To examine reproductive onset as a function of AD history, without and with adjustment for control variables, Cox proportional hazards regression (Cox, 1972) was used. For both cohorts, AD was modeled as a time-invariant predictor of reproductive onset, as were all control variables. AD and relevant control variables (MDD, suicide attempt, panic attack, and CSA), were also examined as time-varying predictors in the young cohort to ensure onset before or at the same time of first reproduction. In these analyses, the Efron approximation (Effron, 1977) was used for survival ties. For Cohort I, examination of AD, MDD, suicide attempt, panic attack, and CSA as time-varying predictors of reproductive onset was not possible because age of onset was not available for these measures.
