Our full dataset, and the methods used to analyze it, depart in a number of respects from those in the original report by Caspi et al. [2003]. Therefore, we also attempt to replicate their findings by restricting our data set and analyses to match theirs as closely as possible. To do this, we restricted age at time of SLE reporting to 23–29, removing 80% of the participants, and used a model which included sex, PLE (truncated here to a maximum of five events, not seven), an additive model for 5HTTLPR genotypes, and a PLE × 5HTTLPR interaction.
