Data analytic plan

Analyses were conducted in PASW Statistics Version 21.0. To test the independent and interactive effects of abuse history and depression severity on IL-6 and CRP, linear regression analyses were conducted separately for each time point and each outcome controlling for age and BMI; additionally, at T3, we also controlled for T2 inflammatory markers. Abuse and depression were centered and entered on the first step along with covariates, and an interaction term reflecting the product of these centered variables was entered on the second step. To examine whether abuse or depression interacted with circulating levels of inflammatory markers to predict poor birth outcomes, separate linear regressions were conducted for gestational age at birth and birth weight and logistic regression was used for any birth complications. Separate models were estimated for the interactions between abuse and inflammatory markers and the interactions between depression and the inflammatory markers. For all birth outcomes, mode of delivery (0=spontaneous vaginal delivery, 1=vaginal delivery with induction/augmentation, or 2= cesarean section) was included as an additional categorical covariate. Finally, to examine whether cortisol at each time point mediated associations between inflammatory markers at that trimester and poor birth outcome, we tested indirect effects models with bootstrapped standard errors specifying that inflammatory markers predicted cortisol at the same occasion, which in turn predicted poor birth outcomes.