Omitting the subjects with bipolar 1 and 2 from the analysis did not change the significance or effect size estimates for rs9804190 or the peak SNPs for either PTSD or externalizing (Supplementary Table 1). Additionally, as 60% of our data occur in couple pairs, we explored whether there might be non-independence of either PTSD or externalizing within couples that could bias results. We found no evidence that PTSD was correlated within couples (estimated OR for lifetime PTSD diagnosis=0.63, Fisher’s exact test p=0.24). However, externalizing was moderately correlated within couple pairs (r=0.39, p=2.97×10−7). Therefore, we additionally examined externalizing using GEE models, explicitly allowing for correlation within couple pairs. GEE results for externalizing-associated SNPS (Table 1 SNPs 1 to 4) were very similar to those obtained without adjustment for couple status (β = −0.39 to −0.43, p = 0.0002 to 0.0006). Hence, it does not appear that the strength of the observed associations was inflated by dependence between subjects.
