regarding replication. For example, several studies only found significant effects for 1 genotype (SS vs SL plus LL or SS plus SL vs LL) rather than the dose effect reported by Caspi et al21,22,24,26,35,37 or found an interaction effect in the opposite direction from that in the original report (the L allele was associated with stressful life events and depression15,20,24,32,35,41). Third, several studies that did not replicate the overall association between genotype, stressful life events, and depression reported replication based on post hoc subgroup analyses of specific age or sex subgroups. 18,20,24,35,41 Although such subgroup analyses may test the generalizability of the original findings, they do not constitute replications unless their findings occur using the same stratification and directionality as those of the original study.
