Validity of methods for analyzing urinary steroid data to detect ovulation in athletes

Purpose: The purpose of this study was to determine whether the accuracy of ovulation detection algorithms is compromised when applied to menstrual cycles exhibiting subclinical hormonal abnormalities, which are particularly prevalent in female athletes. Methods: The validity of five ovulation detection algorithms was compared between 25 regularly exercising women and 15 sedentary controls. Subjects collected daily urine samples for an entire menstrual cycle for analysis of estrone-3-glucuronide (EIG), pregnanediol-3-glucuronide (PDG), and luteinizing hormone (LH). The algorithms were applied to determine their sensitivity (% of true ovulatory cycles), specificity (% of true anovulatory cycles), and the deviation from the reference day of ovulation (difference scores). Results: The sensitivity was > 80% in all algorithms except Baird's EIG/PDG ratio algorithm (74%) and Kassam's PDG ratio algorithm (78%). All algorithms, except Kassam's PDG ratio algorithm (80%), were found to exhibit specificities < 70%. Baird's EIG/PDG ratio algorithm was the most accurate in estimating the day of ovulation by deviating only -0.2 ± 0.3 d from the reference day in the exercising female cycles and -0.5 ± 0.3 d in the controls. No statistical differences in the sensitivities of the algorithms were found between the exercising and control cycles. When comparing the deviation from the reference day of ovulation between subject groups, no statistical difference was found. Conclusion: The algorithms display similar validity in determining the presence and day of ovulation between subject groups, and thus may be applied to cycles exhibiting subclinical hormonal abnormalities as commonly observed in exercising women.