Using Natural Spline Models to Explore the Trajectories of Empirically Derived Domains of Premenstrual Symptoms

Abstract

Premenstrual symptoms are distressing and impairing for individuals and costly to society. These symptoms are heterogeneous within and across people, dimensional, and dynamic. While some efforts have been made to understand the trajectories of premenstrual symptoms, two major gaps in the literature remain. First, we lack understanding of the covariation among symptoms over the course of the menstrual cycle. Second, we know little about the trajectories of these symptoms and why symptoms might take different courses. To address these gaps, a sample of female undergraduates (N = 85) who reported no use of hormonal birth control and regularly occurring menstrual periods were recruited for a 4-month-long electronic daily diary study of premenstrual symptoms. We explored the covariation of symptoms over the cycle by conducting a multilevel exploratory factor analysis of the daily diary items.We identified six distinct but correlated symptomdomains at the withinperson level which were affective, cognitive, interpersonal, pain, and somatic. Next, we characterized the trajectories of each symptom domain using multilevel natural spline models and their first/second derivatives. Somatic symptoms increased/decreased more sharply and quickly than other symptom domains, pointing to a unique trajectory. Interpersonal and affective symptoms, on the other hand, were milder throughout. We demonstrated the importance of investigating the differences among symptom domain trajectories and underscored the need for future research to elucidate the unique mechanisms that underlie each trajectory.