Logistic regression models were used to test the associations between the explanatory (number of S alleles; number of life events) and response variables (depression). Inclusion of the interaction term tests for deviation from a log-additive model of risk, and may be significant even when terms for main effects are not. To test for heterogeneity in all analyses, we performed χ2 tests of heterogeneity, based on the derived regression β coefficients and their standard errors, assuming them to be normally distributed. Because of potential heterogeneity of the effects of explanatory variables across the individual studies, we used random-effects models to estimate the average effect across studies. Inclusion of an ordinal variable in logistic regression provides a statistical test for trends in the effects of this variable on a dichotomous outcome. We had 80% power to reject the null hypothesis of no interaction between stressful life events and 5-HTTLPR genotype in favor of a positive interaction if the true OR was 1.10 or greater. All statistical analyses were conducted using SAS version 9 (SAS Institute Inc, Cary, North Carolina), Microsoft Excel 2007 (Microsoft, Redmond, Washington), and Stata version SE 9 (StataCorp, College Station, Texas).
