Abstract
Background: There are two primary phenotypic models of comorbidity between post-traumatic stress disorder (PTSD) and drug use disorder (DUD), i.e. self-medication (PTSD precedes and causes DUD) and susceptibility (DUD precedes and causes PTSD). We sought to clarify the longitudinal relationship between PTSD and DUD, while examining sex differences.Method: We used approximately 23 years of longitudinal data from Swedish population registries to conduct two complementary statistical models: Cox proportional hazard models (N ≈ 1.5 million) and a cross-lagged panel model (N ≈ 3.8 million).Results: Cox proportional hazards models, adjusting for cohort and socioeconomic status, found strong evidence for the self-medication hypothesis, as PTSD predicted increased risk for DUD among both women [hazard ratio (HR) = 5.34, 95% confidence interval (CI) 5.18, 5.51] and men (HR = 3.65, 95% CI 3.54, 3.77), and moreover, that the PTSD to DUD association was significantly higher among women (interaction term 0.68, 95% CI 0.65, 0.71). The results of the susceptibility model were significant, but not as strong as the self-medication model. DUD predicted risk for PTSD among both women (HR = 2.43, 95% CI 2.38, 2.50) and men (HR = 2.55, 95% CI 2.50, 2.60), and HR was significantly higher in men (interaction term 1.05, 95% CI 1.02, 1.08). Investigating the pathways simultaneously in the cross-lagged model yielded support for both pathways of risk. The cross-paths instantiating the susceptibility model (0.10-0.22 in females, 0.12-0.19 in males) were mostly larger than those capturing the self-medication model (0.01-0.16 in females, 0.04-0.22 in males).Conclusions: We demonstrate that the relationship between PTSD and DUD is bidirectional, with evidence that future research should prioritize examining specific pathways of risk that may differ between men and women.