Summary Objective We investigated whether treatment response is predicted by hypothalamus–pituitary–adrenal (HPA) axis parameters, or by genetic polymorphisms in the glucocorticoid receptor (GR), that regulates its feedback. Methods Ninety-eight outpatients completed 8 weeks of paroxetine treatment. Treatment response was defined as a 50% decrease in Hamilton Rating Scale for depression (HRSD) ratings. At baseline, 24 h urinary cortisol excretion, and cortisol and ACTH concentrations in a DEX/CRH test were measured. The presence of polymorphisms in the GR DNA sequence (BclI, ER22/23EK, N363S) was determined. Prediction of treatment response was analysed by calculating response rates per tertile of an HPA-axis parameter and per GR genotype. Results The response rate in the high ACTH tertile was significantly lower as compared to the intermediate tertile, but not compared to the low tertile (response rates from high to low tertile: 33%, 67% and 42%). Carriers of the BclI polymorphism had higher ACTH values than non-carriers (baseline ACTH: 3 versus 5 ng/l, p=0.02) and showed a trend towards lower decrease of HRSD rates than non-carriers (HRSD decrease: 8 versus 11, respectively, p=0.07). In a subgroup of BclI carriers, patients in the high ACTH tertile had a lower decrease in HRSD and lower response rates than patients in the low ACTH tertiles (HRSD decrease from high to low tertile: 5, 9 and 11, p < 0.01 ). Conclusion The results suggest that hyperactivity of the HPA-axis predict worse treatment outcome. The BclI polymorphism explains, in part, DEX/CRH test results and tends to be associated with worse treatment outcome.