Incre-mental increase in SD was associatedwith an increased risk of nephropathy(1.22 [1.05–1.42], two studies), end-stage renal failure (1.53 [1.35–1.73]),microalbuminuria (1.20 [1.03–1.39]),macro/microvascular events (1.12

Table

2—Continued

StudyID

Timefram

eandnumber

ofsamplesusedto

defi

neHbA1cvariability

Casedefi

nition,ascertainmen

t,and

assessmen

tfreq

uen

cy,10

%loss

tofollow-up

Adjustmen


Skriver20

15(19)

Med

ianfollow-up:6years

Number

ofH

bA1cmeasuremen

tsper

patientwas

atleast3

All-cause

mortalityfrom

record

linkage

with

nationwideDanishCivilRegistration

System

Unclear

Age,sex,med

ications,priorCVD,

dem

entia,chronicpulm

onarydisease,

connective

tissuedisease,u

lcer

disease,m

ildliver

disease,hem

iplegia,

moderateto

severe

renaldisease,

diabetes

withen

d-organ

dam

age,any

tumor,leukemia,lym

phoma,

moderateorsevere

liver

disease,

metastaticsolid

tumor,AIDS,andindex

HbA1c

Sugawara20

12(27)

Average

follow-up:4.3years

Med

iannumber

ofHbA1cmeasuremen

tsper

patient:11

(5–12

)

Microalbuminuriawas

defi

ned

asACR$3.4

mg/mmolforat

leasttw

oofthree

measuremen

tsDuringfollow-upperiod,A

CRwas

evaluated

every6months

Yes

Age,sex,durationofdiabetes,b

lood

pressure,B

MI,serum

lipids,and

smokingstatus

Takao20

14(13)

Med

ianfollow-up:15

.9years

Med

iannumber

ofHbA1cper

patient:79

(40–

117)

Unclearmethodofmortalityascertainmen

tandfreq

uen

cyofevaluation

No,2

7.5%

lost

tofollow-up

Age,sex,meanHbA1c,number

ofHbA1c

measuremen

ts,d

urationofdiabetes,

BMI,bloodpressure,serum

lipids,and

smokingstatus

ACR,albumin-to-creatinineratio;AER

,albumin

excretionrate;CKD

,chronickidney

disease;CVD,cardiovasculardisease;eG

FR,estim

ated

glomerularfiltrationrate;ETDRS,EarlyTreatm

entofDiabetic

RetinopathyStudy;IQR,interquartile

range.






Table

3—Resu

ltsofstudiesthat

eva

luatedglyce

mic

variabilityandadve

rseoutcomes

StudyID

Defi

nitionofglycem

icvariability

Outcomes

evaluated

Studyfollow-up

Results

Studiesofparticipan

tswithtype1

diabetes

Hermann20

14(14)

HbA1cCV

Diabeticretinopathy

1990–March

2013

Coxproportionalhazardsmultipleregressionfordiabetic

retinopathywithHbA1cCVbased

onparticipants

above

orbelowthe50

thcentile(HR1.11[1.10–

1.12])

HbA1cvariability

ledto

anadditionalrise

inrisk

(3.5%

higher

risk

ofdiabeticretinopathyper

1-unitincrease

ofHbA1cCVat

10yearsofdiabetes

duration

Hietala20

13(11)

HbA1cCV

Proliferativeretinopathy

Incohortwithnopriorlaser

treatm

ent,mean5.26

2.2

years

Unclearin

other

cohort

Amongparticipantswithverified

retinopathystatusand

indicationsforlasertreatm

ent,FineandGray

regressionmodelforrisk

ofproliferativeretinopathy

accordingto

quartilesofHbA1cCV:

Firstquartile:HR1.00(referen

ce)

Secondquartile:HR1.3(0.97–

1.8)

Thirdquartile:HR1.5(1.1–2.0)

Fourthquartile:HR1.7(1.3–2.2)

FineandGrayregressionmodelforretinopathyam

ong

patientswithnopriorlasertreatm

entrequiringlaser

treatm

entbyHbA1cvariability

firstquartilevs.fourth

quartile(HR1.6[1.1–2.5])

Kilpatrick

2008

(6)

HbA1cSD

Developmen

tandprogressionofdiabetic

retinopathyandnep

hropathy

6.5years

Coxproportionalhazardsmultiple

regressionofrisk

of

retinopathywithHbA1cSD

(1%increase

SD):HR2.11

(1.54–

2.89)

Riskofn

ephropathywithHbA1cSD

(1%increase

SD):HR

1.86

(1.41–

2.47)

Marcovecchio

2011(29)

HbA1cSD

Microalbuminuria

Unclear

Coxproportionalhazardsmultipleregressionforrisk

of

developmen

tofmicroalbuminuriabyHbA1cSD

(for

every1-unitincrease

ineach

covariate):HR1.31

(1.01–

1.35)

Nazim

2014

(35)

HbA1cCV

Microalbuminuria

9.26

3.4years


of

developingfirstepisodeofm

icroalbuminuriabyHbA1c

CV(per

unitincrease):HR1.04(1.00–

1.08)

Ram

an20

11(25)

HbA1cSD

Microalbuminuria

7.006

2.85

years


regressionfor

microalbuminuriabyHbA1cSD

(per

unitincrease):HR

1.91

(1.37–

2.66)

Wad

en20

09(10)

HbA1cSD

Cardiovasculareven

tandprogressionin

renalstatus(higher

albuminurialevel

inanytw

oofthreeconsecutive

urine

collectionsorto

end-stage

renal

disease)

Med



of

progressionin

renalstatusbyHbA1cSD

(defi

ned

accordingto

quartilesofHbA1cSD

):HR1.92

(1.49–

2.47)

Riskofcardiovasculareven

tbyHbA1cSD

(defi

ned

accordingto

quartilesofHbA1cSD

):HR1.98

(1.39–

2.82)

Con

tinuedon

p.23

64



Table

3—Continued

StudyID

Defi

nitionofglycem

icvariability

Outcomes

evaluated

Studyfollow-up

Results

Studiesofparticipan

tswithtype2

diabetes

Lin20

13(28)

HbA1cCV

Diabeticnep

hropathy

4.40years

Coxproportionalhazardsmultipleregressionfordiabetic

nep

hropathywithHbA1cCV,6.68(HR1.00

[referen

ce]),6

.68–13

.4(HR1.18[0.88–

1.58

]),

and.13

.4(HR1.58

[1.19–

2.11])

Cummings

2011

(23)

Average

excessofH

bA1c.7%

Increase

ofoneormore

CKD

stages

7.66

1.9years

Multiplelogisticregressionmodelofw

orsen

ingbyoneor

more

CKD

stages

withaverageexcess

ofHbA1c.7%

(OR1.173

[1.031

–1.335

])Foo20

14(20)

HbA1cSD

Moderatediabeticretinopathy

2years

Multivariable

logisticregressionformoderatediabetic

retinopathywithHbA

1cSD

(adjustedOR1.49

[0.72–3.07])

Hirakaw

a20

14(18)

HbA1cCVandSD

Composite

ofmajormacrovascular

(death

from

cardiovascularcause,

nonfatalm

yocardialinfarction,o

rnonfatalstroke)andmajor

microvascular(new

orworsen

ing

nep

hropathyorretinopathy)

even

tsMicrovasculareven

tsMacrovasculareven

tsAll-cause

mortality

Med

ian3years


regressionmodels

HbA1cCV1-SD

increase

andrisk

ofoutcomes:

Macro/m

icrovasculareven

ts:HR1.11(1.02–

1.21)

Majormacrovasculareven

ts:HR1.18(1.05–

1.34)

Majormicrovasculareven

ts:HR1.07(0.96–

1.2)

All-cause

mortality:HR1.31(1.16–

1.48)

ContinuousHbA1cSD

1-SD

increase

andrisk

of

outcomes:

Macro/m

icrovasculareven

ts:HR1.12(1.02–

1.22)

Majormacrovasculareven

ts:HR1.21(1.06–

1.38)

Majormicrovasculareven

ts:HR1.08(0.96–

1.21)

All-cause

mortality:HR1.34(1.18–

1.53)

Hsu

2012

(24)

HbA1cSD

Developmen

tofmicroalbuminuria

6.2years


regressionfor

inciden

ceofmicroalbuminuriawithHbA1cSD

quartiles:

Quartile1:

HR1.00(referen

ce)

Quartile2:

HR1.03(0.72–

1.48)

Quartile3:

HR1.09(0.75–

1.57)

Quartile4:

HR1.48(1.03–

2.12)

Lang20

15(36)

HbA1cCV

All-cause

mortality

Med

ianfollow-up:3.3(0.9–7.5)

years


regressionfor

mortalitywith0.01increase

inCV:

From

0.036

(Q1)

to0.046

:adjusted

HR1.04

(1.02–

1.07)

From

0.064

(Q2)

to0.074

:adjusted

HR1.03

(1.01–

1.05)

From

0.11(Q3)to

0.12

:adjusted

HR1.02(1.01–

1.03)

Con

tinuedon

p.23

65


Table

3—Continued

StudyID

Defi

nitionofglycem

icvariability

Outcomes

evaluated

Studyfollow-up

Results

Luk20

13(12)

HbA1cSD

IncidentCKD

(eGFR

,60

mL/min/1.73m

2),

incident

CVD(m

yocardialinfarction,

ischem

icheartdisease,periph

eral

vascular

disease,heartfailure,ischemic

stroke),andend-stagerenaldisease

Med



of

adverseoutcomewithadjusted

HbA1cSD

:Inciden

tCKD

:HR1.16

(1.10–

1.22)

End-stage

renalfailure:HR1.53

(1.35–

1.73)

Inciden

tCVD:HR1.27(1.15–

1.40

)Ma20

12(15)

HbA1cSD

andCV

All-cause

mortality

4.76

2.3years


of

all-cause

mortalitywith:

HbA1cSD

(.50

thcentilevs.,50

thcentile):HR1.99

(1.11–

3.54)

HbA1cCV(.

50th

centile

vs.,50

thcentile):HR1.06

(1.01–

1.11)

Penno20

13(16,17

)HbA1cSD

Diabeticnep

hropathybyalbuminuriaand

eGFR

Diabeticretinopathy

CVD;acute

myocardialinfarction;stroke;

footulcer

organgren

e;am

putation;

coronary,carotid,andlower-lim

brevascularization;andsurgeryfor

aorticaneu

rysm

Unclear

Multiple

logisticregressionofoutcomes

byHbA1cSD

quartiles:

Microalbuminuria

Quartile1:

OR1.00(referen

ce)

Quartile2:

OR1.03(0.878

–1.22)

Quartile3:

OR1.14(0.968

–1.35)

Quartile4:

OR1.31(1.10–

1.56)

Macroalbuminuria

Quartile1:

OR1.00(referen

ce)

Quartile2:

OR0.939

(0.672

–1.31)

Quartile3:

OR1.04(0.757

–1.44)

Quartile4:

OR1.41(1.03–

1.93)

eGFR

,60

mL/min/1.73m

2

Quartile1:

OR1.00(referen

ce)

Quartile2:

OR1.00(0.838

–1.20)

Quartile3:

OR1.23(1.03–

1.48)

Quartile4:

OR1.24(1.02–

1.51)

Multiple

logisticregression1%

increm

entofHbA1cSD

nonad

vanceddiabeticretinopathyvs.noretinopathy:

OR0.91

7(0.758

–1.11)

Multiple

logisticregressionofHbA1cSD

quartilesand

ulceration/gangren

e:Quartile1:

OR1(referen

ce)

Quartile2:

OR1.06

(0.736

–1.52)

Quartile3:

OR1.02

(0.709

–1.46)

Quartile4:

OR1.50

(1.06–

2.12)

Rodrıguez-Segade20

12(26)

HbA1cSD

andCV

Progressionofdiabeticnep

hropathy;if

AER

$10

0mg/24

handhad

been,40

mg/24hat

entry,orifAER

$30

0mg/24handhad

been,20

0mg/24h

aten

try

6.6years


of

progressionofnep

hropathyby:

HbA1cSD

(per

11mmol/mol[1%

]increase):HR1.37

(1.12–

1.69)

HbA1cCV:HR1.03(1.01–

1.04)

Con

tinuedon

p.23

66



[1.02–1.22]), macrovascular events (1.21[1.06–1.38]), cardiovascular disease(1.27 [1.15–1.40]), and mortality in stud-ies with $5 years of follow-up (3.17[1.43–7.03]) and in those with ,5 yearsof follow-up (1.34 [1.18–1.53]). No sig-nificant association was found betweenincremental increase in SD andmicrovas-cular events (1.08 [0.96–1.21]) or retinop-athy (1.03 [0.69–1.53], two studies)(Supplementary Fig. 4B).

A study by Penno et al. (17) reportedadditional nonsignificant associationswith any lower-limb vascular event,any cerebrovascular event, any coro-nary event, acute myocardial infarction,any cardiovascular disease, and stroke.This study could not be included in themeta-analysis because raw data werenot provided. Data on the significant as-sociation of HbA1c CV and all-cause mor-tality reported by Lang et al. (36) (RR[95% CI] 1.02 [1.01–1.03]) were not in-cluded in the meta-analysis because allparticipants had incident chronic heartfailure, increasing heterogeneity withother studies and affecting externalvalidity.

Cummings et al. (23) reported a sig-nificant worsening of one more chronickidney disease stages with an averageexcess HbA1c .7% (53 mmol/mol) (OR1.173 [95% CI 1.031–1.335]). Hirakawaet al. (18) also used other variabilitymeasures, including HbA1c variation in-dependent of the mean, HbA1c residualSD, and HbA1c average real variability.All were significantly associated withmacrovascular complications, macro/microvascular complications, and mor-tality based on data from ADVANCE (Ac-tion in Diabetes and Vascular Disease:Preterax and Diamicron MR ControlledEvaluation) (hazard ratio [HR] [95% CI]:variation independent of the meanHbA1c 1.17 [1.04–1.32], 1.11 [1.02–1.2], 1.30 [1.15–1.46]; residual SDHbA1c 1.20 [1.07–1.35], 1.10 [1.01–1.19], 1.33 [1.19–1.49]; average real var-iability HbA1c 1.21 [1.07–1.37], 1.11[1.02–1.21], 1.38 [1.22–1.55]). Skriveret al. (19) defined HbA1c variability asthe mean absolute residual around theline connecting index value and closingvalue. They reported that for indexHbA1c #8% (64 mmol/mol), variability.0.5 was associated with increasedall-cause mortality (HR 1.3 [95% CI 1.1–1.5]) per HbA1c percentage point vari-ability. However, for individuals with

Table

3—Continued

StudyID

Defi

nitionofglycem

icvariability

Outcomes

evaluated

Studyfollow-up

Results

Skriver20

15(19)

Meanabsolute

residual

aroundthelineconnecting

index

valueandclosing

value

All-cause

mortality

6years

Forindex

HbA1c#8%

(64mmol/mol),variability

.0.5

associated

withincreasedmortality(HR1.3[1.1–1.5])

per

HbA1cpercentage

pointvariab

ility

Forindividualswithindex

HbA1c.8%

(64mmol/mol),

noassociationbetweenHbA1cvariability

and

mortalitywas

iden

tified

Sugawara20

12(27)

HbA1cSD

Microalbuminuria

4.36

2.7years


of

microalbuminuriabyincrem

entalH

bA1cSD

(per

1-SD

increm

ent):HR1.20(1.03–

1.39

)Takao20

14(13)

HbA1cSD

andCV

All-cause

mortality

Med

ianfollow-up:15

.9years


of

all-cause

mortalitywithHbA1cSD

(HR3.17[1.43–

7.03

])andHbA1cCV(HR1.10

[1.04–

1.16])

Coxproportionalhazardsmultipleregressionmodelsfor

all-cause

mortalitywithHbA1cSD

tertiles:

Tertile

1:HR1

Tertile

2:HR1.45

(0.730

–2.88)

Tertile

3:HR3.09

(1.45–

6.58)

Coxproportionalhazardsmultipleregressionmodelsfor

all-cause

mortality,HbA1cCVtertiles:

Tertile

1:HR1

Tertile

2:HR1.21

(0.616

–2.38)

Tertile

3:HR2.89

(1.45–

5.74)

AER

,albumin

excretionrate;CKD

,chronickidney

disease;CVD,cardiovasculardisease;eG

FR,estim

ated

glomerularfiltrationrate.



index HbA1c .8% (64 mmol/mol), noassociation between HbA1c variabilityand mortality could be identified.Sensitivity analyses for study type and

studies that adjusted for duration of di-abetes, number of HbA1c measurements,baseline medications, and comorbiditiesproduced similar results to those that in-cluded all studies (Supplementary Table2). There were too few studies in themeta-analysis to assess publication bias.

CONCLUSIONS

Glycemic variability is emerging as a mea-sure of glycemic control that may be animportant predictor of complications inpatients with diabetes. Our analysissuggests that greater HbA1c variability,irrespective of the definition used, isassociated with adverse outcomes in sev-eral micro- and macrovascular end pointsand mortality. We report that HbA1c vari-ability in type 1 and type 2 diabetes isassociated with renal and cardiovasculardisease. The former is supported by 10studies using both CV and SD as ameasureof HbA1c variability (6,10,12,16,24–29).Only one small cross-sectional study in apediatric cohort using CV did not reportthis significant association (35). The latteris supported by two studies using SD(10,12). Retinopathy appears to be associ-atedwith HbA1c variability in type 1 diabe-tes (6) but not in type 2 diabetes (16,20).However, this was shown with SD as themeasure of variability (6) and not with CV(11,14). Four studies addressed the rela-tionship with mortality in type 2 diabetes(13,15,18,19), with significant associationsreported for SD and CV (13,15,18). Posthoc analysis of the ADVANCE data setshowed HbA1c variability defined by CVand SD to be associated with macrovascu-lar events and combinedmacro/microvas-cular events but not with microvascularevents in type 2 diabetes (18). These find-ings were independent of mean HbA1c,suggesting that HbA1c variability maybe a useful additional risk stratificationtool in both type 1 and type 2 diabetes.The present results add to the find-

ings of a significant association betweenHbA1c SD and renal disease reported inthe 2014 systematic review and meta-analysis by Cheng et al. (9). This meta-analysis of eight articles assessing therelationship between HbA1c variabilityand renal disease in type 1 and type 2diabetes has several limitations. Studieswere excluded that did not report HR

[including the study by Penno et al.(16)], measures of variability otherthan SD or CV were not considered,and different renal outcomes/end pointswere pooled.

The present results also differ fromthe previous systematic reviews ofshort-term glycemic variability and therisk of complications in diabetes (21,22).In previous studies, short-term glycemicvariability was assessed by a variety ofmethods, such as SD, CV, and meanamplitude of glycemic excursions ofdaily glucose readings, including self-monitoring of blood glucose, continu-ous blood glucose monitoring, fastingplasma glucose, and postprandial glu-cose (21). These studies found noconsistent evidence of a relationshipbetween short-term glycemic variabilityand the risk of any complications intype 1 diabetes. However, in six studiesinvolving patients with type 2 diabetes,both previous reviews found a positiveassociation between glucose variabilityand retinopathy. In general agreementwith these two reviews, we found a pos-itive relationship between glycemic var-iability and cardiovascular disease intype 2 diabetes. The present findingof a significant association betweenHbA1c variability and all-cause mortalityin type 2 diabetes is consistent with thefindings of Nalysnyk et al. (22) but notthose of Smith-Palmer et al. (21).

These differing risk prediction resultsfor short- and long-term glycemic vari-ability may indicate differing pathologi-cal mechanisms. Short-term glycemicvariability has been postulated to induceoxidative stress, inflammatory cytokines,and endothelial damage (37–41), mech-anisms linked to diabetes complications(42,43). Additional mechanisms thatmay explain the association of HbA1c var-iability and adverse events include cellu-lar metabolic memory (44–47), insulinresistance (10,48), sensitivity of HbA1cfor detecting glycemic variability (44),and the exponential relationship be-tween HbA1c and risk of microvascularcomplications (16,44).

Confounding factors rather than acausal relationship may explain the asso-ciation of HbA1c variability with compli-cations. These include poor medicationcompliance and self-management(10,12,28); multimorbidity (28); certainmedications, such as steroids and anti-psychotics (49); poor quality of life and

lack of support (50,51); and infections(10).

Eight studies indicated that HbA1c

variability was superior at predictingdiabetes-related complications thanmean HbA1c (6,10,12,13,15,17,24,25).Only one study found a significant asso-ciation of mean HbA1c with diabetes-related complications but not with HbA1cvariability (16,17). Further research is re-quired to assess whether HbA1c variabil-ity might be clinically useful for riskstratification and whether it might be avaluable therapeutic target.

To our knowledge, this systematic re-view and meta-analysis is the first ofHbA1c variability in diabetes and risk ofmortality and complications other thanrenal disease. Limitations of the analysisare exclusion of non-English-languagearticles and studies before 2004. How-ever, inclusion of studies earlier thanthe past 10 years may not be generaliz-able to current practices because cur-rent therapies (long-acting insulins,GLP-1 agonists, and dipeptidyl peptidase-4 inhibitors) were not available before2004. Because of the small number ofavailable studies, wewere unable to usemeta-regression to assess study charac-teristics as moderators. The heteroge-neity estimates vary from very highto zero, and arguably, highly hetero-geneous studies should not be meta-analyzed in the first place. However,homogeneity has been shown to berare and often falsely assumed, espe-cially for small meta-analyses, some-times leading to false conclusions (52).From a statistical point of view, it is bet-ter to identify heterogeneity (which islikely present anyway), which can thenbe successfully accounted for in arandom-effects meta-analysis model(53). Some limitations are inherent tothe available literature, including the ob-servational nature of studies, retrospec-tive design of some, unclear or shortfollow-up periods, exclusion of patientsdeemed as having too few HbA1c mea-surements (13,14,16,17,23,28), and thenonadjustment for different numbers ofHbA1c measurements, duration of diabe-tes, comorbidities, and baseline medica-tions. In addition, there is no acceptedmethod of assessing HbA1c variability,and a single definition of outcomes wasnot used.

The present findings support theneed for further studies investigating





the relationship between HbA1c variabil-ity and diabetes complications. More-sophisticatedmeasuresofHbA1c variabilityare needed as well as consensus abouthow such variability should be defined, in-cluding adjustment for differing intervalsbetween HbA1c measurements and ad-dressing the temporality of the varianceproblem (54). The present findings suggestthat HbA1c variability may be a useful riskstratification tool in both type 1 and type 2diabetes.In conclusion, this meta-analysis

shows significant associations betweenHbA1c variability and all-causemortality,renal disease, and cardiovascular dis-ease in type 2 diabetes and retinopathy,renal disease, and cardiovascular dis-ease in type 1 diabetes. These relation-ships are independent of mean HbA1c,and in the majority of studies, variabilitywas more predictive of adverse out-comes than mean HbA1c.

Funding.C.G. is fundedbyaNational Institute forHealth Research Academic Clinical Fellowship.Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.Author Contributions. C.G. searched data-bases, selected studies, extracted data, andwrote the manuscript. C.S.K. searched data-bases, selected studies, extracted and analyzeddata, and contributed towriting themanuscript.S.A. selected studies and extracted data. I.B.contributed to the discussion. E.K. and M.A.M.selected studies, contributed to the discussion,and reviewed and edited themanuscript. P.K.M.and G.H. reviewed and edited the manuscript.Y.L. extracted and analyzed data, contributedto the discussion, and reviewed and edited themanuscript.M.K.R. contributed to thedesignandreviewed and edited the manuscript.

References1. UK Prospective Diabetes Study (UKPDDS)Group. Intensive blood-glucose control with sul-phonylureas or insulin compared with conven-tional treatment and risk of complications inpatients with type 2 diabetes (UKPDS 33). Lan-cet 1998;352:837–8532. Holman RR, Paul SK, Bethel MA, MatthewsDR, Neil HA. 10-year follow-up of intensive glu-cose control in type 2 diabetes. N Engl J Med2008;359:1577–15893. The Diabetes Control and Complications TrialResearch Group. The effect of intensive treat-ment of diabetes on the development andprogression of long-term complications ininsulin-dependent diabetes mellitus. N Engl JMed 1993;329:977–9864. Nathan DM, Cleary PA, Backlund JY, et al.;Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions andComplications (DCCT/EDIC) Study ResearchGroup. Intensive diabetes treatment and car-diovascular disease in patients with type 1 di-abetes. N Engl J Med 2005;353:2643–2653

5. The Diabetes Control and Complications TrialResearch Group. The relationship of glycemicexposure (HbA1c) to the risk of developmentand progression of retinopathy in the DiabetesControl and Complications Trial. Diabetes 1995;44:968–9836. Kilpatrick ES, Rigby AS, Atkin SL. A1C variabil-ity and the risk of microvascular complicationsin type 1 diabetes: data from the Diabetes Con-trol and Complications Trial. Diabetes Care2008;31:2198–22027. Brownlee M, Hirsch IB. Glycemic variability:a hemoglobin A1c-independent risk factor fordiabetic complications. JAMA 2006;295:1707–17088. Cavalot F. Do data in the literature indicatethat glycaemic variability is a clinical problem?Glycaemic variability and vascular complica-tions of diabetes. Diabetes Obes Metab 2013;15(Suppl. 2):3–89. Cheng D, Fei Y, Liu Y, et al. HbA1C variabilityand the risk of renal status progression in dia-betes mellitus: a meta-analysis. PLoS One 2014;9:e11550910. Waden J, Forsblom C, Thorn LM, Gordin D,Saraheimo M, Groop PH; Finnish Diabetic Ne-phropathy Study Group. A1C variability predictsincident cardiovascular events, microalbuminu-ria, and overt diabetic nephropathy in patientswith type 1 diabetes. Diabetes 2009;58:2649–265511. HietalaK,Waden J, ForsblomC,et al.; FinnDianeStudy Group. HbA1c variability is associatedwith an increased risk of retinopathy requiringlaser treatment in type 1 diabetes. Diabetolo-gia 2013;56:737–74512. Luk AO, Ma RC, Lau ES, et al. Risk associa-tion of HbA1c variability with chronic kidneydisease and cardiovascular disease in type 2 di-abetes: prospective analysis of the Hong KongDiabetes Registry. Diabetes Metab Res Rev2013;29:384–39013. Takao T, Matsuyama Y, Yanagisawa H,Kikuchi M, Kawazu S. Association betweenHbA1c variability and mortality in patientswith type 2 diabetes. J Diabetes Complications2014;28:494–49914. Hermann JM, Hammes HP, Rami-Merhar B,et al.; DPV Initiative the German BMBF Compe-tence Network Diabetes Mellitus. HbA1c vari-ability as an independent risk factor fordiabetic retinopathy in type 1 diabetes: a Ger-man/Austrian multicenter analysis on 35,891patients. PLoS One 2014;9:e9113715. Ma WY, Li HY, Pei D, et al. Variability inhemoglobin A1c predicts all-cause mortality inpatients with type 2 diabetes. J Diabetes Com-plications 2012;26:296–30016. Penno G, Solini A, Bonora E, et al.; RenalInsufficiency and Cardiovascular Events StudyGroup. HbA1c variability as an independent cor-relate of nephropathy, but not retinopathy, inpatients with type 2 diabetes: the Renal Insuffi-ciency And Cardiovascular Events (RIACE) ItalianMulticenter Study. Diabetes Care 2013;36:2301–231017. Penno G, Solini A, Zoppini G, et al.; RenalInsufficiency And Cardiovascular Events (RIACE)Study Group. Hemoglobin A1c variability as anindependent correlate of cardiovascular dis-ease in patients with type 2 diabetes: a cross-sectional analysis of the Renal Insufficiency And

Cardiovascular Events (RIACE) Italian MulticenterStudy. Cardiovasc Diabetol 2013;12:9818. Hirakawa Y, Arima H, Zoungas S, et al. Im-pact of visit-to-visit glycemic variability on therisks of macrovascular andmicrovascular eventsand all-cause mortality in type 2 diabetes: theADVANCE trial. Diabetes Care 2014;37:2359–236519. Skriver MV, Sandbæk A, Kristensen JK,Støvring H. Relationship of HbA1c variability,absolute changes in HbA1c, and all-cause mor-tality in type 2 diabetes: a Danish population-based prospective observational study. BMJOpen Diabetes Res Care 2015;3:e00006020. Foo V, Tan G, Sabanayagam C, et al. HbA1cvariability and diabetic retinopathy in Asiantype 2 diabetes. Invest Ophthalmol Vis Sci2014;55:441221. Smith-Palmer J, Brandle M, Trevisan R,Orsini Federici M, Liabat S, Valentine W. Assess-ment of the association between glycemic var-iability and diabetes-related complications intype 1 and type 2 diabetes. Diabetes Res ClinPract 2014;105:273–28422. Nalysnyk L, Hernandez-Medina M,Krishnarajah G. Glycaemic variability and com-plications in patients with diabetes mellitus:evidence from a systematic review of the lit-erature. Diabetes Obes Metab 2010;12:288–29823. Cummings DM, Larsen LC, Doherty L, LeaCS, Holbert D. Glycemic control patterns andkidney disease progression among primarycare patients with diabetes mellitus. J Am BoardFam Med 2011;24:391–39824. Hsu CC, Chang HY, Huang MC, et al. HbA1cvariability is associated with microalbuminuriadevelopment in type 2 diabetes: a 7-year pro-spective cohort study. Diabetologia 2012;55:3163–317225. Raman S, Delurgio sr SA, Lind M, KosiborodM, Fridlington AG, et al. A1C variability predictsthe risk of microalbuminuria among childrenwith type 1 diabetes mellitus (T1DM). Pre-sented at the 71st Scientific Sessions of theAmerican Diabetes Association, 24-28 June2011, San Diego, CA26. Rodrıguez-Segade S, Rodrıguez J, GarcıaLopez JM, Casanueva FF, Cami~na F. Intraper-sonal HbA(1c) variability and the risk of pro-gression of nephropathy in patients withtype 2 diabetes. Diabet Med 2012;29:1562–156627. Sugawara A, Kawai K, Motohashi S, et al.HbA(1c) variability and the development of mi-croalbuminuria in type 2 diabetes: TsukubaKawai Diabetes Registry 2. Diabetologia 2012;55:2128–213128. Lin CC, Chen CC, Chen FN, et al. Risks ofdiabetic nephropathy with variation in hemo-globin A1c and fasting plasma glucose. Am JMed 2013;126:1017.e1–1017.e1029. Marcovecchio ML, Dalton RN, Chiarelli F,Dunger DB. A1C variability as an independentrisk factor for microalbuminuria in young peo-plewith type 1 diabetes. Diabetes Care 2011;34:1011–101330. Department of Health. National ServiceFramework for Diabetes. London, Departmentof Health, 200131. Health and Social Care Information Centre.Quality and outcomes framework, 2014. Available


from http://www.hscic.gov.uk/qof. Accessed 27May 201532. Davies HT, Crombie IK, Tavakoli M. Whencan odds ratiosmislead? BMJ 1998;316:989–99133. Loke YK, Price D, Herxheimer A. Adverse ef-fects. In Cochrane Handbook for Systematic Re-views of Interventions. Higgins JPT, Green S, Eds.Chichester, U.K., Wiley, 200834. Ioannidis JP, Trikalinos TA. The appropriate-ness of asymmetry tests for publication bias inmeta-analyses: a large survey. CMAJ 2007;176:1091–109635. Nazim J, Fendler W, Starzyk J. Metaboliccontrol and its variability are major risk factorsfor microalbuminuria in children with type 1 di-abetes. Endokrynol Pol 2014;65:83–8936. Lang C, Levin D, Mohan M, et al. Effect ofglycaemic control on outcome in patients withtype 2 diabetes mellitus and chronic heart fail-ure. J Am Coll Cardiol 2015;65(Suppl. 1):A88637. Monnier L, Mas E, Ginet C, et al. Activationof oxidative stress by acute glucose fluctuationscompared with sustained chronic hyperglyce-mia in patients with type 2 diabetes. JAMA2006;295:1681–168738. Ceriello A, Esposito K, Piconi L, et al. Oscil-lating glucose is more deleterious to endothelialfunction and oxidative stress thanmean glucosein normal and type 2 diabetic patients. Diabetes2008;57:1349–135439. Chang C-M, Hsieh C-J, Huang J-C, Huang I-C.Acute and chronic fluctuations in blood glucoselevels can increase oxidative stress in type 2

diabetes mellitus. Acta Diabetol 2012;49(Suppl.1):S171–S17740. Quagliaro L, Piconi L, Assaloni R, et al. Inter-mittent high glucose enhances ICAM-1, VCAM-1and E-selectin expression in human umbilical veinendothelial cells in culture: the distinct role ofprotein kinase C and mitochondrial superoxideproduction. Atherosclerosis 2005;183:259–26741. Ceriello A, IhnatMA. ‘Glycaemic variability’: anew therapeutic challenge in diabetes and thecritical care setting. DiabetMed 2010;27:862–86742. Dhalla NS, Temsah RM, Netticadan T. Roleof oxidative stress in cardiovascular diseases.J Hypertens 2000;18:655–67343. Nishikawa T, Edelstein D, Du XL, et al. Nor-malizing mitochondrial superoxide productionblocks three pathways of hyperglycaemic dam-age. Nature 2000;404:787–79044. Kilpatrick ES. The rise and fall of HbA(1c)as a risk marker for diabetes complications. Di-abetologia 2012;55:2089–209145. Schisano B, Tripathi G, McGee K, McTernanPG, Ceriello A. Glucose oscillations, more thanconstant high glucose, induce p53 activationand a metabolic memory in human endothelialcells. Diabetologia 2011;54:1219–122646. Keating ST, El-Osta A. Glycemic memoriesand the epigenetic component of diabetic ne-phropathy. Curr Diab Rep 2013;13:574–58147. The Diabetes Control and Complications Tri-al/Epidemiology of Diabetes Interventions andComplications Research Group. Retinopathyand nephropathy in patients with type 1

diabetes four years after a trial of intensivetherapy. N Engl J Med 2000;342:381–38948. Groop PH, Forsblom C, Thomas MC. Mech-anisms of disease: pathway-selective insulin re-sistance and microvascular complications ofdiabetes. Nat Clin Pract Endocrinol Metab2005;1:100–11049. Newcomer JW. Antipsychotic medications:metabolic and cardiovascular risk. J Clin Psychi-atry 2007;68(Suppl. 4):8–1350. Maciejewski ML, Dowd B, Call KT, FeldmanR. Comparing mortality and time until death forMedicare HMO and FFS beneficiaries. HealthServ Res 2001;35:1245–126551. Hamer M, Stamatakis E, Kivimaki M, PascalKengne A, Batty GD. Psychological distress, gly-cated hemoglobin, and mortality in adults withandwithout diabetes. PsychosomMed 2010;72:882–88652. Kontopantelis E, Springate DA, Reeves D. Are-analysis of the Cochrane Library data: thedangers of unobserved heterogeneity in meta-analyses. PLoS One 2013;8:e6993053. Kontopantelis E, Reeves D. Performance ofstatistical methods for meta-analysis when truestudy effects are non-normally distributed: asimulation study. Stat Methods Med Res 2012;21:409–42654. Sperrin M, Thew S, Weatherall J, Dixon W,Buchan I. Quantifying the longitudinal value ofhealthcare record collections for pharmacoepi-demiology. AMIA Annu Symp Proc 2011;2011:1318–1325


http://www.hscic.gov.uk/qof


long-term glycemic variability and risk of adverse ... · group or whether variability was...

Documents