Two dimensions of compliance (drop-outs and adherence) were investigated in patients treated with antidepressant drugs. Efficacy, compliance and its determinants were investigated in 66 patients suffering from major depressive disorder and treated in a double-blind manner with fluoxetine 20 mg/day or amitriptyline 150 mg/day for 9 weeks. Overall effectiveness [50% decrease in the initial Hamilton Rating Scale for Depression (HAM-D)] was similar in both groups (62.8% for fluoxetine, 58.1% for amitriptyline). The dropout rate due to side effects was 35.5% for amitriptyline and 5.7% for fluoxetine. A logistic regression analysis revealed that the initial HAM-D score was not predictive for dropping out, but this outcome was instead determined by sex (increased risk for males), age (increased risk for being younger) and occurrence of severe side effects. Adherence was estimated using electronic Medication Event Monitoring System and defined as the percentage of days when the correct dose was taken out of the medication container. Of the patients studied 37% had an adherence of less than 70%. There was no relationship between adherence and efficacy and adherence was similar in patients on fluoxetine or amitriptyline. Side effects were not predictive of being adherent or not, but a higher initial HAM-D score predicted a higher adherence to the medication regimen. The demographic variables had no significant effect. The present study suggests that the link between efficacy, side effects and compliance or adherence is more complex than is generally believed and that early termination and non-adherence seem to be determined by different factors.