Women with polycystic ovary syndrome (PCOS) are insulin resistant, have insulin secretory defects, and are at high risk for glucose intolerance. We performed this study to determine the prevalence of glucose intolerance and parameters associated with risk for this in PCOS women. Two-hundred and fifty-four PCOS women, aged 14-44 yr, were prospectively evaluated at 2 centers, 1 urban and ethnically diverse (n = 110) and 1 rural and ethnically homogeneous (n = 144). The rural PCOS women were compared to 80 control women of similar weight, ethnicity, and age. A 75-g oral glucose challenge was administered after a 3-day 300-g carbohydrate diet and an overnight fast with 0 and 2 h blood samples for glucose levels. Diabetes was categorized according to WHO criteria. The prevalence of glucose intolerance was 31.1% impaired glucose intolerance (IGT) and 7.5% diabetes. In nonobese PCOS women (body mass index, <27 kg/m2), 10.3% IGT and 1.5% diabetes were found. The prevalence of glucose intolerance was significantly higher in PCOS vs. control women (χ2 = 7.0; P = 0.01; odds ratio = 2.76; 95% confidence interval = 1.23-6.57). Variables most associated with postchallenge glucose levels were fasting glucose levels (P < 0.0001), PCOS status (P = 0.002), waist/hip ratio (P = 0.01), and body mass index (P = 0.021). The American Diabetes Association criteria applied to fasting glucose significantly underdiagnosed diabetes compared to the WHO criteria (3.2% vs. 7.5%; χ2 = 4.7; P = 0.046; odds ratio = 2.48; 95% confidence interval = 1.01-6.69). We conclude that 1) PCOS women are at significantly increased risk for IGT and type 2 diabetes mellitus at all weights and at a young age; 2) these prevalence rates are similar in 2 different populations of PCOS women, suggesting that PCOS may be a more important risk factor than ethnicity or race for glucose intolerance' in young women; and 3) the American Diabetes Association diabetes diagnostic criteria failed to detect a significant number of PCOS women with diabetes by postchallenge glucose values.