Study Importance What is already known? ► Excess adiposity and adipose tissue dysfunction are features of pregnancies complicated by gestational diabetes (GDM), but whether excess adiposity or persistent adipose tissue dysfunction contributes differentially to risk of subsequent type 2 diabetes (T2D) in parous women with and without a history of GDM is unknown. What does this study add? ► Among parous women, alanine aminotransferase (as a surrogate marker of liver fat content) and visceral adipose tissue area are positively associated with incident T2D, whereas adiponectin level is inversely associated with T2D. The association between leptin and T2D depends on history of GDM. How might these results change the direction of research? ► Given the interrelationships of leptin, adiponectin, and glucose metabolism, future studies should include measures of insulin sensitivity and secretion, as well as body composition, in cohorts of women with and without a history of GDM. INTRODUCTION

Gestational diabetes mellitus (GDM) affects 10% to 15% percent of parous women ((1)) and confers up to a sevenfold higher risk of type 2 diabetes (T2D) compared with parous women without a history of GDM ((2)). In the Diabetes Prevention Program (DPP), during the initial 3 years of the intervention, the rate of progression to diabetes in women with a history of GDM in the placebo group was much higher than in women without a history of GDM (15.2 vs. 8.9 cases per 100 person-years, respectively), despite similar levels of glycemia at baseline ((3)). Although insulin resistance and reduced β-cell function both contribute to the development of GDM and subsequent T2D ((4-7)), the specific mechanisms responsible remain poorly understood.

Adipose tissue dysfunction is a feature of pregnancies complicated by GDM ((8)). Persistence of adipose tissue dysfunction and insulin resistance following parturition may also contribute to the chronically elevated risk of T2D among women with a history of GDM. Consistent with this hypothesis, at 1 year post partum, a history of GDM is independently associated with lower circulating adiponectin levels ((9)), an adipokine that is involved in fat metabolism and the regulation of blood glucose ((10)). Sex hormone-binding globulin (SHBG) has a consistent inverse association with pregestational and gestational obesity, as well as with metabolic biomarkers such as glucose, C-peptide, and adiponectin ((11)). Furthermore, women with a history of GDM exhibit ectopic fat distribution, having 2.5-fold higher odds of hepatic steatosis when compared with parous women without a history of GDM up to 25 years after parturition ((12)), a risk factor that is strongly linked to the development of T2D ((13)).

Adiponectin was strongly and inversely associated with risk of T2D in the full DPP cohort ((14)); however, adjustment for baseline adiponectin levels did not explain associations of GDM with subsequent risk of T2D among parous DPP participants ((15)). To our knowledge, whether the association of adipose tissue depots and adiposity-related biomarkers with incident T2D differs in women who do and do not report history of GDM has not been systematically assessed. We hypothesized that these depots and biomarkers would be positively and more strongly associated with T2D risk among women with a history of GDM compared with women who did not report a history of GDM. Therefore, we investigated whether associations of the circulating biomarkers leptin, adiponectin, SHBG, and alanine aminotransferase ([ALT] as a surrogate marker for liver fat content) ((16)) or imaging-based visceral (VAT) and subcutaneous (SAT) adipose tissue depots with incident T2D might vary among women who did or did not report a history of GDM in the DPP.

METHODS Study setting

The DPP was a NIH-funded, randomized controlled trial designed to determine whether lifestyle intervention (ILS) or pharmacological therapy (metformin) would prevent or delay the onset of T2D in individuals at high risk for the disease. The study was performed at 27 clinical centers throughout the United States. Details of the DPP study setting ((17)) and population ((18)) have been reported previously. In brief, eligible participants were >25 years old with a fasting plasma glucose concentration of 5.3 to 6.9 mmol/L (or ≤6.9 mmol/L in the American Indian Centers), a 2-hour value after a 75-g glucose load of 7.8 to 11.1 mmol/L, and BMI ≥ 24 kg/m2 (≥22 for Asian American individuals). Women with a history of GDM were intentionally recruited ((17)). After enrollment, participants were randomized to one of three study arms (placebo, metformin, and ILS; n = 3,234). Over a 3.2-year period of study follow-up, they were assessed annually with a 2-hour oral glucose tolerance test and semiannually with a fasting plasma glucose. Diabetes was diagnosed according to American Diabetes Association criteria, requiring confirmation by a second test, usually within 6 weeks ((19)).

Study population

Data on past pregnancies were available for n = 2,190 women, of whom n = 1,766 reported at least one live birth. Of these, n = 350 reported a past diagnosis of GDM, and n = 1,416 reported no history of GDM. Among women with a history of GDM, 122 were assigned to placebo, 111 to metformin, and 117 to ILS, and among women without a history of GDM, 487 were assigned to placebo, 464 to metformin, and 465 to ILS. In an ancillary substudy of the DPP, baseline measurement of adipose tissue by computed tomography (CT) was done in 957 participants at 18 of the 27 sites, of whom 477 were parous women. Leptin, adiponectin, SHBG, and ALT were available for n = 1,691, n = 1,648, n = 1,559, and n = 1,711, respectively. The protocol was approved by institutional review boards at each center, and written informed consent was obtained from all participants prior to enrollment.

Data collection

At baseline, women completed a questionnaire reporting past pregnancies, live births, and whether they experienced a previous pregnancy complicated by gestational diabetes. For the 207 women who reported data on date of GDM diagnosis, there was an average 12-year interval between the delivery of their first pregnancy and enrollment in the DPP. Information on time since first pregnancy was not available for women who did not report a history of GDM. Data on lactation duration were not available.

We measured levels of circulating biomarkers (leptin, adiponectin, SHBG, and ALT) and obtained imaging-based VAT and SAT areas at the baseline DPP visit. Plasma leptin and adiponectin concentrations were measured by radioimmunoassay (RIA) (MilliporeSigma). Plasma SHBG was measured using enzyme-linked immunosorbent assay (ELISA) (Bioline AgroSciences). Serum ALT values were measured using the Hitachi 917 chemistry autoanalyzer (Hitachi, Ltd.). Coefficients of variation, as previously reported, were 2.6% to 9.3% for leptin ((14)), 1.1% to 2.0% for adiponectin ((14)), and 5.0% to 7.8% for SHBG ((20)). VAT and SAT areas were calculated from cross-sectional CT-based measurements of body composition obtained in a subset of DPP participants at baseline, and details of the measurements have been reported previously ((21)). In brief, four measurements were obtained (two at each vertebral level, L2-L3 region and L3-L4 region, which were averaged) for each participant. Each participant was used as his/her own control to create a bimodal histogram depicting the distribution of Hounsfield units in the image, resulting in separate peaks for muscle and fat. The adipose tissue area was defined as the area under the fat peak (number of fat pixels) multiplied by the area of one pixel.

Statistical analyses

Baseline characteristics were summarized for the overall group and by history of GDM. Continuous variables with normal distributions (summarized by mean [SD]) were compared between the two groups of women using the Student t test, whereas characteristics with non-normal distributions (summarized as median, 25th percentile, and 75th percentile) were compared using the nonparametric Wilcoxon rank sum test. Categorical variables were summarized as frequency (percentage) and compared between groups using the χ2 test of independence. We examined plots such as those of martingale residuals to explore the functional forms of covariates and confirm that the model was correctly specified. Cox proportional hazards regression models were used to assess the association of biomarkers and fat areas with the risk of T2D, overall and stratified by history of GDM. Multiplicative interaction between history of GDM and the biomarkers and fats areas of interest was evaluated using models with interaction terms. Models were adjusted for treatment assignment, age at randomization, BMI, White race, education level, smoking status, and hormone use. Overall models were also adjusted for history of GDM. Associations were estimated using hazard ratios (HR) for 1-SD increases in each biomarker or fat area and corresponding 95% confidence intervals (CI) and p values. In order to explore potential differences in associations of these measures with incident T2D by race, sensitivity analyses were performed stratifying by race. We also fit models with and without adjustment for menopausal status and BMI. The statistical analysis software SAS version 9.4 (SAS Institute Inc.) was used for the analysis. All tests performed were two-sided at an α level of significance equal to 0.05.

RESULTS

Women were followed for an average of 12.9 years. There were 871 incident cases of T2D. On average, women with a history of GDM were younger and they had lower mean levels of adiponectin and VAT area at the L3-L4 region than women without a history of GDM (Table 1). Among parous women overall, ALT concentration, L2-L3 VAT area, and L3-L4 VAT area were positively associated with risk of T2D (HR for 1-SD differences 1.073 [95% CI: 1.009-1.141], p = 0.024; 1.251 [95% CI: 1.056-1.481], p = 0.009; 1.272 [95% CI: 1.079-1.498], p = 0.004, respectively), whereas adiponectin concentration was negatively associated with risk of T2D (HR 0.762 [95% CI: 0.697-0.832], p < 0.001). Overall, leptin concentration was not associated with risk of T2D; however, in GDM-stratified models, a 1-SD increase in leptin was associated with greater risk of T2D in women without a history of GDM (HR 1.126 [95% CI: 1.022-1.239], p = 0.016) but with lower risk among women with a history of GDM (HR 0.776 [95% CI: 0.635-0.949], p = 0.013; Table 2). The association of leptin with risk of T2D varied significantly by reported history of GDM (p for interaction = 0.002). There were no other significant interactions of a history of GDM with any other biomarkers or adipose tissue depots. In sensitivity analyses to further characterize associations of leptin with incident T2D risk, we did not identify an interaction of race and leptin on T2D risk (p for interaction = 0.319). In a model with adjustment for menopausal status, we found an HR for T2D of 1.067 (95% CI: 0.977-1.165, p = 0.151), compared with an HR of 1.064 in the model without adjustment for menopausal status (95% CI: 0.975-1.162, p = 0.164; Supporting Information Tables S1-S3). In GDM-stratified models without adjustment for BMI, we found an HR for T2D of 1.179 (95% CI: 1.097-1.268, p < 0.001) and 0.880 (95% CI: 0.757-1.022, p = 0.093) in women without and with a history of GDM, respectively, compared with BMI-adjusted HR of 1.126 (95% CI: 1.022-1.239, p = 0.016) and 0.776 (95% CI: 0.635-0.949, p = 0.013) in the primary models.

TABLE 1. Baseline characteristics of parous women in the DPP, overall and stratified by history of GDM Overall (n = 1,711) No history of GDM (n = 1,376) History of GDM (n = 335) p value Age (y) 49.8 ± 9.9 51.5 ± 9.7 43.0 ± 7.7 <0.001 Race (% White) 866 (50.6%) 688 (50.0%) 178 (53.1%) 0.303 Attained ≥high school education (% yes) 1,174 (68.6%) 909 (66.1%) 265 (79.1%) <0.001 Treatment 0.985 Placebo 588 (34.4%) 473 (34.4%) 115 (34.3%) Metformin 552 (32.3%) 445 (32.3%) 107 (31.9%) Lifestyle 571 (33.4%) 458 (33.3%) 113 (33.7%) Weight (kg) 90.6 ± 19.1 90.6 ± 19.4 90.6 ± 17.8 0.962 Waist circumference (cm) 102.9 ± 14.3 103.2 ± 14.4 102.0 ± 13.5 0.184 BMI (kg/m2) 34.4 ± 6.6 34.5 ± 6.8 34.1 ± 6.1 0.280 Fasting glucose (mmol/L) 5.85 ± 0.45 5.85 ± 0.44 5.88 ± 0.47 0.248 2-hour glucose (mmol/L) 9.13 ± 0.95 9.11 ± 0.94 9.21 ± 1.00 0.064 Fasting insulin (pmol/L) 24.0 (16.0, 33.0) 24.0 (16.0, 33.0) 24.0 (17.0, 34.0) 0.442 Alcohol (g/d) 0.00 (0.00, 0.94) 0.00 (0.00, 0.94) 0.00 (0.00, 0.94) 0.091 Statin use (% yes) 64 (3.7%) 62 (4.5%) 2 (0.6%) <0.001 Hormone use (% yes) 438 (25.6%) 399 (29.0%) 39 (11.6%) <0.001 Smoked in the past 30 days (% yes) 114 (6.7%) 95 (6.9%) 19 (5.7%) 0.417 Number of live births 2.0 (2.0, 3.0) 2.0 (2.0, 3.0) 2.0 (2.0, 3.0) 0.489 ALT (mg/dL) 15 (11, 21) 15 (11, 21) 15 (11, 21) 0.430 Leptin (ng/mL) 27.2 (20.2, 36.0) 27.4 (20.2, 36.3) 27.0 (20.2, 33.8) 0.159 Adiponectin (mg/mL) 7.80 (5.90, 10.20) 8.00 (6.20, 10.50) 7.00 (5.40, 9.10) <0.001 SHBG (nmol/L) 44.1 (31.3, 71.7) 43.9 (31.4, 72.4) 44.4 (30.0, 68.2) 0.680 Characteristics among parous women in the DPP body composition substudy Overall (n = 477) No history of GDM (n = 397) History of GDM (n = 80) VAT area at L2-L3 (cm2) 164.93 ± 62.91 166.18 ± 64.56 158.71 ± 53.89 0.333 SAT area at L2-L3 (cm2) 329.34 ± 121.22 332.89 ± 121.85 311.71 ± 117.22 0.154 VAT area at L3-L4 (cm2) 149.90 ± 56.26 152.33 ± 57.28 137.79 ± 49.52 0.035 SAT area at L3-L4 (cm2) 481.54 ± 140.15 484.32 ± 140.07 467.75 ± 140.59 0.335 Data are presented as mean ± SD or median (interquartile range) for continuous variables and n (percentage) for categorical variables. Overall sample size (n = 1,711) includes parous women with available data on leptin, adiponectin, SHBG, and/or ALT. P value from the t test or nonparametric Wilcoxon test, as appropriate, for continuous variables and the χ2 test for categorical variables. Abbreviations: ALT, alanine aminotransferase; DPP, Diabetes Prevention Program; GDM, gestational diabetes mellitus; SAT, subcutaneous adipose tissue; SHBG, sex hormone-binding globulin; VAT, visceral adipose tissue. TABLE 2. Cox proportional hazards regression models testing the association of adiposity-related biomarkers with progression to T2D per 1-SD increase in each biomarker, overall and stratified by history of GDM Overall Women without history of GDM Women with history of GDM p value interaction* HR 95% CI p value HR 95% CI p value HR 95% CI p value ALT (mg/dL) 1.073 1.009-1.141 0.024 1.093 1.018-1.173 0.014 1.040 0.913-1.186 0.553 0.269 Leptin (ng/mL) 1.064 0.975-1.162 0.164 1.126 1.022-1.239 0.016 0.776 0.635-0.949 0.013 0.002 Adiponectin (μg/mL) 0.762 0.697-0.832 <0.001 0.756 0.684-0.837 <0.001 0.776 0.648-0.930 0.006 0.936 SHBG (nmol/L) 0.951 0.879-1.030 0.215 0.936 0.855-1.025 0.156 1.007 0.852-1.190 0.934 0.208 VAT area at L2-L3 (cm2**) 1.251 1.056-1.481 0.009 1.238 1.032-1.486 0.022 1.293 0.806-2.075 0.287 0.691 SAT area at L2-L3 (cm2**) 1.084 0.868-1.354 0.475 1.102 0.861-1.410 0.442 1.076 0.620-1.866 0.796 0.163 VAT area at L3-L4 (cm2**) 1.272 1.079-1.498 0.004 1.246 1.041-1.491 0.017 1.431 0.970-2.112 0.071 0.913 SAT area at L3-L4 (cm2**) 1.064 0.837-1.353 0.610 1.112 0.848-1.457 0.443 1.028 0.551-1.917 0.931 0.068 Models are adjusted for treatment assignment, age at randomization, BMI, White race, attained education level, smoking status, and hormone use. Overall models are also adjusted for history of GDM. Abbreviations: ALT, alanine aminotransferase; DPP, Diabetes Prevention Program; GDM, gestational diabetes mellitus; HR, hazard ratio; SAT, subcutaneous adipose tissue; SHBG, sex hormone-binding hormone; VAT, visceral adipose tissue. * p value for interaction term (GDM × adiposity-related biomarker) from the corresponding regression model testing associations with progression to T2D among all parous DPP participants. ** VAT and SAT tested jointly in a single model. DISCUSSION

In this cohort of parous women with impaired glucose tolerance, elevated fasting glucose, and overweight or obesity, we identified positive (ALT concentration and VAT quantity) and negative (adiponectin concentration) associations of circulating adiposity-related biomarkers and adipose tissue areas with risk of T2D. These associations did not differ significantly among women with and without a history of GDM. Conversely, the association between leptin concentration and T2D risk varied significantly depending on history of GDM. Among parous women with live births who did not have a history of GDM, higher leptin levels at baseline were associated with increased risk of T2D over 3 years of follow-up during the DPP. However, among those with a history of GDM, higher leptin levels were inversely associated with risk of T2D.

ALT concentration and VAT area were positively associated with risk of T2D, as in other populations ((22-25)). Furthermore, these associations did not vary between women with and without a history of GDM, suggesting that mechanisms of adipose tissue dysfunction on T2D risk reflected by these measures do not depend on a history of GDM. However, we did not demonstrate an association of SHBG with risk of T2D in this subgroup of parous women. This is consistent with results in the full DPP cohort ((26)), but this differs from previous studies that have demonstrated an inverse association between SHBG and T2D risk in women ((27)). Differences in the populations studied (e.g., BMI) or adjustment for obesity or body composition might be important in associations of SHBG with incident T2D. In the current study, we also demonstrated a strong and negative association of adiponectin concentration with risk of T2D in parous women, consistent with results in the full DPP cohort ((14)), and no evidence of effect modification by past GDM status on the association of adiponectin with diabetes risk.

Higher circulating leptin levels have previously been associated with increased risk of incident T2D or worsening hyperinsulinemia, independent of BMI or adiposity in large longitudinal cohorts ((28-30)). In the Jackson Heart Study, a longitudinal cohort of Black participants (n = 3,363), higher circulating leptin levels were associated with higher odds of T2D over 7 years, independent of BMI, overall and among male participants only (HR for 1-unit increase in log leptin level in ng/mL overall = 1.3 [95% CI: 1.1-1.6]; in men = 1.3 [95% CI: 1.1-1.7]). The association did not achieve statistical significance among female participants (HR = 1.2 [95% CI: 0.9-1.6]) ((28)). Similarly, in male but not female participants in the Japanese American Community Diabetes Study (JACDS, n = 370), higher circulating leptin levels were associated with higher odds of T2D over 5 to 6 years. The magnitude of the association was similar with and without adjustment for total CT-measured fat (odds ratio [OR] for 1-SD increase in leptin after adjustment for total fat in male participants = 1.8 [95% CI: 1.02-3.17]) ((29)). Finally, in the Ely Study, a longitudinal cohort of White individuals in the UK (n = 748), higher circulating leptin levels were associated with worsening hyperinsulinemia over 10 years in men and women, independent of BMI ((30)). In sum, positive leptin-T2D and leptin-insulin resistance associations were present in several populations (US Black individuals, Japanese American individuals, and UK White individuals) and primarily driven by results in men. The findings in these previous reports, primarily among male individuals, of a direct association of leptin with T2D risk were similar to our finding among women without a history of GDM. In a sensitivity analysis, we did not identify an interaction of race and leptin on T2D risk (p for interaction = 0.767).

On the other hand, findings from subgroup analyses in these and other studies hint at a heterogeneous relationship of leptin with incident diabetes risk. An analysis in the Atherosclerosis Risk in Communities (ARIC) Study reported a positive association of leptin with incident diabetes risk in both men and women (HR in women = 1.3 for tertile 2, 2.5 for tertile 3; p for trend < 0.001) ((31)) that reversed after adjustment for BMI and waist-hip ratio in women (HR in women = 0.7 for tertile 2, 0.5 for tertile 3; p for trend = 0.03). Among female JACDS participants (n = 175, of whom 17 had incident diabetes), although results did not reach statistical significance, the point estimate for the OR for a 1-SD increase in leptin with odds of incident diabetes was 1.1 (95% CI: 0.7-1.8) before adjustment for total fat and 0.7 (95% CI: 0.3-1.5) after adjustment ((29)). These observations indicate a potential role for BMI or for sex-specific risk factors such as menopausal status in the relationship between leptin and risk of T2D; however, in a case-control analysis of n = 2,308 women aged 30 to 55 years from the Nurses’ Health Study, leptin was not associated with risk of incident T2D over 5 years of follow-up in models either with or without adjustment for BMI and menopausal status ((32)). In our data set, we did not see a significant difference in BMI between women with and without a history of GDM (34.1 vs. 34.5). In order to evaluate the possibility of confounding by BMI, we fit models testing the association of leptin and incident T2D without adjustment for BMI and found an HR for T2D of 1.179 (p < 0.001) and 0.880 (p = 0.093) for a 1-SD change in women without and with a history of GDM, respectively, compared with a BMI-adjusted HR of 1.126 (p = 0.016) and 0.776 (p = 0.013), suggesting that confounding by BMI did not explain the difference in associations of leptin with T2D risk that we observed in women without and with a history of GDM. Differences in age across the cohorts are also unlikely to explain differences in the observed association of leptin with T2D risk, as average age was similar among the previous reports (ranging from 52 to 56 years among studies reporting a mean value for age) ((28-30)(32)). Another possibility is menopausal status. In our data set, 57.5% (791/1,376) of women without a history of GDM were postmenopausal compared with 17.9% (60/335) of women with a history of GDM. In a sensitivity analysis adjusting for menopausal status, the HR for leptin was 1.067 (p = 0.151), which was very similar to our overall results (HR 1.064, p = 0.164). Women with and without a history of GDM might also differ by lifestyle factors such as duration of lactation, which is associated with lower risk of T2D over the long term ((33)), although longer duration of lactation was not associated with leptin or adiponectin levels at 3 years post partum after adjustment for factors including BMI and maternal glycemic status during pregnancy ((34)). Data on lactation and other peripartum factors were not available in the DPP.

Leptin, a polypeptide hormone secreted by white adipose tissue in proportion to total fat mass, acts in the brain to reduce food intake and increase energy expenditure ((35)) via its hypothalamic receptor ((36)). In addition to the hypothalamus, the receptor is also expressed in muscle, liver, ovary, and adipose tissue, with pleiotropic metabolic effects ((37)). Among women without a history of GDM, higher circulating leptin levels (as markers of central leptin resistance) ((35)) might explain the positive association we observed between leptin and risk of T2DM. A history of GDM might be hypothesized to affect the relationship as follows. A history of GDM has been associated with lower β-cell function in cross-sectional studies ((38, 39)). Because insulin stimulates leptin production in humans ((40)), when some threshold of β-cell dysfunction is reached, leptin concentration might be lower than would be expected for a given BMI. Among women with a history of GDM in our cohort, lower leptin levels (as markers of β-cell function) might, therefore, reflect more severe β-cell dysfunction and explain the negative association we observed between leptin and risk of T2DM. In our cohort, leptin was positively correlated with the insulinogenic index, a surrogate measure of β-cell function, and inversely with the inverse of fasting insulin, a surrogate measure of insulin sensitivity in both women with and without a history of GDM, although the correlations were somewhat attenuated among women without a history of GDM (Spearman correlation coefficients were 0.25 and −0.40 for women with a history of GDM vs. 0.16 and −0.15 for women without a history of GDM, respectively). This potential mechanism, as well as others, might contribute to the variable associations of leptin with T2D risk in parous women that we observed in the current study, depending on whether or not there was a history of GDM. Topics such as age/menopause-related leptin resistance ((35, 41)) and the long-term consequences of GDM on leptin metabolism require further study.

The present study has several strengths, most importantly, a well-characterized multiethnic cohort at high risk of T2D, including a large sample of women with a history of GDM. Several limitations also deserve consideration. A history of GDM was self-reported, and women may underreport diagnoses of prior GDM ((42)). These issues may result in exposure misclassification. The use of a cohort comprising randomized trial participants may limit external validity ((43)). CT-based adipose areas were available for only a subset of participants (n = 477), which might lead to bias because of differences in baseline characteristics between this subset and the entire DPP cohort or to type II error in these analyses. Because leptin exhibits diurnal variability ((44)), the single fasting time point measured here may not adequately reflect the physiological relationship between serum leptin and T2D. Finally, as described earlier, both age and menopausal status differed between women with and without a history of GDM in our cohort, limiting comparisons.

CONCLUSION

Among women with impaired glucose tolerance in the DPP, the association of circulating leptin levels at baseline with subsequent risk of progression to T2D varied depending on the presence of a history of GDM. Among women without a known history of GDM, higher baseline leptin levels were associated with greater risk of T2D, whereas, conversely, among women with a history of GDM, higher baseline leptin levels were associated with lower risk of T2D. Given the close interrelationships of leptin, adiponectin, and glucose metabolism, studies designed to unravel these mechanisms should include measures of insulin sensitivity and secretion, as well as body composition, in cohorts of age-matched women with and without a history of GDM. Finally, a history of GDM should be considered in future studies of the role of leptin in the development of T2D in women with impaired glucose tolerance.

ACKNOWLEDGMENTS

The Diabetes Prevention Program (DPP) Research Group gratefully acknowledges the commitment and dedication of the participants of the DPP and DPP Outcome Study (DPPOS). During the DPP and DPPOS, the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) of the NIH provided funding to the clinical centers and the coordinating center for the design and conduct of the study and the collection, management, analysis, and interpretation of the data (U01 DK048489). The Southwestern American Indian Centers were supported directly by the