Early Phenotypes in Girls at Risk for PCOS Replicate Metabolic and Reproductive Subtypes: An Unsupervised Clustering Analysis

Both daughters of women with PCOS (PCOS-d) and overweight girls (OW-g) are proposed to be at increased risk of PCOS because they have peripubertal increases in testosterone (T) levels, a cardinal feature of PCOS. We are testing this hypothesis by performing longitudinal studies in these girls after menarche. In adult women with PCOS, we have recently identified reproductive and metabolic subtypes using unsupervised cluster analyses. These subtypes were associated with novel PCOS susceptibility genetic loci suggesting that the subtypes reflect biologically discrete entities. We performed similar analyses in our cohort of early postmenarchal PCOS-d and OW-g to test the hypothesis that these subtypes are present in girls at risk for PCOS. Fifteen PCOS-d and 10 OW-g aged 11-16 years and with postmenarchal age less than 2 years were studied. Mothers of PCOS-d fulfilled NIH criteria for PCOS, mothers of OW-g were reproductively normal with no history of irregular menses or clinical hyperandrogenism. OW-g had a BMI above the 85th percentile for age. There was no BMI inclusion criterion for PCOS-d; four PCOS-d had a BMI above the 85th percentile. The girls were of comparable age, post-menarchal age and BMI z score. A fasting morning blood sample was drawn for T, SHBG, DHEAS, glucose and insulin. Leuprolide 10 mcg/kg SC was administered. LH and FSH levels were measured at baseline, 30 min, and 60 min following leuprolide. Unsupervised hierarchical cluster analysis adjusted for age was performed on quantitative traits including BMI, T, fasting insulin, fasting glucose, DHEAS, SHBG, and LH and FSH. These are the same quantitative traits used for clustering in adult PCOS. The clustering revealed 2 distinct PCOS subtypes: a reproductive group (41%), characterized by higher SHBG levels, LH and FSH with relatively low BMI and insulin levels, and a metabolic group (41%), characterized by higher BMI and insulin levels and lower SHBG, LH, and FSH. Jaccard coefficients indicated cluster stability (0.70 reproductive, 0.69 metabolic). There was a significant difference in the distribution of the two subgroups in PCOS-d and OW-g: PCOS-d 60% reproductive, 13% metabolic, 27% indeterminate; OW-g 25% reproductive, 50% metabolic, 25% indeterminate (Chi Sq P=0.05). We found that early postmenarchal PCOS-d and OW-g demonstrate reproductive and metabolic subtypes similar to those identified in adult women with PCOS. The majority of PCOS-d had the reproductive subtype. These findings suggest that this subtype, which is characterized by disordered gonadotropin secretion, is an early harbinger of PCOS. Longitudinal studies are ongoing to test this hypothesis.