Rationale Cytochrome P450 enzymes are important in the metabolism of antidepressants. The highly polymorphic nature of these enzymes has been linked to variability in antidepressant metabolism rates, leading to hope regarding the use of P450 genotyping to guide treatment. However, evidence that P450 genotypic differences underlie the variation in treatment outcomes is inconclusive. Objectives We explored the links between both P450 genotype and serum concentrations of antidepressant with antidepressant side effects, using data from the Genome-Based Therapeutic Drugs for Depression Project (GENDEP), which is a large ( n  = 868), pharmacogenetic study of depressed individuals treated with escitalopram or nortriptyline. Methods Patients were genotyped for the enzymes CYP2C19 and CYP2D6, and serum concentrations of both antidepressant and primary metabolite were measured after 8 weeks of treatment. Side effects were assessed weekly. We investigated associations between P450 genotypes, serum concentrations of antidepressants and side effects, as well as the relationship between P450 genotype and study discontinuation. Results P450 genotype did not predict total side effect burden (nortriptyline: n  = 251, p  = 0.5638, β  = −0.133, standard error (SE) = 0.229; escitalopram: n  = 340, p  = 0.9627, β  = −0.004, SE = 0.085), study discontinuation (nortriptyline n  = 284, hazard ratio (HR) = 1.300, p  = 0.174; escitalopram n  = 376, HR = 0.870, p  = 0.118) or specific side effects. Serum concentrations of antidepressant were only related to a minority of the specific side effects measured: dry mouth, dizziness and diarrhoea. Conclusions In this sample where antidepressant dosage is titrated using clinical judgement, P450 genotypes do not explain differences between patients in side effects with antidepressants. Serum drug concentrations appear to only explain variability in the occurrence of a minority of specific side effects.