Commentary on Joyce et al.: Studying menstrual cycle effects on behavior requires within-person designs and attention to individual differences in hormone sensitivity

Abstract

The science of menstrual cycle effects on behavior can be accelerated by addressing two long-standing issues. First, models predicting uniform cycle effects must be updated to reflect evidence for marked individual differences in behavioral hormone sensitivity. Secondly, longitudinal approaches must be embraced, as cross-sectional measures of cyclical change show poor prospective validity.

Studies of menstrual cycle changes in behavior hold promise for developing a mechanistic understanding of sex differences. However, progress in this field remains disjointed and slow, with limited translation of scientific findings into sex-specific treatments for behavioral disorders. Joyce et al. [1] provide an excellent demonstration of two recurrent, related barriers to scientific progress in this area. First, scientists often erroneously predict uniform cycle effects, whereas the preponderance of data instead support an individual differences model in which cyclical behavior change varies between people according to their degree of hormone sensitivity. Secondly, it has become clear that scientists must measure these individual differences in cyclical change prospectively (i.e. using repeated measures throughout the cycle) and not retrospectively (i.e. using single time-point interviews or surveys), as the two generally do not converge. These challenges increase the complexity, cost, and burden of conducting rigorous studies of behavior throughout the menstrual cycle. However, directly engaging these challenges will allow for the accumulation of more actionable knowledge regarding sex-specific mechanisms in behavioral disorders.

Steroid changes (e.g. throughout the menstrual cycle) do not influence behavior uniformly among individuals; rather, these effects are shaped by marked between-person differences in neurobehavioral hormone sensitivity. In longitudinal studies most cycling individuals do not show recurrent changes in mood, cognition, or behavior throughout the cycle [2,3], whereas a minority show changes ranging from mild to severe [4,5]. These cyclical changes in symptoms are not caused byendocrine or gyne-cological pathology; instead, experimental studies implicate an abnormal brain sensitivity to normal hormone changes [6]. Joyce et al. demonstrate that this differential hormone sensitivity may also shape substance use and misuse throughout the cycle. Because focusing upon global effects of the cycle fundamentally compromises the validity of conclusions, these findings highlight the need to measure and model these between-person differences in hormone sensitivity.

However, studying these between-person differences in susceptibility to cyclical symptom change remains challenging, due to the perennial lack of correspondence between retrospective and prospective measures of cyclical symptoms. Retrospective (single time-point) measures of cyclical change continue to tempt scientists with their low burden and strong face validity; however, they have repeatedly failed to demonstrate convergent validity with actual cyclical changes [7–10]. Specifically, retrospective measures suffer from low specificity (i.e. reports of cyclicity when none exists in daily ratings; false positives) in studies of premenstrual dysphoric disorder (PMDD). They may also suffer from in adequate sensitivity (i.e. reports of no cyclicity when it exists in daily ratings; false negatives) in chronically distressed groups (e.g. borderline personality disorder [11]). Joyce et al. conceptually replicate these findings by demonstrating poor agreement between the Structured Clinical Interview for DSM-5 Disorders (SCID-5) interview-based provisional diagnosis of PMDD and prospective patterns of cyclical depression change. Although they may have observed greater or lesser concordance if other core emotional symptoms of premenstrual disorders were considered (e.g. irritability, mood swings, anxiety), their results cast doubt on the validity of the SCID-5 interview for PMDD. These historical validity problems among retrospective measures are well known among clinical assessment experts in PMDD, as reflected in the unprecedented decision of the DSM-5 PMDD working group to require prospective daily ratings throughout two menstrual cycles for an official diagnosis [12]. In sum, retrospective methods lack sufficient validity for measuring behavior change across the menstrual cycle and should be used only in screening protocols.

While prospective, within-person studies of behavior across the menstrual cycle can be complex, costly and time-consuming, they remain essential for identifying who is at risk for cyclical changes and for clarifying the pathophysiology of this hormone-sensitive response. For interested readers, recent reviews provide helpful overviews of the field of cyclical hormone sensitivity [13–15] and recommendations with practical tools for conducting within-person menstrual cycle studies [16,17]. Further, automated algorithms exist for standardized quantification of cyclicity in daily ratings [9,18,19]. Wider adoption of these rigorous methods will clarify hormonal mechanisms among various behavioral disorders and may finally yield the insights necessary for the targeted development of sex-specific treatments.