Background Melancholic and atypical depression are widely thought to moderate or predict outcome of pharmacological and psychological treatments of adult depression, but that has not yet been established. This study uses the data from four earlier trials comparing cognitive behavior therapy (CBT) versus antidepressant medications (ADMs; and pill placebo when available) to examine the extent to which melancholic and atypical depression moderate or predict outcome in an “individual patient data” meta‐analysis. Methods We conducted a systematic search for studies directly comparing CBT versus ADM, contacted the researchers, integrated the resulting datasets from these studies into one big dataset, and selected the studies that included melancholic or atypical depressive subtyping according to DSM‐IV criteria at baseline (n = 4, with 805 patients). After multiple imputation of missing data at posttest, mixed models were used to conduct the main analyses. Results In none of the analyses was melancholic or atypical depression found to significantly moderate outcome (indicating a better or worse outcome of these patients in CBT compared to ADM; i.e., an interaction), predict outcome independent of treatment group (i.e., a main effect), or predict outcome within a given modality. The outcome differences between patients with melancholia or atypical depression versus those without were consistently very small (all effect sizes g < 0.10). Conclusions We found no indication that melancholic or atypical depressions are significant or relevant moderators or predictors of outcome of CBT and ADM.