Metabolic and polycystic ovary syndromes in indigenous South Asian women with previous gestational diabetes mellitus


Prof CN Wijeyaratne, Department of Obstetrics & Gynaecology, Faculty of Medicine, University of Colombo, PO Box 271, Kynsey Road, Colombo 007, Sri Lanka. Email


Objective  To determine the risk of metabolic syndrome (MS) and polycystic ovary syndrome (PCOS) in a cohort of indigenous South Asian women with a recent history of gestational diabetes mellitus (GDM).

Design  Case–control study.

Setting  Department of Obstetrics & Gynaecology, University of Colombo, Sri Lanka.

Sample  Two hundred and seventy-four indigenous Sri Lankan women with previous GDM and 168 ethnically matched controls. Of these, 147 with previous GDM and 67 controls not taking hormonal contraception participated in an in-depth endocrine study.

Methods  Assessing the prevalence of MS and PCOS based on clinical features, biochemistry and ovarian ultrasound examination at 3 years postpartum.

Main outcome measures  Prevalence of MS and PCOS.

Results  Women with previous GDM and controls were studied at a mean (range) of 34.6 (13.4–84.1) and 46.5 (17.5–78) months postpartum, respectively. Those with previous GDM had a larger mean ± 95% confidence interval waist circumference (90.9 ± 1.3 versus 81.2 ± 2.8 cm, P= 0.0004) and were more likely to have hypertension (17.6 versus 7.4%, P= 0.001), glucose intolerance (51.7 versus 10.4%, P= 0.00001), hypertriglyceridaemia (16.3 versus 5.9%, P= 0.02) and a lower level of high-density lipoprotein (70 versus 56.7%, P= 0.04) than the controls. Of the women who had GDM, 72 (49%) had MS, 86 (58.5%) had polycystic ovaries and 59 (40%) had PCOS, significantly more than the control women—4 (6%), 9 (13%) and 2 (3%), respectively (P= 0.00001).

Conclusions  The prevalence of MS and PCOS in indigenous Sri Lankan women 3 years postpartum is significantly higher in those with previous GDM compared with ethnically matched controls. This confirms an association between GDM and subsequent PCOS and MS.


Glucose intolerance, hypertension and hyperlipidaemia are recognised causes of premature atherosclerosis. The metabolic syndrome (MS), a manifestation of the prediabetic state, is associated with a high risk of premature atherosclerosis.1,2 Polycystic ovary syndrome (PCOS), the most common endocrine disorder among women of reproductive age worldwide, is also closely linked to the MS.3 Gestational diabetes mellitus (GDM) is a metabolic disorder considered to be similar to type II diabetes mellitus (T2DM) where pregnancy acts as a ‘stressor’, revealing an individual’s underlying diabetic potential.4 Although the clinical importance of GDM is principally related to pregnancy outcomes, follow-up studies of women who had GDM have shown that many, despite normal short-term postpartum glucose tolerance testing, subsequently develop insulin resistance (IR) and progress to develop T2DM and atherosclerosis.5,6

Global diabetes prevalence estimates for 2025 predict that the greatest increases in diabetes prevalence will occur in the Indian subcontinent.7,8 These predictions are being borne out by recent observations.8,9 This is mirrored by reports of a high incidence of GDM among migrant South Asians10–15 and a high prevalence of obesity and T2DM among indigenous South Asian populations including those in Sri Lanka.16–18 Although there is an accepted relationship between ethnicity, GDM, PCOS and T2DM,4,8 there are few follow-up studies of women with GDM from high-risk indigenous populations that have assessed the risk of longer term MS. Accordingly, we wished to determine the prevalence of PCOS and MS in indigenous Sri Lankan women who had had GDM in their last pregnancy and compare this with ethnically matched women who had not had GDM.


A case–control study was carried out between 2003–05 on consecutive women managed at the University Maternity Unit of De Soysa Hospital for Women (DSHW), Colombo, for GDM between January 1999 and June 2002. GDM was diagnosed and defined as per World Health Organization (WHO) guidelines using a 75-g, 2-hour oral glucose tolerance test (OGTT).19 Women with GDM were managed in a dedicated combined antenatal–diabetic clinic according to American College of Obstetricians and Gynaecologists guidelines.20 All women had a postpartum OGTT performed while on a normal diet at 6–8 weeks after delivery. We included women as cases, whose last pregnancy was managed for impaired glucose tolerance or frank diabetes at the University Unit of DSHW and who were beyond 12 months postpartum at the time of recall. Migrant women, multiple pregnancies, those with established diabetes mellitus before the index pregnancy, Cushing’s disease, acromegaly and phaeochromocytoma and those taking medication for epilepsy, psychiatric illness or steroids were excluded. A random sample of ethnically matched control women from the same catchment area, who delivered in the same unit during the same period and whose hospital records confirmed maternal normoglycaemia during pregnancy were also invited to participate. Ethical approval was obtained from the Ethical Review Committee of The Faculty of Medicine, University of Colombo. Informed written consent was obtained from all responders.

Three medical doctors recorded a detailed history from the previous-GDM group and controls and cross-checked data with hospital case notes and antenatal records. In the absence of documented preconception maternal body mass index (BMI), we analysed the BMI recorded at antenatal booking. A subgroup of women who were currently not pregnant and not taking hormonal contraception were invited to join an in-depth endocrine evaluation following an overnight 12-hour fast and within the first 7 days of the follicular phase of their next natural menstrual cycle. A detailed menstrual history since menarche, problems with ovulation, skin manifestations of PCOS such as hirsutism, acne and acanthosis nigricans, a history of hypertension and medication were recorded. Their current anthropometry (height, weight, hip and waist circumferences [WC]), skin manifestations of PCOS and resting blood pressure (BP) in the sitting posture were recorded. Anthropometry was measured using standardised equipment and procedures, and the WC was measured to the nearest 0.1 cm at the upper border of the iliac crest while the subject was at minimal respiration.

Analysis of fasting plasma glucose (FPG), lipids, fasting insulin, testosterone and sex hormone binding globulin (SHBG) was carried out at the laboratory of the Department of Gynaecology and Obstetrics, Faculty of Medicine, University of Colombo, blinded to the case or control status. Assays were performed on batched samples, consisting of combinations of cases and controls. Plasma glucose was measured using an enzymatic colorimetric assay (Hitachi®; Roche, Japan) with intra-assay coefficient of variation (CV) of 2.9%. Plasma insulin, serum SHBG and testosterone were measured using a time-resolved fluoroimmunoassay (AutoDELFIA®; Perkin Elmer, Los Angeles, CA), with plasma insulin intra-assay CVs of 1.9% and inter-assay CVs of 2.5%, serum SHBG intra-assay CVs of 3.8% and inter-assay CVs of 3.1% and serum testosterone inter-assay CV of 5.4%.

Estimates of IR were derived using the homeostasis model assessment (HOMA) algorithm (version 2.2; Diabetes Trials Unit, University of Oxford, Oxford, UK). Transvaginal pelvic ultrasound examination was carried out at the DSHW, Colombo (Aloka® SSD 1000 computed sonography system; Mitaka-shi, Tokyo, Japan, with a 5-mHz transducer). The principal investigator and two trained obstetrician and gynaecologists performed the scans alternately and were blinded to the clinical details of individual women. Diagnosis of polycystic ovaries (PCO) was based on revised consensus diagnostic criteria, i.e. ovarian volume > 10 cm3 and/or 2- to 9-mm follicles exceeding 12 in number in a single plane.21,22 Calculation of ovarian volume was performed using the simplified formula for a prolate ellipsoid (0.5 × length × width × thickness). Follicle number was estimated both in longitudinal and anteroposterior cross-sections of the ovaries. Follicle size was expressed as the mean of the diameters measured in the two planes perpendicular to each other.

The prevalence of PCOS was determined among cases and controls based on 2003 revised diagnostic criteria.21 The prevalence of the MS was determined using the 2005 diagnostic criteria of the International Diabetes Federation (IDF)—a WC ≥ 80 cm for European women or ethnicity-specific values and any two of the following: fasting plasma triglyceride (TG) > 1.69 mmol/l (>150 mg/dl), high-density lipoproteins (HDL) < 1.3 mmol/l (<50 mg/dl), systolic BP ≥ 130 mmHg or diastolic BP ≥ 85 mmHg or previously diagnosed hypertension with/without treatment, FPG ≥ 5.5 mmol/l (100 mg/dl) or previously diagnosed T2DM.23 In the absence of local gender-specific data for WC, the 95th centile of controls was taken as the cutoff for diagnosis of MS.

Sample size calculation

With a reported prevalence of GDM in Sri Lanka of 10.3% and an estimated prevalence of PCOS of ∼6%, we determined a sample size of 260 consecutive women with previous GDM and 173 postpartum women without diabetes (leaving an allowance for 10% drop-outs) using Epi Info Version 3.2.2 (WHO, Geneva, Switzerland).

Statistical analysis

Data were analysed using SPSS® (SPSS Inc., Chicago, IL) statistical package, version 10.0. To determine differences between GDM and controls, chi-square test was used for categorical data and Student’s t test for continuous variables. Bivariate analysis applying Pearsons’ correlation coefficient tested correlation between PCOS and the MS. Two-tailed P values < 0.05 were considered statistically significant.


Three hundred and ninety-three letters were mailed to women with a history of GDM and to 233 ethnically matched control women. Of these, 274 (69%) and 168 (72%) responded, respectively.

Table 1 denotes demography and first-trimester BMI in the GDM and control groups. The mean ± 95% confidence interval (CI) interval between the previous pregnancy and this study was 34.6 ± 2.2 months for the women with GDM and 46.5 ± 3.7 months for control women. There were no differences with regard to age, race or parity between the two groups. The mean ± 95% CI BMI in early pregnancy of women who developed GDM was 26.3 ± 0.9 kg/m2, compared with 23 ± 0.7 kg/m2 for control women (P= 0.00001).

Table 1.  Demographic data and BMI at antenatal booking of women managed between 1999–2002 for GDM during the index pregnancy compared with controls managed in the same unit
CharacteristicGDM, n= 274Controls, n= 168P Value
  • *

    Significant P < 0.05, Student’s t test.

  • POA, Period of amenorrhoea.

  • NS, Not significant.

Racial groups among Sri Lankan women, n (%)
Sinhalese245 (89)151 (89)NS
Tamil16 (5.8)11 (6.5) 
Muslim12 (4.4)6 (3.4) 
Others1 (0.03)0 
Mean age ± 95% CI at index pregnancy (years)33.4 ± 0.632.3 ± 0.8NS
Parity (P), n (%)
P164 (22)43 (26) 
P2102 (37)53 (31)NS
P357 (21)45 (27) 
>P351 (20)27 (16) 
T2DM in first-degree relative150 (54.7)*59 (35)0.004
Maternal BMI at POA 11 weeks (kg/m2) (mean ± 95% CI)26.3 ± 0.923 ± 0.70.00001
Postpartum duration at recall (months) (mean ± 95% CI)34.6 ± 2.246.5 ± 3.7NS

Prevalence of MS at postpartum review

The mean BMI at postpartum review in women with previous GDM, not taking hormonal contraception, was 24.3 ± 3.4 kg/m2 compared with 21.65 ± 3.8 kg/m2 in controls (P= 0.001). The mean ± 95% CI WC was 90.9 ± 1.3 cm in the previous-GDM group versus 81.2 ± 2.8 cm in controls (P= 0.0004). Hundred and forty women out of 147 (95%) of the previous-GDM group and 35 out of 67 (52%) controls had a WC exceeding 80 cm, which is recommended for European women as per IDF guidelines of 2005.23 In the previous-GDM group, eight women had established hypertension and another 18 were diagnosed by the study, whereas among controls, one had established hypertension and another four were diagnosed by the study (P= 0.001). Twenty women in the previous-GDM group were diagnosed as with T2DM approximately 2 years postpartum, and a further 56 (44%) from the remainder were identified by this study to have a FPG > 5.5 mmol/l, compared with one woman in the control group who had developed T2DM and another six (9%) identified with FPG > 5.5 mmol/l (P= 0.00001). Exclusive of the 20 women with established T2DM, women with previous GDM had significantly impaired glucose handling compared with controls as evidenced by higher mean ± 95% CI fasting glucose (6.8 ± 0.6 versus 5.1 ± 0.2 mmol/l; P= 0.001), higher mean ± 95% CI fasting insulin (100.9 ± 6.8 versus 47.7 ± 2.9 pmol/l; P= 0.001) and increased IR estimated by HOMA algorithm (1.85 versus 1.02; P= 0.0001). Women with previous GDM also had higher mean ± 95% CI fasting TG (1.2 ± 0.05 versus 0.96 ± 0.01 mmol/l; P= 0.001) and lower mean ± 95% CI HDL (1.16 ± 0.08 versus 1.29 ± 0.01 mmol/l; P= 0.03). Overall, the prevalence of MS in the women with previous GDM was 72/147 (49%) compared with 4/67 (6%) among controls (P= 0.00001).

Prevalence of PCOS at postpartum review

Eighty-six (58.5%) women with previous GDM and nine (13.4%) control women fulfilled the ultrasound diagnostic criteria of PCO (P= 0.00001). The prevalence of PCOS in the women with previous GDM versus controls was 40 versus 3% (P= 0.00001). Mean ovarian volume in the PCO group with previous GDM was significantly greater than women from the same group with normal ovaries (see Table 2, P= 0.002). Those with PCO and previous GDM had significantly greater prevalence of oligomenorrhoea, acanthosis nigricans, hirsutism and infertility than the women without PCO from the same group. Although BMI and BP were similar between subjects with PCO and those without PCO within the GDM subgroup, the waist–hip ratio was significantly greater (P= 0.01) among the subjects with PCO who had GDM previously. FPG and insulin concentrations were significantly higher, with HDL and SHBG concentrations significantly lower in those with PCO and previous GDM than those without PCO who had previous GDM (Table 2).

Table 2.  Characteristics of PCOS and related metabolic features among GDM women with and without PCO
Characteristics of PCOSGDM group, n= 147P value
PCO, n= 86Non-PCO, n= 61
  • FG, Ferriman–Galleway score; NS, Not significant; WHR, waist-to-hip ratio.

  • *

    Mean ± 95% CI.

Ovarian volume (cm3)
Menstrual irregularity16 (18.6%)4 (6.5%)0.02
Acanthosis nigricans47 (55%)20 (32%)0.03
Hirsutism, FG score > 711 (13%)4 (5%)0.009
Infertility23 (27%)9 (15%)0.007
BMI (kg/m2)25.4 ± 3.3*23.2 ± 4.2*0.04
WHR0.97 ± 0.1*0.92 ± 0.1*0.01
Systolic BP (mmHg)116115NS
Diastolic BP (mmHg)79.677.9NS
FPG (mmol/l)7.2 ± 0.8*6.6 ± 0.3*0.01
Insulin (pmol/l)
IR (HOMA)2.4 ± 0.18*1.61 ± 0.13*0.001
Cholesterol (mmol/l)4.874.99NS
TG (mmol/l)
HDL (mmol/l)0.83 ± 0.18*1.140.004
Testosterone (nmol/l)2.21 ± 0.04*1.98 ± 0.07*NS
SHBG (nmol/l)1.15 ± 0.02*1.34 ± 0.03*0.02

Of the 72 women with previous GDM who fulfilled the diagnostic criteria of MS, 18 (25%) also fulfilled the PCOS diagnostic requirements.


The striking revelation of this study is the high prevalence of MS (49%) observed in this cohort of indigenous South Asian women 3 years following previous GDM, who were normoglycaemic at 6–8 week following delivery and currently in their mid-30s. Kousta et al.24 who assessed the impact of ethnicity on the metabolism of women with previous GDM using the same diagnostic criteria at 20 months postpartum reported a similarly high prevalence of MS among migrant South Asians with GDM. Our results findings being similar to those reported in British Asian women with previous GDM suggest that ethnicity rather than migration places South Asian women with previous GDM at risk of developing the MS within 3 years of pregnancy. A high prevalence of MS in these women in their fourth decade of life predicts premature coronary artery disease (CAD). This hypothesis is supported by reports from other studies carried out on European women with previous GDM, who despite regaining glucose tolerance soon after pregnancy showed subtle yet significant differences from controls in fasting lipids, BP and microvascular/macrovascular function that are consistent with an increased risk of diabetes and CAD.5,6 It is noteworthy that Kousta et al.24 who compared different ethnic groups with previous GDM found that far less Europeans had MS (28%) when compared with Asian-Indian (49%) and African-Caribbean women (43%). Hence, the ethnic origin of women with previous GDM must be taken into account in their long-term follow up. This is supported by a recent epidemiological report of a migrant aboriginal population which revealed that the ethnic origin was an independent predictor of GDM, with an associated increase in birthweight, long-term T2DM and diabetic end-stage renal disease, where the author speculates that an ancient survival advantage is now proving to be disadvantageous with a high metabolic risk and a possible intergenerational perpetuation.25

PCO were reported to be common among European women with previous GDM that led to the recommendation that all women known to have PCO/PCOS should be screened for GDM.26 A Swedish group also reported similarly that women with a history of GDM with PCO formed a distinct subgroup that is more prone to develop the IR syndrome.27 Our data support this to be indeed the case among indigenous South Asian women where women with PCO and previous GDM had IR as a dominant component. We previously reported that nonpregnant indigenous Sri Lankan women with PCOS in the third decade of life had a far greater degree of IR and cardiovascular risks than their white European counterparts.28 More recently, a South-East Asian study confirmed a similar association between GDM, IR and PCOS among Thai women.29 It is noteworthy that Ehrmann et al. who evaluated the predictive value of race and a family history of T2DM on the metabolic status of a large cohort of women with PCOS found that these factors have a substantial impact on the metabolic and glycaemic status of women. They also evaluated these parameters in women without PCOS and found that a positive family history of T2DM correlated with central obesity and IR, which is mirrored in our study.30 Although the link between GDM, PCOS and the MS are well known, this study highlights the need for continuing monitoring and management of women with GDM from the Asian communities.

Limitations of our study were that it involved a recall method and was therefore vulnerable to selection bias by responders. Nevertheless, this is likely to have been minimised by a response rate of ∼70% from a consecutive sample managed in a single unit and the 81% response from those among the responders who qualified for an in-depth metabolic and endocrine assessment. Moreover, having a control arm that was matched for ethnicity should have further minimised possible effects from bias. The absence of pre-pregnancy BMI and hence the study of first-trimester BMI data of those who booked early was another drawback. Nevertheless, the observation that first-trimester obesity was a significant risk factor for developing GDM matched the high 3-year postdelivery BMI and WC observed among the same group of women with previous GDM when compared with ethnically matched controls.

In the past few decades, Sri Lanka has experienced rapid urbanisation and a change in lifestyle that has contributed to an exponential rise in the prevalence of obesity and the burden of chronic noncommunicable diseases such as diabetes mellitus and cardiovascular disease.31,32 This has increased the burden of GDM that was confirmed in 2003 by a study based in a defined rural population in Sri Lanka that revealed an age-standardised GDM prevalence of 10.3%, which is a three-fold rise from that of a similar population in 1987.33,34 The current study was conducted in a tertiary referral centre, which confirmed that the first-trimester maternal BMI exceeding 23.8 kg/m2 has a risk for developing GDM. It is noteworthy that the mean WC of Sri Lankan women measured 3 years after the pregnancy complicated by GDM was significantly greater with an ‘overweight’ BMI (for Asians) than ethnically matched postpartum controls. However, the cutoff for WC was much greater than the IDF recommendation for European women and those recommended for South-East Asian women.23,35 The greater WC observed among the Sri Lankan control women and a high positive family history of T2DM (35%) support the hypothesis proposed by Dhawan et al.15 that South Asians have a greater risk of central adiposity that correlates positively with IR. Hence, it is reasonable to suggest that South Asian descent alone makes a woman more vulnerable to central obesity but has additional contributory factor(s) that lead them to develop GDM, T2DM and/or PCOS.

Central obesity among Asians and its impact on chronic disease is further highlighted by a recent national survey of 15 540 Chinese adult men and women aged 35–74 years that revealed WC was significantly associated with diabetes and cardiovascular risks.36 Furthermore, a comparison of visceral adipose tissue content and T2DM among postmenopausal Filipino, African-American and white women reported the highest prevalence of T2DM among the Filipino women who also had the highest visceral fat content.37 It is noteworthy that the third National Health and Nutrition Examination Survey, which examined the effects of BMI and WC (independently and combined) on obesity-related morbidity, revealed that the metabolic and cardiovascular risks of obesity are explained by WC and not by BMI.38 Therefore, South Asian women with previous GDM who are centrally obese postpartum are likely to have an increased risk of metabolic disease, which is supported by our findings.


We conclude that there is a strong association between PCOS, previous GDM and long-term MS among indigenous Sri Lankan women. Extrapolating this evidence to high-risk women with previous GDM and/or PCOS in the Sri Lankan setting suggests an urgent need to institute health promotion and preventive measures to these high-risk groups of women in order to minimise long-term effects of the MS. Educating those with PCOS and previous GDM should be targeted towards lifestyle modification to improve insulin sensitivity and towards maintaining an ideal body weight and lower WC through dietary modification and physical activity.


We thank the head, academic staff, consultants and support staff of the Department of Obstetrics & Gynaecology, Faculty of Medicine, University of Colombo, Sri Lanka; Director and administrative staff of DSHW, Colombo, Sri Lanka; laboratory technical officers Mrs ASDS Wijeratne, Mrs AUA Gunewardhena and Mrs M Warnakulasuriya for their excellent laboratory support. Dr Paul Belchetz, consultant endocrinologist, Leeds General Infirmary, UK, for his support. Participants and their families for their ready cooperation.


Association of Physicians of Great Britain & Ireland; Special Trustees of the General Infirmary at Leeds, UK.