Persistent Apparent Pancreatic β-Cell Defects in Premenarchal PCOS Relatives

Abstract

Context:

Polycystic ovary syndrome confers an increased risk for type 2 diabetes in affected women as early as adolescence. First-degree relatives (FDRs) of affected women are at increased risk for associated reproductive and metabolic phenotypes.

Objective:

We sought to prospectively assess insulin sensitivity and secretion and to measure reproductive hormone levels using sensitive techniques.

Design, Setting, and Participants:

Twelve premenarchal FDR girls and 10 control girls of comparable age, Tanner stage, and body mass index were studied at an academic medical center.

Interventions:

Frequently sampled intravenous glucose tolerance tests and oral glucose tolerance tests were performed.

Main Outcome Measures:

Reproductive hormone levels, lipid profiles, glucose tolerance, and frequently sampled iv glucose tolerance test parameters of insulin sensitivity and secretion were investigated.

Results:

Disposition index (DI), insulin secretion corrected for insulin sensitivity, was decreased in FDR compared with control girls at baseline (P = .01), independent of dysglycemia. Decreases in DI persisted in FDR girls during the 2-year follow-up (P = .003). T levels were increased (P = .02) in FDR compared with control girls at baseline, but this difference did not persist because T levels increased in control girls.

Conclusions:

DI is decreased in peripubertal FDR girls, and this decrease persists as puberty progresses. These findings suggest that β-cell dysfunction is an early defect in glucose homeostasis preceding decompensation in glucose tolerance in FDR girls. T levels were increased in FDR girls earlier than previously reported, but these changes did not persist, suggesting an earlier onset of pubertal increases in glandular androgen secretion in FDR girls.

Polycystic ovary syndrome (PCOS) is a common disorder with the classic syndrome of hyperandrogenism and chronic anovulation affecting approximately 7% of reproductive-age women (1). Women with PCOS frequently have metabolic abnormalities including insulin resistance, pancreatic β-cell dysfunction, dyslipidemia, and obesity (1). Affected women have markedly increased prevalence rates of impaired glucose tolerance (IGT) and type 2 diabetes, making PCOS a leading risk factor for these disorders in adolescent and young adult women (1).

The etiology of PCOS remains unknown. Elevated insulin levels secondary to insulin resistance function as a cogonadotropin to increase ovarian androgen production (1). Hyperinsulinemia also decreases hepatic SHBG synthesis, resulting in increased bioavailable T levels (1). Conversely, androgens contribute to insulin resistance and visceral adiposity in affected women (1). PCOS may also have developmental origins because phenocopies can be created in several animal species by androgen administration prenatally (1, 2). Finally, PCOS is highly heritable with approximately 70% concordance in monozygotic twins, suggesting a genetic contribution to its pathogenesis (1). Indeed, several replicated genetic susceptibility loci have been mapped for the disorder (3, 4).

Male as well as female first-degree relatives (FDRs) have reproductive and metabolic abnormalities consistent with a genetic contribution to these phenotypes (1, 5–10). Accordingly, investigation of these phenotypes in FDR children has provided insight into the ontogeny of the syndrome. During infancy and persisting into childhood, FDR girls have increased antimullerian hormone (AMH) levels (9), suggesting altered ovarian follicular development. Dehydroepiandrosterone sulfate (DHEAS) levels are elevated in FDR compared with control girls, suggesting exaggerated adrenarche (5). Elevations in circulating T levels have not been noted until later pubertal Tanner stages IV and V in FDR girls (10). However, these studies were limited by the use of less sensitive and precise RIA measurements of T (11).

Glucose-stimulated hyperinsulinemia develops as early as age four years and persists throughout puberty in FDR girls (6). One study (7) reported evidence for insulin resistance and β-cell dysfunction in premenarchal FDR girls aged 8–14 years using fasting proxies for insulin resistance and insulin responses during an abbreviated iv glucose tolerance test to assess β-cell function (7). However, these imprecise measures of insulin sensitivity and secretion (6, 7) are confounded by other factors including differences in pancreatic β-cell function, insulin clearance, and glucose absorption (12).

We undertook this study to investigate the mechanisms of abnormalities in glucose homeostasis by prospectively assessing insulin sensitivity and secretion in peripubertal FDR girls over a 2-year period. To gain insight into the ontogeny of metabolic and reproductive phenotypes, we also assessed reproductive hormone levels in these girls using a highly sensitive T assay technique (13).

Materials and Methods

Premenarchal FDR girls (n = 12) of women with PCOS and control girls (n = 10) of comparable age, body mass index (BMI), visceral adipose tissue (VAT), and breast Tanner stage were studied. Only overweight or obese subjects were studied due to provisions in the Code of Federal Regulations (CFR) that forbid research procedures presenting more than minimal risk to healthy child subjects (45CFR46.406/21CFR50.53) (14).

Subjects were recruited by contacting women who have previously participated in our studies of PCOS and control adult women as well as by advertisements in local media and on-line. All girls were in good health and not taking any medications known to alter reproductive hormone metabolism or glucose homeostasis for at least 1 month prior to study. The inclusion criteria were ages 8–12 years, Tanner Stage I-III breast development, and 85th to 98th BMI percentile. FDR girls had a mother or sister who fulfilled National Institutes of Health criteria for PCOS [hyperandrogenism and oligoanovulation with exclusion of other reproductive disorders (1)] as confirmed by us prior to the current study or by their personal physician. Mothers of control girls had regular menses every 27–35 days as well as no history of reproductive disorders and no signs or symptoms of androgen excess by validated questionnaire (15). The Institutional Review Boards of the Feinberg School of Medicine, Northwestern University, and Children's Memorial Hospital approved this study. Written informed consent was obtained from the parent of all girls, and written assent was obtained from each girl prior to participation.

Twenty-five potential control girls and 14 potential FDR girls were screened (visit 1). Studies were performed in the postabsorptive state after a 10-hour overnight fast. At visit 1, after an iv catheter for blood sampling had been inserted and a 30-minute rest period, baseline blood samples were obtained for reproductive hormone and lipid levels followed by an oral glucose tolerance test (OGTT) with a glucose dose of 1.75 g/kg body weight, up to a maximum of 75 g, with blood sampling at 0, 30, 60, 90, and 120 minutes. We have validated the assessment of circulating total and bioavailable T levels in a single early morning blood sample in adult women (15). A single early morning blood sample has also been used to detect elevated total and/or free T levels in obese compared with the lean girls across puberty (16).

Control girls with dysglycemia [fasting glucose ≥ 100 mg/dL and/or 2 h postglucose challenge glucose ≥ 140 mg/dL (17)] were excluded from further study. Because IGT is extremely common in adult women with PCOS (1), we hypothesized that this finding may be part of the at risk phenotype identified in FDR girls. Accordingly, FDR girls with fasting glucose 100–125 mg/dL or 2-hour postglucose challenge glucose 140–199 mg/dL were included in the study.

Twelve potential control girls failed screening for the following reasons: 11 did not meet the BMI criterion and one had dysglycemia detected during the OGTT. Three additional control girls were disqualified after enrollment because they were unable to tolerate the study procedures. Two potential FDR girls failed screening for the following reasons: one did not meet the BMI criterion and one did not meet the breast Tanner stage criterion. Twelve FDR (11 daughters, one sister) and 10 control girls qualified for further study.

Within 10 days of visit 1, a frequently sampled iv glucose tolerance test (FSIGT) was performed after a 10-hour overnight fast in 11 FDR and nine control girls; one FDR and one control girl did not complete the FSIGT for technical reasons. Glucose (0.3 g/kg) was administered iv at time 0 and 0.03 U/kg regular insulin (Humulin R; Eli Lilly) was administered iv at time 20 minutes (18). Frequent blood samples were obtained serially for 15 minutes before and 180 minutes after the iv glucose (18). VAT volume was measured by magnetic resonance imaging (MRI; 1.5 Telsa Sonata MRI; Siemens Medical Solutions) (19). Visit 1 and 2 studies were repeated in seven FDR and eight control girls 1 year later and in six FDR and eight control girls 2 years later (Figure 1).

Figure 1.

Study time line. Details of longitudinal study visits are depicted.

Assays

Glucose, insulin, androstenedione (A), ultrasensitive estradiol (E2), SHBG, DHEAS, LH, FSH, AMH, total cholesterol, high-density lipoprotein (HDL), and triglyceride levels were measured as reported (18, 20, 21). Low-density lipoprotein (LDL) was calculated using the Friedewald equation. T levels were analyzed by liquid chromatography tandem mass spectrometry (LC-MS/MS) (sensitivity 0.07 nmol/L; intraassay coefficient of variation 9.0% at 0.56 nmol/L; interassay coefficient of variation 15.8% at 0.42 nmol/L) (10). Bioavailable T was calculated as reported (22).

Calculations

Insulin sensitivity [sensitivity index (SI)] and secretion [acute insulin response to glucose (AIRg)] were determined by minimal model analysis of FSIGT glucose and insulin values using MINMOD Millennium 5.7.7 (MINMOD Inc) (23). The disposition index (DI) was calculated as the product of SI × AIRg (23). Glucose clearance (KG), a parameter of glucose tolerance, was calculated as the slope of the least square regression line of the natural log of the glucose level vs time from 10 to 19 minutes after the glucose infusion during the FSIGT (18). MRI scans were analyzed using Slice-O-Matic (version 4.3b; Tomovision) image analysis software (24) in a blinded random fashion by the same observer on two separate occasions; the average is reported as VAT volume.

Statistical analysis

Due to our small sample size, we did not assume Gaussian distribution of our data and used nonparametric Wilcoxon sign rank tests for two-group comparisons of baseline data. Categorical variables were compared by Fisher's exact test. Baseline control DI data were fit to a line and individual FDR girl DI data plotted. Repeated measures of ANOVA using a generalized estimating equation model in SAS version 9.2 (SAS Institute, Inc) was applied to assess the longitudinal data. This analysis assumes a Gaussian distribution of the data. Accordingly, data were log transformed when necessary to achieve homogeneity of variance. Transformation did not result in a Gaussian distribution for two end points (total cholesterol, LH). These data were analyzed by three nonparametric two-group comparisons (baseline vs year 1; baseline vs year 2; year 1 vs year 2) with the level of α adjusted with a Bonferroni correction to .017 (.05/3) for multiple comparisons. Spearman's correlation coefficients were determined to assess association between independent-variables BMI, breast Tanner stage or VAT and dependent variables DI or T. Data are reported as the untransformed median (25th to 75th interquartile range) with the level of α set at .05, except as noted.

Results

The racial/ethnic composition of the groups was as follows: non-Hispanic white, 10 PCOS, four control; black, one PCOS, four control; Hispanic, one PCOS, 0 control; mixed ethnicity (black/white), 0 PCOS, one control; and Asian, 0 PCOS, one control (P = .07). There were no significant differences in age, BMI z score, VAT volume, or breast or pubic hair Tanner stage between the groups at baseline (Table 1) or follow-up (data not shown). Eight of 12 FDR and 6 of 10 control girls had a family history of type 2 diabetes (P = .99). One of 12 FDR mothers and none of 10 control mothers had a history of gestational diabetes. All girls were premenarchal at baseline with a median breast Tanner stage II for the FDR girls and Tanner stage I for control girls (P = .31). At the study's conclusion, two FDR and two control girls were postmenarchal, and the median breast Tanner stage had advanced to stage V in both groups.

Table 1.

Baseline Clinical Characteristics and Reproductive Hormones

	FDR (n = 12)	Control (n = 10)	P Value
Age, y	10.4 (8.8–12.1)	9.7 (9.0–10.1)	.34
BMI z score	1.7 (1.5–1.8)	1.9 (1.4–2.0)	.31
VAT, L	0.45 (0.37–0.98)	0.55 (0.42–0.61)	.86
Breast Tanner stage	I, 25%, II, 58%, III, 17%	I, 60%, II, 30%, III, 10%	.31a
Pubic hair Tanner stage	I, 55%, II, 27%, III, 9%, IV, 9%	I, 50%, II, 50%	.72a
SHBG, nmol/L	37 (23–48)	46 (35–61)	.31
DHEAS, ng/mL	58 (29–78)	59 (24–72)	.97
A, ng/dL	54 (44–109)	47 (37–56)	.37
LH, IU/L	0.1 (0.1–2.0)	0.1 (0.1–0.4)	.99
FSH, IU/L	2.3 (1.1–4.3)	2.1 (1.2–5.1)	.77
E2, pg/mLb	12 (9–23)	11 (7–15)	.36
AMH, ng/mLb	2.6 (1.6–4.3)	2.0 (1.1–2.5)	.15

Data are presented as median (25th to 75th interquartile range) with the exception of the categorical variables of breast and pubic hair Tanner stage, in which the percentage of subjects included in each group are noted. P values listed from Wilcoxon sign rank tests unless noted otherwise.

^aCategorical variables breast and pubic hair Tanner stage were analyzed by Fisher's exact test.

^bDue to insufficient sample volume, E2 and AMH assays were performed in 11 FDR and eight control girls. To convert DHEAS from nanograms per milliliter to nanomoles per liter, multiply by 2.714; A from nanograms per deciliter to nanomoles per liter, multiply by 0.0349; E2 from picograms per milliliter to picomoles per liter, multiply by 3.671; AMH from nanograms per milliliter to picomoles per liter, multiply by 7.1429.

At baseline, FDR girls had significantly higher total (P = .02) and bioavailable T (P = .007) levels. The difference in T levels did not persist over the subsequent years as T concentrations in control girls increased to a greater degree than those in FDR girls (Figure 2). Levels of the other reproductive hormones did not differ significantly at baseline (Table 1) or follow-up. There was a trend toward increased AMH levels in FDR girls at baseline (Table 1) and longitudinally (P = .10, repeated measures of ANOVA). There was no significant correlation between T and BMI z score or T and breast Tanner stage at baseline or in the follow-up years 1 or 2 (data not shown).

Figure 2.

T levels and FSIGT-derived SI and secretion over 2 years of follow-up. The increased T (upper left panel) and Bioavailable T (upper right panel) at baseline in the FDR girls (*, P = .02 T and **, P = .007 uT, Wilcoxon sign rank tests) did not persist as the control girls had greater increases in T (P = .48 T, P = .54 uT, repeated measures of ANOVA). There was no significant difference in SI between the 2 groups (P = .20, repeated measures of ANOVA, lower left panel). DI was persistently decreased in FDR compared with control girls at baseline (***, P = .01, Wilcoxon sign rank tests, lower right panel) and over the 2-year longitudinal study (†, P = .003, repeated measures of ANOVA). The untransformed mean ± SE are graphed. ●, FDR girls; ○, control girls.

Fasting glucose levels did not differ significantly at baseline (Table 2) or longitudinally. Two-hour postchallenge glucose levels were significantly increased at baseline in FDR girls (P = .04, Table 2). Three FDR girls had IGT at baseline and one also had impaired fasting glucose (17). There was no significant difference in baseline 2-hour postchallenge glucose levels after exclusion of the dysglycemic FDR girls (P = .20). Two of the dysglycemic FDR girls did not return for follow-up. IGT normalized by year 2 in the FDR girl with IGT who returned for follow-up. No control girls developed dysglycemia over the study period. Therefore, 2-hour postchallenge glucose levels no longer differed at the follow-up visits. Fasting and 2-hour postchallenge insulin levels did not differ significantly between the groups at baseline (Table 2) or longitudinally. Lipid levels did not differ at baseline (Table 2) or longitudinally.

Table 2.

Baseline Metabolic Data

	FDR (n = 12)	Control (n = 10)	P Value
Fasting glucose, mg/dL	92 (89–95)	90 (88–90)	.46
Two-hour postchallenge glucose, mg/dL	121 (108–136)	108 (96–120)	.04
Fasting insulin, μIU/mL	18 (15–26)	17 (12–20)	.53
Two-hour postchallenge insulin, μIU/mL	78 (53–167)	77 (67–121)	.69
Total cholesterol, mg/dL	146 (128–152)	148 (134–154)	.51
LDL, mg/dL	88 (69–98)	87 (82–97)	.87
HDL, mg/dL	41 (38–45)	41 (32–49)	.87
Triglycerides, mg/dL	75 (65–98)	48 (37–86)	.11

Data are presented as median (25th to 75th interquartile range). P values are listed from Wilcoxon sign rank tests. Bold indicates differences in end points reaching statistical significance. To convert glucose values from milligrams per deciliter to millimoles per liter, multiply by 0.0556; insulin values from microinternational units per milliliter to picomoles per liter, multiply by 6.945; total cholesterol, LDL, HDL, and triglyceride values from milligrams per deciliter to millimoles per liter, multiply by 0.02.

KG was lower at baseline (P = .02) and longitudinally (P = .01, repeated measures of ANOVA) in FDR compared with control girls. After exclusion of the three dysglycemic FDR girls, there was still a nonsignificant trend toward decreased KG in FDR girls at baseline (P = .06) and during longitudinal follow-up (P = .06, repeated measures of ANOVA). There was a nonsignificant (P = .11) trend toward decreased SI in FDR girls at baseline and longitudinally (P = .20, Figure 2). AIRg did not differ significantly between the groups at baseline or longitudinally.

There was a 46% decrease in DI (P = .01) in FDR compared with control girls at baseline. When the three FDR girls with dysglycemia were excluded, the difference in DI was still significant (P = .03). Similarly, the difference in DI remained significant when the 1 FDR sister was removed from the analysis (P = .01). Nine of 11 (80%) FDR girls' DI values were below the line fit of the control girls, suggesting that those FDR girls were at high risk for type 2 diabetes (Figure 3) (25). Decreased DI persisted in the FDR compared with the control girls during the longitudinal study (P = .003, repeated measures of ANOVA; Figure 2). There was no correlation between DI and BMI z score, breast Tanner stage, or VAT at baseline or in follow-up years 1 or 2 (data not shown). Some studies have suggested that blacks have higher DI values than whites because of differences in insulin secretion and clearance (26, 27). However, DI did not differ significantly in the black control girls compared with control girls of other races (P = .49).

Figure 3.

Individual DIs for FDR girls at baseline. Hyperbolic line fit for the control girl data, the FDR girl data are plotted individually (○). FDR girls with dysglycemia are denoted by black circles (●). Most FDR girls had a DI below the control population average, indicating that they are at high risk to progress to type 2 diabetes (25).

Discussion

DI was significantly decreased at baseline, and this decrease persisted over 2 years of prospective follow-up in peripubertal FDR compared with control girls of comparable age, pubertal status, BMI, and VAT. Insulin sensitivity did not differ significantly in the 2 groups. These findings suggest that the first change in glucose homeostasis in the FDR girls studied was a defect in pancreatic β-cell function that was evident in early puberty. Furthermore, the difference in DI preceded decompensation in glucose tolerance because it was present in FDR girls without dysglycemia. In addition, DI remained significantly decreased when the analysis was limited to daughters to ensure consistent exposure to a PCOS prenatal environment. Because DI is the most powerful predictor of type 2 diabetes risk (25), FDR girls are already at high risk in early puberty.

Total and bioavailable T levels were increased in our FDR compared with control girls during early puberty, in contrast to earlier reports in Chilean FDR girls (10), suggesting that increased glandular steroidogenesis was also present in early puberty in FDR girls. However, the increased T levels did not persist as puberty progressed, suggesting that there is an earlier increase in glandular androgen secretion during puberty in FDR compared with control girls, rather than an ongoing increase in androgen production during puberty. DHEAS and AMH levels were not increased in our FDR girls, in contrast to findings in Chilean FDR girls (5, 9). Consistent with findings in the Chilean FDR girls (10), there were no changes in other circulating androgen levels or in gonadotropin levels. The glandular source(s) of increased T production in the present study is unknown because neither DHEAS nor gonadotropin levels were increased.

In the setting of normal β-cell function, as insulin sensitivity decreases, insulin secretion increases in a compensatory manner to maintain a constant hyperbolic relationship (28). This relationship, the product of insulin secretion and insulin sensitivity, is known as DI. Individuals with DI below the 50th percentile in the normal population are at increased risk of type 2 diabetes (25). Eighty percent of the FDR girls had a DI below the control population average, indicating that they were at high risk to progress to type 2 diabetes (Figure 3) (25). This finding is of particular concern, given the young age of our FDR girls. Furthermore, β-cell dysfunction occurred in the absence of dysglycemia in most of the FDR girls as it does in adult women with PCOS (1). Accordingly, screening for dysglycemia would not detect most high-risk FDR girls. However, there was evidence for decreased glucose tolerance in FDR girls because there was a trend toward decreased KG (29, 30), even after the exclusion of the dysglycemic FDRs. There was no deterioration in glucose tolerance, but the duration of follow-up was brief (31) and two dysglycemic FDRs did not return for further study.

This metabolic phenotype could reflect the impact of PCOS susceptibility genes or programming effects of the intrauterine environment. In adult PCOS women, the type 2 diabetes TCF7L2 susceptibility locus was associated with evidence of β-cell dysfunction (32). This finding suggests that genetic factors could contribute to the β-cell dysfunction in FDR girls. In addition, in mouse models, epigenetic modifications in genes regulating pancreatic development, including PDX1, have been implicated in the development of β-cell dysfunction in the setting of intrauterine growth restriction (33). Similar epigenetic modifications could contribute to the development of β-cell dysfunction in FDR girls.

We identified decreased DI in FDR girls at a younger age than most other studies in populations at high risk for type 2 diabetes (34–36). Indeed, there have been few studies accurately assessing DI in children due to concerns about the volume of blood sampling and subject compliance during prolonged testing protocols (14). However, in a study from The Netherlands (37), DI was decreased by 20% in former full-term small-for-gestational-age infants with a mean age of 6.8 years compared with former small-for-gestational-age infants born prematurely. Decreased DI has also been reported in adolescent cohorts older than our subjects. A similar approximately 50% reduction in DI to our FDR girls was reported in both adolescents with a family history of type 2 diabetes compared with those without a family history (34), and in adolescent girls with PCOS and IGT compared with PCOS girls with normal glucose tolerance (35). The prevalence of a family history of type 2 diabetes was comparable in our FDR and control girls, suggesting that factors other than a family history of type 2 diabetes contributed to the decreased DI in FDR girls.

Most previous investigations of metabolic phenotypes in FDR girls have used imprecise assessments of glucose homeostasis (6, 7, 38). Fasting parameters of insulin sensitivity, such as homeostasis model assessment index of insulin resistance and quantitative insulin sensitivity check index (7), as well as glucose-stimulated hyperinsulinemia (6) and abbreviated iv glucose tolerance tests (7) are confounded by differences in pancreatic β-cell function, insulin clearance, and glucose absorption (12). The one study (8) that did use FSIGTs with minimal model analysis found decreased SI and DI as well as decreased suppression of free fatty acids by insulin in FDR girls. However, the FDR girls were significantly heavier than the control girls, so the findings were confounded by the independent impact of adiposity on insulin sensitivity and secretion (23). Furthermore, the girls were later in puberty than those in the present study because 44% of the FDR and 20% of the control girls were already postmenarchal (8), whereas all of our girls were premenarchal at baseline.

The other novel finding of our study was the observation that T levels were increased in FDR girls during the early peripubertal years, Tanner stages I-III. Sir-Petermann et al (10) also found elevated T levels in FDR girls, but only during late puberty, in Tanner stages IV-V. These discrepant findings may be due to ethnic differences in the FDR populations because the subjects in the study by Sir-Petermann et al were Chilean. However, it is also possible that differences in the T assay technique accounted for the different observations in T levels. Sir-Petermann et al (10) measured T by RIA that was validated in a subset of samples using LC-MS/MS. All T levels were determined by LC-MS/MS in our study, which has greater sensitivity compared with RIA for the measurement of T levels in the normal female and pediatric range (13).

An important strength of our study was the use of FSIGT with minimal model analysis to precisely quantify insulin sensitivity and secretion. Despite the small sample size, we identified a significant difference in DI, the primary study end point. Furthermore, the changes in DI were corroborated by their persistence in the longitudinal study. A limitation of our study was the relatively small sample size. Accordingly, we may have lacked adequate statistical power to detect differences in FDR girls in the secondary metabolic and reproductive end points, such as AMH.

The racial/ethnic heterogeneity of the control population was another potential limitation of our study. The control group had a higher prevalence of black girls than the FDR group, although this difference was not statistically significant. Black youth are more hyperinsulinemic than white youth due to decreased insulin sensitivity (39) and clearance coupled with increased insulin secretion (26, 27). These changes result in increased DI values in black youth (26, 27). However, DI did not differ significantly in the black control girls compared with control girls of other races in the present study. This finding suggests that racial/ethnic differences in the control girls did not confound our finding of decreased DI in FDR girls.

Finally, differences in pubertal stage were a potential constraint. Breast Tanner stage did not differ significantly between the groups, but there was a higher percentage of breast Tanner stage I girls in the control group. Nevertheless, DI did not correlate with breast Tanner stage. Furthermore, DI remained significantly decreased in FDR compared with control girls over 2 years of prospective follow-up as puberty advanced in both groups. These findings suggest that differences in pubertal stage between FDR and control girls did not contribute to differences in DI.

In conclusion, peripubertal FDR girls have apparent persistent pancreatic β-cell dysfunction beginning in early puberty without significant alterations in insulin sensitivity. This phenotype in FDR girls could reflect the impact of PCOS susceptibility genes or programming effects of the intrauterine environment. We speculate that therapies targeting insulin secretion rather than insulin action may be more successful for diabetes prevention in girls at risk for PCOS (40). Further prospective studies in larger cohorts are necessary to validate our findings and to understand the progression of metabolic and reproductive phenotypes in FDRs.