Reproductive
A probabilistic estimate of anovulation using basal body temperature recorded in a mobile phone application (app) Anne Marie Jukic* Anne Marie Jukic D. Robert McConnaughey Carlotta Favaro Eleonora Benhar Agathe van Lemsweerde Jack T Pearson Allen J Wilcox Donna D Baird Anne Marie Z Jukic
Factors that increase anovulation have been difficult to study. We propose a method of ascertaining likely anovulation from basal body temperature (BBT). We use test data from the Natural Cycles mobile app (NC app) to calculate the difference in mean temperature between two portions of the menstrual cycle that are iteratively created by dividing the cycle on each cycle day from day 5 (from cycle start) to -5 (from next cycle start). The maximum temperature difference (MTD) among these differences was retained for each cycle. We hypothesized that a small MTD is correlated with anovulation as defined by a proprietary NC app algorithm. Our test sample of cycles (N=114,004) excluded those of extreme lengths or with a high proportion of missing BBT days, and those impacted by pregnancy, breastfeeding, illness, or hormone use. When we stratified the MTD distribution by the NC app’s definition of anovulatory and ovulatory, there was little overlap between the two (median MTDs of 0.26 °F and 0.63°F, respectively). Using a receiver operator characteristic (ROC) curve, the MTD was highly concordant with the NC app’s measure of anovulation (area under the curve: 0.89). Using the Youden point of the ROC curve, we identified the optimal MTD cut point below which a cycle was likely to be anovulatory. We used generalized linear regression and binomial regression with a compound symmetric covariance structure to estimate associations of MTD and MTD-defined anovulation with older age. Older participants had lower estimated MTDs, and the risk of anovulation was also higher. (Table) These findings replicated in a second sample of NC app cycles. The MTD is the first estimate of anovulation based solely on daily BBT. It is predictive of NC app-defined anovulation and shows the expected association with older age. The MTD could be useful across many digital platforms to understand factors associated with anovulation.
