Chunk #2 — Bias in a protocol for a meta-analysis of 5-HTTLPR, stress, and depression — Issue 1: “Primary Analysis Plan 2” to study lifetime depression does not allow for establishing temporal order between stress and depression
consequence of their mood disorder [3]. This well-known phenomenon is referred to in the literature as “stress generation” [4]. For example, depressed individuals have elevated rates of intimate partner violence and divorce. To use retrospective reports of lifetime depression in a test of this GxE hypothesis is tantamount to using lifetime weight to test hypotheses about the cause of low birth-weight, or to use lifetime IQ to test hypotheses about causes of IQ decline in Alzheimer’s dementia; the measure sounds the same, but it is not. Timing is everything. The importance of temporal order in hypothesis testing in studies having observational designs is nicely explained in a powerpoint lecture "What Do Survey Data Really Mean? Considering Issues of Causality and Temporality in Survey Research," by Seth Noar (http://www.nidcr.nih.gov/Research/DER/BSSRB/PowerPointPresentations/default.htm). Strong GxE tests documenting that stress antedated depression exist, but are not included in the meta-analysis (e.g., [5]).
