Objective: To determine optimal criteria for polycystic ovary morphology. Design: Cross-sectional study. Setting: Multidisciplinary polycystic ovary syndrome (PCOS) clinic at a tertiary academic center. Patient(s): Subjects with PCOS were seen between 2006 and 2015 and met PCOS 1992 National Institutes of Health criteria. Controls were from the Ovarian Aging (OVA) study (2006-2011), a longitudinal study including healthy women with regular menstrual cycles. Intervention(s): Clinical data collection. Main Outcome Measure(s): Follicle number per ovary (FNPO) and ovarian volume (OV). Result(s): A total of 245 subjects with PCOS and 756 OVA study subjects were included in the FNPO analysis and had a mean (+/- SD) FNPO of 22.6 +/- 12.4 and 10 +/- 5.3, respectively. Receiver operating characteristic curves were created for both FNPO and OV and analyzed across age group categories (25 to < 30, 30 to < 35, and 35 to < 40 years). Youden's and minimum distance (d) were used to compare efficacies of FNPO and OV thresholds. The optimal threshold for distinguishing PCOS from OVA controls was FNPO > 13. There was a decreasing trend in FNPO threshold with increasing age group (> 15, > 14, and > 12, respectively). A total of 297 PCOS subjects and 756 OVA study subjects were included in the OV analysis and had a mean maximum OV of 10 +/- 5 cm(3) and 6.5 +/- 5 cm(3), respectively. The overall threshold was OV > 6.75 cm(3), with a trend toward decreasing OV with increase in age group (> 8.5, > 7, and > 6.25 cm(3), respectively). Conclusion(s): Our findings reflect that age-stratified thresholds demonstrate superior diagnostic performance, with an improved balance of sensitivity and specificity compared with a single threshold. We propose age-specific thresholds for better diagnostic performance. (C) 2018 by American Society for Reproductive Medicine.