Predictive validity of a DNA methylation-based screening panel for postpartum depression

Abstract

AbstractPrenatal maternal depression increases the risk of negative maternal-infant health outcomes but often goes unrecognized. As a result, biomarker screening tests capable of identifying women at risk for depression are highly desirable. This study tested how demographic and clinical factors affect the predictive validity of a DNA methylation-based screening test for postpartum major depression (MD) using data from a longitudinal study of birth outcomes. Lifetime history of MD and current levels of postpartum depressive symptoms were assessed using an extended self-report version of the Composite International Diagnostic Interview Short Form and the Edinburgh Postnatal Depression Scale (EPDS), respectively. Predictive validity of the test was estimated in the PREG cohort using the area under the receiver operator characteristic curve (AUC), and sensitivity analyses were performed to assess the impact of self-reported race, age, and pre-pregnancy history of MD. Data for N=103 pregnant participants (African-American=49; European-American=54) were available. The prediction model identified women who would develop high levels of postpartum depressive symptoms better within the subset of women with previous histories of MD (AUC = 0.94, 95% CI 0.79-1.00) compared to the full pregnant cohort (AUC = 0.62, 95% CI 0.46-0.79). This observation prompted secondary analyses to test the model specificity for postpartum depression. The model predicted lifetime history of MD moderately well in never-pregnant, mixed-sex cohort of adolescents (N=150; ages 15-20; AUC = 0.75, 95% CI 0.57-0.92) and performed slightly better in males versus females. Additional sensitivity analyses are needed to determine the extent of the model’s specificity for MD subtypes and if demographic or clinical factors influence the predictive validity of this model.