We welcome this opportunity to provide further information on our study. Screening for our patients took place after clinical remission. This carried 2 important consequences. First, patients were treated in a naturalistic way,1 pushing the antidepressants dosages as high as reasonably possible according to clinical judgment.2 Had we endorsed a fixed treatment strategy with a specific antidepressant, we would have obtained a more homogeneous, but less representative, sample. The second consequence of screening after remission is that it allows a better evaluation of comorbidity (certain types of comorbidity wane with the abatement of depressive symptoms).3 Of the 58 consecutive patients who had 3 or more episodes of unipolar depression, 6 patients (10%) were excluded for unsatisfactory clinical response and 7 (12%) met other exclusion criteria (mainly presence of personality disorder and/or double [ie, and/or antecedent dysthymia] depression). Such percentages would have been higher if the patients had been screened before treatment. Probably because of the satisfactory response, no patient declined to participate in the study. Discontinuation of treatment with antidepressant drugs was not feasible for 5 of the 45 patients who had been randomized to cognitive behavioral treatment or clinical management. These patients, in fact, displayed the emergence of symptoms during drug tapering and were brought back to the original dose. We have no way of knowing whether a relapse would have indeed occurred with the discontinuation of therapy with antidepressants. Intent-to-treat analysis is important for a correct interpretation of studies with unbalanced dropout rates. In this case, it was not. Considering the 3 patients in the cognitive behavioral treatment and the 2 in the clinical management groups as relapsers does not change the results: patients in the cognitive behavioral therapy group had a significantly higher (t43 = 2.25, P<.05) survival time (mean, 79.8 weeks; SD, 37.9 weeks) than patients in the clinical management group (mean, 56.5 weeks; SD, 31.2 weeks), and treatment assignment was still significant (log-rank test, χ21 = 8.47; P<.01) by survival analysis. We did not test the ability of the blind rater to identify correctly the treatment assignment. However, most of the assessments were simply concerned with the presence of major depression.