The figure shows the numbers of patients recruited and followed up. There was no systematic difference between the proportions of patients completing the 12 month questionnaire in terms of treatment or whether they were randomised (P=0.34 for heterogeneity between the four groups). Sixty five (63%) patients in the randomised trial completed the Beck depression inventory and SF-36 at 12 months compared with 142 (65%) in the patient preference trial. The proportions of patients in each group who kept their 12 month appointment differed (P=0.01), with attendance ranging between 25% for patients choosing antidepressants and 53% for those randomised to antidepressants.
We abstracted the general practitioner's notes for 96% (99/103) of patients in the randomised trial and 96% (212/220) in the patient preference. There was sufficient information to carry out the psychiatrist's overall assessment on 79% (81/103) of patients in the randomised trial and 74% (163/220) in the patient preference trial.
Patient characteristics at entry
The patients were predominantly women, there were roughly equal numbers in social classes I/II, III, and IV/V, and 57% were married (table ). Patients preferring counselling were less severely depressed than the randomised patients or those preferring antidepressants.3
Summary of baseline characteristics. Values are numbers (percentages) unless stated otherwise
Beck inventory scores at 12 months
Mean Beck scores did not differ significantly between the two groups in the randomised controlled trial (P=0.49, table ). There was no evidence for an interaction between treatment and preference (P=0.6) so we combined the randomised and patient preference groups. Mean Beck scores were similar in counselled patients and those receiving antidepressants (13.2 v 12.8, 95% confidence interval for difference −2.7 to 3.5).
Scores on Beck depression inventory scale at 12 months
We investigated the effect of patient preference by comparing the randomised and patient preference groups for the two treatments separately. There was no difference in the mean Beck scores for the patients treated with antidepressants (mean difference 3.1, 95% confidence interval –1.8 to 7.8), but the patients choosing counselling did better than those randomised to counselling (mean difference 4.6, 95% confidence interval 0.0 to 9.2).
Global outcome and time to remission
We found no differences in global outcome between the randomised or patient preference trials when outcome was split into good or moderate versus poor (table ). Stratification by randomised or patient preference status gave similar outcomes for antidepressants and counselling (Mantel Haenszel P=0.63). Similar results were obtained by splitting the outcomes into good versus moderate or poor. The treatment allocations were guessed correctly by the assessor in nine of the 20 sets of notes, indicating that the masking was satisfactory.
Global outcome, remission, and relapse in randomised and patient preference groups. Values are numbers (percentages) of patients. Totals exclude missing data but include patients in whom outcome was uncertain
Overall, 68% (221/323) of patients had a remission or 83% (221/265) of those with a known outcome. The proportions of patients who had a remission were similar in each group (table ). Median time to remission was three months in all groups except the group randomised to antidepressants, where the median time to remission was two months (comparing randomised groups logrank statistic 2.74, P=0.1; pooled logrank statistic for randomised and patient preference trials 0.82, P=0.36). Thirty three (15%) of the 221 patients had a relapse. There were no differences between the groups.
Research diagnostic criteria scores
Of the randomised patients who kept their 12 month follow up, nine (47%) in the counselling group were no longer depressed compared with 21 (78%) in the antidepressant group (P=0.07, table ). When we assumed that all those failing to attend had recovered, then 81% in the counselling group and 88% in the antidepressant group were no longer depressed. The figures when we assumed that patients failing to attend were treatment failures were 17% and 41% respectively (P=0.01).
Comparison of outcome at 12 months on research diagnostic criteria in patients receiving counselling or antidepressants
In the patient preference trial, 48 (80%) patients choosing counselling and 17 (85%) choosing antidepressants had recovered at 12 months (P=0.87). If missed appointments were assumed to be treatment successes the outcomes were similar, but if all missed appointments were treatment failures, patients choosing counselling would do better than those choosing antidepressants.
There were no differences in scores between patients randomised to counselling or antidepressants on any of the SF-36 domains (data not shown). The scores were also similar when the randomised and preference arms were combined. Patients who chose counselling did consistently better than those randomised to counselling on all dimensions of the SF-36.
Two (2%) randomised patients and 15 (7%) preference patients were referred for outpatient assessment (P=0.1, table ). Few patients received inpatient treatment relating to depression, and there were no differences between the groups. There were no differences in the proportions of patients attending as inpatients for unrelated conditions.
One 46 year old man committed suicide two months after enrolling in the study in the patient preference antidepressants arm. His Beck score at entry was 43. In accordance with the study protocol, the patient's general practitioner was notified of this score.