Articles

Type 2 diabetes mellitus after gestational diabetes:

a systematic review and meta-analysis
Leanne Bellamy, Juan-Pablo Casas, Aroon D Hingorani, David Williams

Summary
Background Women with gestational diabetes are at increased risk of developing type 2 diabetes, but the risk and time Lancet 2009; 373: 1773–79
of onset have not been fully quantified. We therefore did a comprehensive systematic review and meta-analysis to See Editorial page 1735
assess the strength of association between these conditions and the effect of factors that might modify the risk. See Comment page 1738
Northwick Park Hospital,
Methods We identified cohort studies in which women who had developed type 2 diabetes after gestational diabetes London, UK (L Bellamy MBBS);
were followed up between Jan 1, 1960, and Jan 31, 2009, from Embase and Medline. 205 relevant reports were hand Department of Epidemiology
and Population Health,
searched. We selected 20 studies that included 675 455 women and 10 859 type 2 diabetic events. We calculated and London School of Hygiene and
pooled unadjusted relative risks (RRs) with 95% CIs for each study using a random-effects model. Subgroups analysed Tropical Medicine, London, UK
were the number of cases of type 2 diabetes, ethnic origin, duration of follow-up, maternal age, body-mass index, and (J-P Casas MD); Department of
diagnostic criteria. Epidemiology and Public
Health, University College
London, London, UK
Findings Women with gestational diabetes had an increased risk of developing type 2 diabetes compared with those (Prof A D Hingorani FRCP); and
who had a normoglycaemic pregnancy (RR 7·43, 95% CI 4·79–11·51). Although the largest study (659 164 women; Institute for Women’s Health,
University College London,
9502 cases of type 2 diabetes) had the largest RR (12·6, 95% CI 12·15–13·19), RRs were generally consistent among
London, UK (D Williams FRCP)
the subgroups assessed.
Correspondence to:
Dr David Williams, Institute for
Interpretation Increased awareness of the magnitude and timing of the risk of type 2 diabetes after gestational diabetes Women’s Health, University
among patients and clinicians could provide an opportunity to test and use dietary, lifestyle, and pharmacological College London, 250 Euston
interventions that might prevent or delay the onset of type 2 diabetes in affected women. Road, London NW1 2PG, UK
d.j.williams@ucl.ac.uk

Funding None.

Introduction elucidation of the causes of these disorders, and for the

Gestational diabetes mellitus is defined as glucose
intolerance that is first detected during pregnancy.1 In the
USA, this condition affects 135 000 pregnancies (3–5% of
all pregnancies) per year.2,3 Shortly after delivery, glucose
homoeostasis is restored to non-pregnancy levels, but
affected women remain at high risk of developing type 2
diabetes mellitus in the future.4,5
For any population and ethnic group, the risk of
gestational diabetes indicates the underlying frequency of
type 2 diabetes.2,6 The incidences of gestational diabetes
and type 2 diabetes are rising throughout the world, with
huge health-care and economic costs.6,7 Diabetes pre-
disposes individuals to cardiovascular, renal, and retinal
diseases, costing US$91·8 billion per year in the USA.8
Women who have had gestational diabetes are advised
to have their glucose tolerance assessed 6 weeks after
delivery.9 However, low rates of attendance at the 6-week
follow-up10,11 suggest that health-care professionals,
women with gestational diabetes, or both, do not realise
the importance of this disorder as an early warning sign
of the susceptibility to develop type 2 diabetes in the
future; therefore an opportunity to promote health and
prevent disease is missed. Moreover, no consensus exists
on how and whether mothers should continue to be
monitored after this period.
The association between gestational diabetes and
type 2 diabetes mellitus has implications for the
prediction and possible prevention or delay of the
development of type 2 diabetes in women. We therefore
did a comprehensive systematic review and meta-
analysis to quantify the overall risk of women with
gestational diabetes mellitus developing type 2 diabetes
mellitus, and to assess the effect of factors that might
modify this risk.
Methods
Search strategy and selection criteria
We did an electronic search of Embase from 1974 to
Jan 31, 2009, and Medline from 1960 to Jan 31, 2009,
without language restrictions. Search term combinations
were “gestational diabetes”, “diabetic pregnancy”,
“diabetes mellitus”, “type 2 diabetes mellitus”, “NIDDM”,
and “non-insulin dependent diabetes mellitus”. All
reference lists from the main reports and relevant reviews
were hand searched for additional eligible studies.
We identified retrospective and prospective cohort
studies, reported between Jan 1, 1960, and Jan 31, 2009, in
which pregnant women of any parity or ethnic origin were
identified as having gestational diabetes (exposed group)
and normoglycaemic pregnancies (control group).
The outcome studied was the development of type 2
diabetes at least 6 weeks after the end of the index
pregnancy, and was defined with an oral glucose tolerance
test or fasting plasma glucose concentration, or both,

www.thelancet.com Vol 373 May 23, 2009 1773

 Articles

measured to assess heterogeneity with RevMan and

435 potentially relevant titles and Phase 1
summaries identified
230 excluded:
No desirable outcomes
Outcomes during pregnancy only
Editorials
Letters
News articles
205 (47%) full-text studies reviewed Phase 2
137 excluded (webappendix pp 1–9)
30 reviews
2 case-control studies
4 cross-sectional studies
1 follow-up <6 weeks
18 outcome inferred only
12 no association with study question
23 insulin resistance after gestational
diabetes
15 only gestational diabetes assessed
10 only type 2 diabetes assessed
1 foreign, not available in
British Library
20 repeat study group
1 normoglycaemic group lost to
follow-up
STATA (version 10.0). We investigated potential sources
of identified heterogeneity among studies by stratification
according to the number of cases of type 2 diabetes
(<100, 100–500, and >500); ethnic origin (white,
non-white, and mixed race populations); average
follow-up (<5 years and ≥5 years); average maternal age
at index pregnancy (<30 years and ≥30 years); average
body-mass index of mothers at the end of their index
pregnancy and follow-up (<25 kg/m², 25–30 kg/m²,
>30 kg/m²); study design (prospective and retrospec-
tive); and variation in criteria used in each study to
diagnose gestational diabetes and type 2 diabetes. In
these analyses, we used 99% CIs to reduce the potential
for chance differences arising from multiple testing.
Small-study bias and publication bias were assessed with
funnel-plot analysis. RRs were unadjusted because they
were calculated from raw data.
Role of the funding source
There was no funding source for this study. The
corresponding author, on behalf of all authors, had full
access to all the data in the study and had final
responsibility for the decision to submit for publication.

68 (16%) relevant studies—ie,

cohort studies in women with Phase 3
gestational and type 2 diabetes mellitus
48 excluded (webappendix pp 10–13)
because no control group
20 (5%) included in meta-analysis Phase 4
Figure 1: Study selection process
Modified with permission from Centre for Reviews and Dissemination.20
according to the National Diabetes Data Group,12 WHO
criteria,13 countrywide guidelines,14 or specified local
criteria.15,16 Cohort studies of women with diabetes
mellitus before the index pregnancy were excluded.
Data gathering and statistical analysis
Two independent reviewers (LB and DW) checked each
full-text report for eligibility, and extracted and tabulated
all relevant data. Disagreement was settled by consensus
between all authors. If there was more than one report
relating to the same cohort, the report with the information
See Online for webappendix most relevant to our analysis was included (webappendix
for studies excluded at full-text review). Additional tabular
data, including the largest study in the meta-analysis, were
provided by direct contact with the corresponding
authors.17,18 All procedures conformed to the guidelines for
meta-analysis of observational studies in epidemiology.19
We used RevMan (version 5) to calculate unadjusted
summary relative risks (RRs) with 95% CIs, using a
random-effects model for all analyses. τ² and I² were also
Results
Figure 1 shows the study selection process (reasons for
exclusion are listed in webappendix pp 1–13). 48 of
68 full-text reports meeting all the inclusion criteria were
subsequently excluded because of the absence of an
appropriate control or reference population. The
20 remaining studies (table),5,15–18,21–35 contributed 675 455
women with type 2 diabetes to the meta-analysis, and
31 867 of these had previous pregnancies affected by
gestational diabetes with a total of 10 859 incident cases
of type 2 diabetes.
Figure 2 shows the overall RR of women developing
type 2 diabetes after a pregnancy complicated by
gestational diabetes with evidence of heterogeneity in
the risk estimate. Generally, small studies had large
effect estimates (figure 3), but the largest study18 included
in the meta-analysis had the largest effect estimate
(figure 2).
Figure 4 shows the potential sources of heterogeneity
by study characteristics, participant characteristics, and
diagnostic criteria for gestational diabetes and type 2
diabetes. Effect estimates were broadly consistent in all
the subgroups analysed (figure 4B; figure 4C). There was
some heterogeneity in effect estimates when studies
were grouped according to the number of cases of type 2
diabetes (<100, 100–500, and >500; figure 4A).
Heterogeneity was reduced by exclusion of the largest
study (χ²=1·48, df=1, p=0·22, I²=32·2%). No significant
heterogeneity was evident from the subgroup analysis of
the criteria used for the diagnosis of gestational diabetes
and type 2 diabetes in the studies included in the analysis
(figure 4C).

1774 www.thelancet.com Vol 373 May 23, 2009

 Articles

Study type, year, country Ethnic Mean maternal age GDM criteria Total women Mean follow-up Definition of T2DM
origin (years; SD or 95% CI) studied (degree of (SD or 95% CI)
of women with matching GDM/
GDM/non-GDM non–GDM)
Feig et al18 Retrospective cohort, 2008, Mixed 29·3 (5·5) Canadian Institute for 659 164 5·2 yearsi Ontario Diabetes
Canada Health Information Databaseii,37
(discharge summary)36
Lee H et al15 Prospective cohortiii, 2008, Non-white 33·6 (4·8) National Diabetes Data 1736v (248 IGT) 2·1 yearsi Localvi
Korea Group, 1979iv,38
Madarász Retrospective cohortiii, 2008, White 33·1/30·0 (5·9) WHO, 1999vii,39 107 3·6 years (GDM; 0·8)/ WHO, 1999viii,40
et al21 Hungary 8·1 years (non-GDM; 5·1)
Gunderson Prospective cohort, 2007, USA Mixed Matched range 18–30 Obstetric laboratory 2408 Total 20 year follow-up American Diabetes
et al22 reports (72% followed for entire Association, 1997ix/diabetes
time) medication/self report
Vambergue Prospective cohort, 2007, Mixed 27·0 (5·2)/28·8 (5·8) Carpenter and 581xi,xii (175 AGT) 6·75 years (0·8) American Diabetes
et al23 France Coustanx,41 Association, 1997ix
Lee A et al24 Retrospective cohort, 2007, Mixed 30·7 (5·1)/30·5 (4·6) Australian Diabetesxiii,42 6253xiv 2·2 years (GDM),i WHO, 1998viii,40
Australia (pregnancy guidelines) 8·6 years (non-GDM)i
Ferraz et al17 Prospective cohort, 2007, Brazil Non-white 26·9/25·1 WHO, 1999xv,43 178xvi 6·2 years (0·8) WHO, 1999viii,40
Krishnaveni Prospective cohort, 2007, India Non-white Matched age range Carpenter and 524 5 years WHO, 1999viii,40
et al25 Coustanx,41
Morimitsu Prospective cohort, 2007, Brazil Mixed 32/27 (7) American Diabetes 34xviii 16–24 weeks American Diabetes
et al26 Association, 1997xvii,14 Association, 1997ix
Järvelä et al5 Retrospective cohortiii, 2006, White 31·6 (17·7–46·5)/31·3 Finnish Diabetes 870v,xii,xiv 5·7 years (GDM; 1·0–11·6) Medication for T2DM
Finland (18·8–46·0) Associationxix,44 6·1 (non-GDM; 1·5–13·1) linked to databasexx,45
Albareda Prospective cohort, 2003, White 30·7/30·4 Second and third GDM 766xiv 6·16 years (0·05–13·73) WHO, 1998viii,40
et al27 Spain workshop
conferencexxi,46,47
Åberg et al28 Retrospective cohort, 2002, White Matched range 20–45 European Association 290 1 year WHO, 1985xxiii
Sweden for Study of
Diabetesxxii,48
Linné et al16 Retrospective cohort iii, 2002, White 32·6/30·6 Localxxiv 80v,xvi,xxv 15 years Localxxvi
Sweden
Bian et al29 Retrospective cohort, 2000, Non-white 29/29 (23–40) National Diabetes Data 84v,xiv,xviii,xxvii 5–11 years WHO, 1985xxiii
China Group, 1979iv
Ko et al30 Prospective cohort, 1999, Non-white 34·0 (4·1)/34·4 (6·4) Localxxviii 1232v 6 weeks WHO, 1985xxiii
China
Osei et al31 Retrospective cohort, 1998, Non-white 31·3 (2·0)/36·0 (0·9) National Diabetes Data 65xxix 7 years National Diabetes Data
USA Group, 1979iv Group, 1979xxx,49
Damm et al32 Retrospective cohort, 1994, White 30·1/26·7 Localxxxi 298xiv 7·5 yearsi WHO, 1985xxiii
Denmark
Benjamin Retrospective cohort, 1993, Mixed 27·2/26·5 Localxxxii,50 94v,xii,xvi,xxxiii,51 4·8 years (GDM)/ National Diabetes Data
et al33 New Mexico 5·5 years (non-GDM) Group, 1979xxx,49
O’Sullivan34 Prospective cohort, 1991, USA Mixed ·· Localxxxii 943 22–28 years WHO, 1985xxiii
Persson Retrospective cohort, 1991, White 31 (20–46)/30 (16–43) WHO, 1985 xxxiv,51
186 v,xii,xix,xxxv
3–4 years WHO, 1985xxiii
et al35 Sweden

AGT=abnormal glucose tolerance. FPG=fasting plasma glucose. IGT=impaired glucose tolerance. i=median values at follow-up only. ii=includes all individuals diagnosed with diabetes ≥90 days after delivery
living within Ontario, Canada. iii=report is cited as a case-control study in the summary but as a cohort study in the methods. iv=two or more FPG concentrations ≥5·8 mmol/L or 1 h glucose ≥10·6 mmol/L, 2 h
glucose ≥9·2 mmol/L, 3 h glucose ≥8·1 mmol/L after 100 g glucose load. v=matched by age. vi=FPG concentrations ≥7 mmol/L only. vii=FPG concentrations ≥7 mmol/L or 2 h glucose ≥11·1 mmol/L, or both, after
75 g oral glucose load. viii=FPG concentrations ≥7 mmol/L, or 2 h glucose ≥11·1 mmol/L after 75 g glucose load. ix=FPG concentrations ≥7 mmol/L or 2 h glucose ≥11·1 mmol/L, or both, after 75 g glucose load.
x=FPG concentrations ≥5·3 mmol/L or 2 h plasma glucose ≥8·6 mmol/L, or 3 h glucose ≥7·7 mmol/L after 100 g oral glucose load. xi=matched by ethnic origin. xii=matched by parity. xiii=FPG concentrations
≥5·5 mmol/L or 2 h glucose ≥8·0 mmol/L after 75 g glucose load (before Jan 01, 1999, FPG ≥7·8 mmol/L, 1 h glucose ≥9 mmol/L and 2 h glucose ≥7·0 mmol/L after 50 g glucose load). xiv=matched by year of
delivery. xv=FPG concentrations ≥7 mmol/L or 2 h glucose ≥11·1 mmol/L, or both, (group also included gestational impaired glucose tolerance=FPG <7 mmol/L and 2 h glucose 7·8–11·1 mmol/L) after 75 g oral
glucose load. xvi=matched by body-mass index. xvii=two or more FPG concentrations ≥5·3 mmol/L, 1 h glucose ≥10·0 mmol/L, 2 h glucose ≥8·6 mmol/L, 3 h glucose ≥7·8 mmol/L after 100 g glucose load.
xviii=matched by gestational age at study entry. xix=FPG concentrations ≥4·8 mmol/L or 1 h glucose ≥10·0 mmol/L, or 2 h glucose ≥8·7 mmol/L after 75 g oral glucose load. xx=medications obtained from
central drug register with complete population coverage plus questionnaire. xxi=two or more FPG concentrations ≥5·8 mmol/L, 1 h glucose ≥10·6 mmol/L, 2 h glucose ≥9·2 mmol/L, 3 h glucose ≥8·1 mmol/L
after glucose (75 g) load. xxii=FPG concentrations ≥6 mmol/L or 2 h glucose ≥9 mmol/L after 75 g glucose load. xxiii=FPG concentrations ≥7·8 mmol/L, or 2 h glucose ≥11·1 mmol/L after 75 g oral glucose load.
xxiv=2 h glucose concentration ≥9 mmol/L after 75 g glucose load. xxv=matched by waist to hip ratio. xxvi=2 h glucose concentration >10 mmol/L after 75 g glucose load. xxvii=matched by social background.
xxviii=two or more FPG concentrations ≥5·0 mmol/L or 1 h glucose ≥9·5 mmol/L, 2 h glucose ≥8·1 mmol/L, 3 h glucose ≥7·0 mmol/L after 50 g glucose load. xxix=study included a third group of 15 women who
were not included in this meta-analysis. xxx=FPG concentrations ≥7·8 mmol/L or 2 h glucose 11·1 mmol/L after 75 g glucose load. xxxi=two or more venous plasma glucose concentrations >3 SD from the mean
in women without gestational diabetes after 50 g glucose load (0 min: 6·4 mmol/L; 30 min: 10·1 mmol/L; 1 h: 10·1 mmol/L; 90 min: 8·7 mmol/L; 2 h: 7·6 mmol/L; 150 min: 7·6 mmol/L; 3 h: 6·6 mmol/L).
xxxii=fasting venous whole blood glucose concentration ≥5 mmol/L or 2 h venous whole blood glucose ≥8·0 mmol/L or 3 h glucose ≥6·9 mmol/L) after 100 g glucose load. xxxiii=matched by length of follow-up.
xxxiv=FPG concentrations ≥7 mmol/L and 2 h glucose ≥7·8 mmol/L after 75 g oral glucose load. xxxv=matched by prepregnancy weight and weight gain.

Table: Gestational diabetes mellitus (GDM) and development of type 2 diabetes mellitus (T2DM)

www.thelancet.com Vol 373 May 23, 2009 1775

 Articles

Country T2DM/GDM T2DM/no GDM Relative risk (95% CI)

Feig et al,18 1995–2002 Canada 2874/21 823 6628/637 341 12·66 (12·15–13·19)
Lee H et al,15 1995–97 Korea 71/620 22/868 4·52 (2·83–7·21)
Madarasz et al,21 1995 Hungary 21/68 0/39 24·93 (1·55–400·64)
Gunderson et al,22 1985–206 USA 43/166 150/2242 3·87 (2·87–5·22)
Vambergue et al,23 1992 France 53/295 1/111 19·94 (2·79–142·47)
Lee A et al,24 1971–2003 Australia 405/5470 16/783 3·62 (2·21–5·93)
Ferraz et al17* Brazil 6/70 7/108 1·32 (0·46–3·78)
Krishnaveni et al,25 1997–98 India 13/35 8/489 22·70 (10·09–51·08)
Morimitsu et al,26 1999–2001 Brazil 7/23 0/11 7·50 (0·47–120·11)
Järvelä et al,5 1984–94 Finland 23/435 0/435 47·00 (2·86–771·65)
Albareda et al,27 1966–93 Spain 44/696 0/70 9·07 (0·56–146·25)
Åberg et al,28 1991–99 Sweden 21/229 1/61 5·59 (0·77–40·66)
Linné et al,16 1964–65 Sweden 10/28 0/52 38·38 (2·33–631·74)
Bian et al,29 1964–65 China 15/45 1/39 13·00 (1·80–93·93)
Ko et al,30 1988–95 China 105/801 7/431 8·07 (3·79–17·19)
Osei et al,31 1990–91 USA 10/15 0/35 47·25 (2·95–757·28)
Damm et al,32 1978–85 Denmark 33/241 0/57 16·06 (1·00–258·06)
Benjamin et al,33 1961–88 New Mexico 14/47 3/47 4·67 (1·43–15·21)
O’Sullivan,34 1954–60 and 1962–70 USA 224/615 18/328 6·64 (4·19–10·52)
Persson et al,35 1961–84 Sweden 5/145 0/41 3·16 (0·18–55·76)

Total 3997/31 867 6862/643 588 7·43 (4·79–11·51)

Test for heterogeneity: τ 2 =0·50, χ 2=126·67, df=19 (p<0·0001), l2=85·0% (95%CI 78–90)
Test for overall effect: Z=9·39 (p<0·0001)

0·01 0·1 1 10 100

Decreased risk Increased risk

Figure 2: Risk of type 2 diabetes mellitus (T2DM) after gestational diabetes mellitus (GDM)
x-axis is log scale. Each solid square represents a relative risk. Horizontal lines indicate 95% CIs. df=degrees of freedom. *Dates not available.

association between gestational diabetes and type 2

0·0 diabetes, and the knowledge that many of the risk factors
are the same (ie, a family history of diabetes, raised
0·4
body-mass index, increased age, and Asian and black
ethnic origin), suggest that the two disorders might have
an overlapping cause.4 Results of candidate gene studies,
SE of log risk ratio

0·8 giving support to this hypothesis,52,53 show that frequency

1·2
1·6
0·01 0·1 1 10
Relative risk
Figure 3: Funnel plot of 20 cohort studies included in meta-analysis
x-axis is log scale. Dotted line is the summary relative risk (7·43).
In nine5,17,25,28–31,33,35 of 20 studies, none of the women
dropped out and data were not missing. In four
studies,18,22,23,32 633 449 (96%) of 659 164, 1696 (70%) of
2408, 713 (71%) of 1009, and 241 (81%) of 298 women were
followed up. In another study,24 2739 (44%) of 6253 were
followed up. In six studies,15,16,21,26,27,34 the dropout rate was
not recorded.
Discussion
Women who have had gestational diabetes have at least a
seven-fold increased risk of developing type 2 diabetes
mellitus in the future compared with those who have had
a normoglycaemic pregnancy. The strength of the
of some alleles associated with the increased risk of
development of type 2 diabetes were increased in women
who had gestational diabetes. Irrespective of the precise
biological link between these two disorders, the
development of gestational diabetes might help to identify
women at high risk of developing type 2 diabetes.
Although women who have had gestational diabetes
100 are recommended to have a glucose tolerance test at
6 weeks postpartum, most do not attend.11 The increased
risk of type 2 diabetes reported in this meta-analysis
might help to motivate mothers to attend screening
programmes, and health-care professionals to increase
uptake to these programmes or perhaps suggest the best
time for reassessment. Since the risk of type 2 diabetes
seems to be maintained for several years, consideration
of whether any form of continuous assessment would
lead to health gains is important. Women who have had
gestational diabetes also have increased lipid concentra-
tions and blood pressure, and type 2 diabetes is estimated
to confer an equivalent risk of ageing 15 years.54 Early
identification and treatment of these factors could also
help to reduce premature cardiovascular and renal
diseases in this group of individuals.10
We noted heterogeneity in the overall effect estimate and
did subgroup analyses to determine the potential sources.

1776 www.thelancet.com Vol 373 May 23, 2009

 Articles

Total number of Relative risk (99% CI) l2 (%; 99% CI) τ2 χ 2 test of
cases of T2DM heterogeneity
(p value)

A
Ethnic origin 0 (0–95) 0 0·78 on 2 df (0·68)
White (7 studies) 158 12·76 (2·31–70·63) 0 (0–80) 0 3·31 on 6 df (0·77)
Non-white (6 studies) 265 7·45 (2·43–22·83) 78 (38–92) 0·76 22·97 on 5 df (0·0003)
Mixed (7 studies) 10436 6·36 (2·06–19·61) 93 (87–97) 0·55 92·20 on 6 df (<0·0001)

Follow-up time 60 (N/A) 0·14 2·49 on 1 df (0·11)

<5 years (8 studies) 698 4·69 (2·84–7·75) 0 (0–77) 0 4·66 on 7 df (0·70)
≥5 years (12 studies) 10161 9·34 (3·42–25·54) 88 (77–93) 0·53 88·10 on 11 df (<0·0001)

Study design 0 (N/A) 0 0·62 on 1 df (0·43)

Prospective (9 studies) 779 6·07 (2·68–13·74) 71 (29–88) 0·28 27·64 on 8 df (0·0005)
Retrospective (11 studies) 10080 9·29 (3·02–28·60) 68 (27–86) 0·46 31·30 on 10 df (0·0005)

Number of incident cases 88 (54–97) 0·05 16·78 on 2 df (0·0002)

<100 (15 case studies) 389 8·71 (3·00–25·31) 52 (0–72) 0·58 29·37 on 14 df (0·009)
100–500 (4 studies) 968 4·88 (2·65–8·97) 55 (0–90) 0·07 6·74 on 3 df (0·08)
>500 (1 study) 9502 12·66 (11·79–13·60) N/A N/A N/A

B
Maternal age 0 (0–92) 0 0·25 on 3 df (0·97)
<30 years (5 studies) 9602 6·67 (1·08–41·16) 81 (14–94) 1·02 20·65 on 4 df (0·0004)
≥30 years (9 studies) 739 6·14 (2·66–14·15) 33 (0–76) 0·14 11·93 on 8 df (0·15)
Age range reported only (4 studies) 478 7·40 (2·03–26·94) 83 (14–95) 0·41 17·50 on 3 df (0·0006)
Not similar* (2 studies) 40 10·97 (0·37–325·68) 0 (N/A) 0 0·14 on 1 df (0·70)

BMI at index pregnancy 42 (0–88) 0·10 3·46 on 2 df (0·18)

<25 kg/m2 (4 studies) 205 8·92 (2·83–28·12) 0 (0–92) 0 0·38 on 3 df (0·94)
25–30 kg/m2 (1 study) 428 3·70 (1·59–8·58) 0 (N/A) 0 0·26 on 1 df (0·61)
Not recorded (15 studies) 10226 7·66 (3·12–18·79) 88 (79–93) 0·52 109·78 on 13 df (<0·0001)

BMI at follow-up 0 (0–92) 0 0·20 on 3 df (0·98)

<25 kg/m2 (5 studies) 424 6·66 (2·29–19·40) 78 (29–93) 0·32 17·98 on 4 df (0·001)
25–30 kg/m2 (3 studies) 105 8·94 0·58–138·54) 64 (0–90) 1·98 11·20 on 4 df (0·02)
≥30 kg/m2 (2 studies) 27 10·42 (0·25–433·97) 56 (N/A) 1·49 2·26 on 1 df (0·13)
Not recorded (10 studies) 10303 8·10 (2·82–23·25) 79 (49–91) 0·36 33·33 on 7 df (<0·0001)

C
GDM criteria 0 (0–80) 0 5·27 on 6 df (0·51)
WHO (3 studies) 39 3·28 (0·17–62·56) 48 (0–90) 1·16 3·87 on 2 df (0·14)
NDDG (4 studies) 163 6·77 (1·61–28·47) 20 (0–93) 0·20 3·75 on 3 df (0·29)
Countrywide guidelines (3 studies) 451 6·54 (0·61–69·89) 40 (0–87) 0·70 3·33 on 2 df (0·19)
Carpenter and Coustan (2 studies) 75 22·27 (6·10–81·25) 0 (N/A) 0 0·01 on 1 df (0·90)
EASD (1 study) 22 5·59 (0·18–172·63) N/A N/A N/A
Local (5 studies) 414 7·03 (3·73–13·25) 0 (0–87) 0 2·40 on 4 df (0·66)
Other† (2 studies) 9695 7·07 (0·95–52·45) 98 (N/A) 0·69 58·89 on 1 df (<0·00001)

T2DM criteria 22 (0–75) 0·05 5·14 on 4 df (0·27)

WHO (11 studies) 950 6·60 (2·58–16·89) 61 (7–83) 0·37 25·45 on 10 df (0·005)
NDDG (2 studies) 27 10·42 (0·25–433·97) 56 (N/A) 1·49 2·26 on 1 df (0·13)
ADA (3 studies) 254 5·46 (1·12–26·69) 29 (0–96) 0·28 2·80 on 2 df (0·25)
Local (2 studies) 103 8·28 (0·32–216·45) 54 (N/A) 1·24 2·18 on 1 df (0·14)
Other† (2 studies) 9525 12·66 (11·79–13·60) 0 (N/A) 0 0·84 on 1 df (0·36)

0·01 0·1 1 10 100

Decreased risk Increased risk

Figure 4: Risk of type 2 diabetes mellitus (T2DM) grouped by study characteristics (A), participant characteristics (B), and diagnostic criteria (C)
x-axis is log scale. Each solid square represents a relative risk. Horizontal lines indicate 99% CIs. ADA=American Diabetes Association. BMI=body-mass index. df=degrees of freedom. EASD=European
Association for Study of Diabetes. GDM=gestational diabetes mellitus. N/A=heterogeneity not applicable because one study analysed. NDDG=National Diabetes Data Group. *Average ages of women
with and without GDM were not similar; therefore effect of maternal age could not be assessed. †Includes databases and obstetric reports (table).

Effect estimates were similar when studies were grouped
according to patient characteristics, suggesting that much
of the heterogeneity was unexplained. The effect estimates
reported in studies in which different criteria were used
for the diagnosis of gestational diabetes mellitus and type 2
diabetes mellitus were similar. The number of cases of
type 2 diabetes included in our analysis contributed to the
heterogeneity, which was reduced by exclusion of the
largest study. However, this study was of high quality and
resulted in an effect size that was larger, instead of smaller,
than the estimates from small studies. Although we did
not identify the main sources of heterogeneity of effect
size, a meta-analysis of summary data from reported
studies has little capacity to do so. These sources could be

www.thelancet.com Vol 373 May 23, 2009 1777

 Articles
efficiently assessed in an individual-level meta-analysis
with existing datasets, or in a new large prospective study
that might help to delineate useful features for identi-
fication of a mother with gestational diabetes at high risk
of developing type 2 diabetes.
We were unable to stratify results according to family
history of diabetes and previous gestational diabetic
pregnancy because of the absence of studies that
adequately addressed this issue. With the exception of
the largest study18—the only study in which women were
not analysed according to their late-onset diagnosis of
diabetes (type 1 or 2)—we excluded studies or parts of
studies in which women who developed type 1 diabetes
in later life were identified. Therefore, inclusion of this
study might have resulted in an overestimation of the
number of women with gestational diabetes who
developed type 2 diabetes. However, because type 1
diabetes is rare compared with type 2 disease and
because of the large number of diabetes events within
the largest study,18 we do not think that the inclusion of
cases of type 1 diabetes distorted the large effect estimate
obtained within this study. Both types 1 and 2 diabetes
are more common after a pregnancy complicated by
gestational diabetes than after a normoglycaemic
pregnancy. This issue was addressed in most of the
studies by separation of the two conditions at follow-up.
Meta-analyses of reported data can also be affected by
publication bias, and the funnel plot provides some
evidence that studies with fewer than 100 cases resulted
in larger effect estimates than those with 100–500 cases.
However, in our meta-analysis, the largest effect estimate
was seen in the largest study with 9502 cases of type 2
diabetes.18
Studies were included from three decades in which the
widely accepted WHO and National Diabetes Data Group
criteria for type 2 diabetes were regularly revised. In 1997,
the concentration of fasting plasma glucose that was
used to define type 2 diabetes was lowered, thus reducing
the threshold for diagnosis. Therefore, the inclusion of
old studies might lead to an underestimation of the
future risk of type 2 diabetes in women who have had
gestational diabetes.
The increase in RR of future type 2 diabetes after
gestational diabetes in our systematic review and
meta-analysis provides an incentive for women with a
history of gestational diabetes to attend postpartum and
continuous assessments for glucose tolerance. A history of
gestational diabetes therefore provides a low-cost, natural
screening test for future type 2 diabetes. Resolution of the
metabolic changes after pregnancy could provide an
opportunity to test the effectiveness of interventions for
the primary prevention of type 2 diabetes with changes in
diet and lifestyle, and the introduction of treatments to
prevent or delay the onset of type 2 diabetes.55,56 The
effectiveness and cost-effectiveness of the interventions
would need to be tested in clinical trials. These interventions
might have an important effect on women’s health.
1778
Contributors
LB contributed to protocol design, data extraction, statistical analysis,
and data interpretation. J-PC contributed to statistical analysis and data
interpretation. ADH helped with data interpretation. DW contributed to
protocol design, and data extraction and interpretation. All authors
contributed to the writing and revision of this report, and have seen and
approved the final version.
Conflicts of interest
ADH is a member of the editorial board of Drug and Therapeutics
Bulletin and has acted as an adviser to GlaxoSmithKline and London
Genetics. He has received honoraria for speaking at educational
meetings sponsored by the pharmaceutical industry and has donated all
or most of these to charity. The other authors declare that they have no
conflicts of interest.
Acknowledgments
A proportion of DW’s salary was received by the University College
London/University College London Hospital from the Department of
Health’s National Institute for Health Research Biomedical Research
Centre. ADH is a British Heart Foundation Senior Research Fellow
(FS051/125). We thank Reecha Sofat for statistical support.
References
1 Proceedings of the 4th International Workshop-Conference on
Gestational Diabetes Mellitus. Diabetes Care 1998;
21 (suppl 2): B1–167.
2 Ben Haroush A, Yogev Y, Hod M. Epidemiology of gestational
diabetes mellitus and its association with type 2 diabetes.
Diabet Med 2004; 21: 103–13.
3 Centers for Disease Control and Prevention. National diabetes fact
sheet: general information and national estimates on diabetes in
the United States, 2007. Atlanta: Department of Health and Human
Services, Centers for Disease Control and Prevention, 2008.
4 Kim C, Newton KM, Knopp RH. Gestational diabetes and the
incidence of type 2 diabetes. Diabetes Care 2002; 25: 1862–68.
5 Järvelä IY, Juutinen J, Koskela P, et al. Gestational diabetes
identifies women at risk for permanent type 1 and type 2 diabetes in
fertile age. Predictive role of auto-antibodies. Diabetes Care 2006;
29: 607–12.
6 Hunt KJ, Schuller KL. The increasing prevalence of diabetes in
pregnancy. Obstet Gynecol Clin North Am 2007; 34: 173–99.
7 Lipscombe LL, Hux JE. Trends in diabetes prevalence, incidence,
and mortality in Ontario, Canada 1995–2005: a population-based
study. Lancet 2007; 369: 750–56.
8 American Diabetes Association. Economic costs of diabetes in the
US in 2002. Diabetes Care 2003; 26: 917–32.
9 Report of the expert committee on the diagnosis and classification
of diabetes mellitus. Diabetes Care 1997; 20: 1183–97.
10 Kim C, McEwen LN, Piette JD, et al. Risk perception for diabetes
among women with histories of gestational diabetes mellitus.
Diabetes Care 2007; 30: 2281–86.
11 Kim C, Tabaei BP, Burke R, et al. Missed opportunities for type 2
diabetes mellitus screening among women with a history of
gestational diabetes mellitus. Am J Public Health 2006; 96: 1643–48.
12 National Diabetes Data Group (NDDG). Classification and
diagnosis of diabetes mellitus and other categories of glucose
intolerance. Diabetes 1979; 28: 1039–57.
13 WHO. Diabetes Mellitus: report of a WHO study group. Technical
report series, number 727. Geneva: World Health Organization,
1985, 1998, and 1999.
14 Expert committee on the diagnosis and classification of diabetes
mellitus: report of the expert committee on the diagnosis and
classification of diabetes mellitus. Diabetes Care 1997; 20: 1183–97.
15 Lee H, Jang H, Park H, Metzger B, Cho N. Prevalence of type 2
diabetes among women with a previous history of gestational
diabetes mellitus. Diabetes Res Clin Pract 2008; 81: 124–29.
16 Linné Y, Barkeling B, Rössner S. Natural course of gestational
diabetes mellitus: long term follow up of women in the SPAWN
study. Br J Obstet Gynaecol 2002; 109: 1227–31.
17 Ferraz TB, Motta RS, Ferraz CL, Capibaribe DM, Forti AC,
Chacra AR. C-reactive protein and features of metabolic syndrome
in Brazilian women with previous gestational diabetes.
Diabetes Res Clin Pract 2007; 78: 23–29.
www.thelancet.com Vol 373 May 23, 2009

18 Feig D, Zinman B, Wang X, Hux J. Risk of development of diabetes
mellitus after a diagnosis of gestational diabetes. CMAJ 2008;
179: 229–34.
19 Stroup DF, Berlin JA, Morton SC, et al. (MOOSE group)
Meta-analysis of observational studies in epidemiology, a proposal
for reporting. JAMA 2000; 283: 2008–12.
20 Centre for Reviews and Dissemination. Undertaking Systematic
reviews of research on effectiveness: CRD’s guidance for those
carrying out or commissioning reviews. Report number 4. York:
Centre for Reviews and Dissemination, 2001.
21 Madarász E, Tamás Gy, Tabák Á, Szalay J, Kerényi Z. Metabolic
syndrome after pregnancy complicated with gestational diabetes:
four-year follow-up. Orv Hetil 2008; 149: 831–38.
22 Gunderson EP, Lewis CE, Tsai AL, et al. A 20-year prospective study
of childbearing and incidence of diabetes in young women,
controlling for glycaemia before conception. The Coronary Artery
Risk Development in Young Adults (CARDIA) study. Diabetes 2007;
56: 2990–96.
23 Vambergue A, Dognin C, Boulogne A, Réjout MC, Biausquet S,
Fontaine P. Increasing incidence of abnormal glucose tolerance in
women with prior abnormal glucose tolerance during pregnancy:
DIAGEST 2 study. Diabet Med 2008; 25: 58–64.
24 Lee AJ, Hiscock RJ, Wein P, Walker SP, Permezel M. Gestational
diabetes mellitus: Clinical predictors and long-term risk of
developing type 2 diabetes. Diabetes Care 2007; 30: 878–83.
25 Krishnaveni GV, Hill JC, Veena SR, et al. Gestational diabetes and the
incidence of diabetes in the 5 years following the index pregnancy in
South Indian women. Diabetes Res Clin Pract 2007; 78: 398–404.
26 Morimitsu LK, Fusaro AS, Sanchez VH, Hagemann C, Bertini AM,
Dib SA. Fibrinolytic dysfunction after gestation is associated to
components of insulin resistance and early type 2 diabetes in Latino
women with previous gestational diabetes. Diabetes Res Clin Pract
2007; 78: 340–48.
27 Albareda M, Caballero A, Badell G, et al. Diabetes and abnormal
glucose tolerance in women with previous gestational diabetes.
Diabetes Care 2003; 26: 1199–205.
28 Åberg A, Jönsson E, Eskilsson I, Landin-Olsson M, Frid A.
Predictive factors of developing diabetes mellitus in women with
gestational diabetes. Acta Obstet Gynecol Scand 2002; 81: 11–16.
29 Bian X, Gao P, Xiong X, et al. Risk factors for development of
diabetes mellitus in women with a history of gestational diabetes
mellitus. Chin Med J 2000; 113: 759–62.
30 Ko G, Chan J, Tsang L, Chi-Yin-Li, Cockram C. Glucose intolerance
and other cardiovascular risk factors in Chinese women with a
history of gestational diabetes mellitus. Aust NZ J Obstet Gynaecol
1999; 39: 478–83.
31 Osei K, Gaillard TR, Schuster DP. History of gestational diabetes
leads to distinct metabolic alterations in nondiabetic
African–American women with a parental history of type 2 diabetes.
Diabetes Care 1998; 21: 1250–57.
32 Damm P, Buschard K, Jakobsen B, et al. Prevalence and predictive
value of islet cell antibodies and insulin auto antibodies in women
with gestational diabetes. Diabet Med 1994; 11: 558–63.
33 Benjamin E, Winters D, Mayfield J, Gohdes D. Diabetes in
pregnancy in Zuni Indian woman: prevalence and subsequent
development of clinical diabetes after gestational diabetes.
Diabetes Care 1993; 16: 1231–35.
34 O’Sullivan JB. The Boston gestational diabetes studies: review and
perspectives. In: Sutherland, Stowers, and Pearson, eds.
Carbohydrate metabolism in pregnancy and the newborn IV.
London: Springer-Verlag Berlin Heidelberg, 1984: 287–94.
35 Persson B, Hanson U, Hartling S, Binder C. Follow up of women
with previous GDM. Insulin, C-peptide and proinsulin responses to
oral glucose load. Diabetes 1991; 40 (suppl 2): 136–41.
36 Canadian Institute for Health Information. Discharge Abstract
Database. http://secure.cihi.ca/cihiweb/dispPage.jsp?cw_
page=services_dad_e (accessed May 7, 2009).
www.thelancet.com Vol 373 May 23, 2009
Articles
37 Hux JE, Ivis F, Flintoft V, Bica A. Diabetes in Ontario:
determination of prevalence and incidence using a validated
administrative data algorithm. Diabetes Care 2002; 25: 512–16.
38 National Diabetes Data Group (NDDG). Classification and
diagnosis of diabetes mellitus and other categories of glucose
intolerance. Diabetes 1979; 28: 1039–57.
39 WHO. Definition, diagnosis and classification of diabetes mellitus
and its complications. Report of a WHO consultation, part 1:
diagnosis and classification of diabetes mellitus. Geneva: World
Health Organisation, 1999. http://whqlibdoc.who.int/hq/1999/
WHO_NCD_NCS_99.2.pdf (accessed May 7, 2009).
40 WHO. Definition, diagnosis and classification of diabetes mellitus
and its complications . Report of a WHO consultation, part 1:
diagnosis and classification of diabetes mellitus. Geneva: World
Health Organisation, 1999. http://whqlibdoc.who.int/hq/1999/
WHO_NCD_NCS_99.2.pdf (accessed May 7, 2009).
41 Carpenter MW, Coustan DR. Criteria for screening tests for
gestational diabetes mellitus. Am Obstet Gynecol 1982; 144: 768–73.
42 Hoffman L, Nolan C, Wilson J, Oats J, Simmonds D. Gestational
diabetes mellitus: management guidelines. Med J Aust 1998;
169: 93–97.
43 WHO. Definition, diagnosis and classification of diabetes mellitus
and its complications . Report of a WHO consultation, part 1:
Diagnosis and classification of diabetes mellitus. Geneva: World
Health Organisation 1999. http://whqlibdoc.who.int/hq/1999/
WHO_NCD_NCS_99.2.pdf (accessed May 7, 2009).
44 Teramo K, Leinonen P. Diabetes ja raskaus. Kirjassa: Ylikorkala O,
Kauppila A. toim. Naistentaudit ja synnytykset. Pieksämäki:
Kustannus Oy Duodecim 1996: 513–22. http://www.diabetes.fi/sivu.
php?artikkeli_id=473 (accessed May 7, 2009).
45 Reunanen A, Kangas T, Martikainen J, Klaukka T. Nationwide
survey of comorbidity,use, and costs of all medications in Finnish
diabetic individuals. Diabetes Care 2000; 23: 1265–71.
46 American Diabetes Association: Second Workshop-Conference on
Gestational diabetes: summary and recommendations. Diabetes
Care 1985; 34(suppl 2): 123–26.
47 Metzger BE, the organizing committee. Summary and
recommendations of the Third International Workshop–
Conference on Gestational Diabetes Mellitus. Diabetes 1991;
40 (suppl 2): 197–201.
48 Lind T, Phillips PR. Influence of pregnancy on the 75g OGTT. A
prospective multicentre study. The Diabetic Pregnancy Study Group
of the European Association for the study of diabetes. Diabetes 1991;
40: 8–13.
49 National Diabetes Data Group (NDDG). Classification and
diagnosis of diabetes mellitus and other categories of glucose
intolerance. Diabetes 1979; 28: 1039–57.
50 O’Sullivan JB, Mahan CM. Criteria for the oral glucose tolerance
test in pregnancy. Diabetes 1964; 13: 278–85.
51 WHO. Diabetes mellitus: report of a WHO study group. Technical
report series, number 727. Geneva: World Health Organization, 1985.
52 Lauenborg J, Grarup N, Damm P, et al. Common type 2 diabetes
risk gene variants associate with gestational diabetes.
J Clin Endocrinol Metab 2009; 94: 145–50.
53 Cho YM, Kim TH, Choi SH, et al. Type 2 diabetes-associated
genetic variants discovered in the recent genome-wide association
studies are related to gestational diabetes in the Korean population.
Diabetologia 2009; 52: 253–61.
54 Metzger BE, Cho NH, Roston SM, Radvany R. Prepregnancy weight
and antepartum insulin secretion predict glucose intolerance five
years after gestational diabetes. Diabetes Care 1993; 16: 1598–605.
55 Stage E, Ronneby H, Damm P. Lifestyle change after gestational
diabetes. Diabetes Res Clin Pract 2004; 63: 67–72.
56 Ratner RE, Christophi CA, Metzger BE, et al. Prevention of diabetes
in women with a history of gestational diabetes: effects of
metformin and lifestyle interventions. J Clin Endocrinol Metab 2008;
93: 4774–79.
1779
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.