presents the study’s Consort Chart. All 60 protocol eligible patients agreed to enroll; 38 (63%) were 21–59 years old, and 22 (37%) were 60 or older. The sample of 60 randomized participants was predominantly female and included diverse ethnicities (). Two-thirds of participants self-identified as minority group members. Participants typically experienced depression of moderate severity, extended duration and recurrent nature, with 35% reporting passive suicidal ideation at baseline and 64% having a prior history of antidepressant treatment and/or psychotherapy. Randomized groups did not differ on sociodemographic or clinical variables. Mean baseline HRSD scores for participants assigned to congruent versus incongruent treatment were 22.7 (sd=5.3) and 24.6 (sd=6.3), respectively.
Characteristics and treatment preferences of study participants (N=60)
Of participants with previous antidepressant experience (n=22), 14 (64%) agreed or strongly agreed that “the treatment was effective” and 15 (68%) agreed or strongly agreed that “the treatment resulted in troubling side effects, or made me more distressed.” Of those with previous psychotherapy experience (n=31), 19 (61%) agreed or strongly agreed that “the treatment was effective” and 9 (29%) agreed or strongly agreed that “the treatment resulted in troubling side effects, or made me more distressed.”
Follow-up data on treatment adherence were available for all participants. HRSD 12-week data were available for 26 (90%) congruent and 27 (87%) incongruent participants. HRSD 24-week data were available for 25 (86%) congruent and 24 (77%) incongruent participants. Participants who dropped out by week 12 had lower rates of treatment adherence throughout the study period (mean proportion of attended sessions or pills taken=0.09, sd=9.6 versus 0.67, sd=35.9 for participants who were followed; t=4.20, df=58, p<0.001). No other variable distinguished dropouts from completers.
Ranked treatment preferences are presented in . When patient preferences were restricted to either of the two study treatments, 42 (70%) participants selected individual psychotherapy while the remaining 18 (30%) selected antidepressant medication. An analysis of preference strength for the entire sample revealed stronger preferences for psychotherapy than antidepressants (paired t=5.6, df=58, p<0.001). The mean preference strength for psychotherapy was 4.1, signifying “agreement;” in contrast the mean preference strength of 2.9 for antidepressants, indicating “neutral or indifferent.” Preference strength for the treatments was unrelated to age.
Participants anticipated greater improvement from psychotherapy (mean=0.72, sd=0.2) than from antidepressant medication (mean=0.49, sd=0.3; paired t=4.4, df=54, p<0.001). Expected improvement again did not differ by age group.
Preference strength and treatment initiation
All participants (n=29) randomized to a treatment congruent with their preference initiated treatment; only 23/31 (74%) of the incongruent group did so (Fisher’s Exact Test=0.005). All 8 participants failing to initiate treatment had been assigned antidepressant medication. While these participants were offered extra-protocol psychotherapy referrals, none had pursued such treatment when assessed at periodic follow-ups.
Using logistic regression, treatment initiation was associated with stronger preferences (OR=5.3, 95% CI=4.3, 6.3, df=1, p=0.001; ), expectation of improvement from the assigned treatment (OR=2.5, 95% CI=1.9, 3.1, df=1, p=0.002), but no sociodemographic or clinical variable. In a combined model, preference strength was associated with treatment initiation over and above expected improvement (OR=4.4, 95% CI=3.3, 5.5, df=1, p=0.009). Interaction terms of preference strength by age group (mid-life versus elderly) and depression severity (HRSD score) were not significant.
Treatment initiation by preference strength
Preference strength and treatment adherence
Preference strength for assigned treatment, but not simply congruence, was associated with higher 12-week treatment adherence rates (Beta=13.40, p=0.002; ). Differences in adherence rates between participants assigned to psychotherapy (mean=0.68, sd=31.9) and antidepressant medication (mean=0.52, sd=42.9) were not significant. No other variables were associated with adherence, although expected improvement from the assigned treatment approached significance (Beta=6.65, p=0.060). Interaction terms of preference strength by age group (mid-life versus elderly), treatment type (antidepressant medication versus psychotherapy), and depression severity (HRSD score) were not significant.
Average level of adherence by preference strength
Preference strength and outcome
Across groups, mean HRSD ratings at 12 and 24 weeks were 16.4 (sd=8.3, range=3–38) and 16.3 (sd=9.9, range=2–42) respectively. Remission rates (HRSD≤7) at 12 and 24 weeks were 21% (11/53) and 29% (14/49), respectively.
Congruent treatment was not significantly related to 12 and 24-week HRSD scores in linear regression models controlling for baseline HRSD score. Contrary to prediction, preference strength was negatively associated with symptom severity at 12 weeks (Beta=1.9, p=0.028) and had no effect at 24 weeks. Neither treatment congruence nor preference strength was related to 12 and 24-week remission (using Fisher’s Exact Test and logistic regression, respectively). .
Participants assigned psychotherapy achieved significantly lower 24-week HRSD scores (mean=14.0, sd=9.4) than those assigned medication (mean=18.9, sd=10.1; F=4.12, df=1, p=0.048). These treatment groups did not differ in 12-week HRSD scores, or in 12 or 24-week remission rates.
Interaction terms of preference strength by age group, treatment type, and depression severity were not significant. Results did not change when available 8-week HRSD scores for 2 participants were substituted for missing 12-week scores. Finally, application of mixed-effects regression models as sensitivity analyses that accommodated all observed data across all assessment time points yielded results consistent to that which are reported above.
Other predictors of outcome
Regressions controlling for baseline HRSD scores found degree of adherence (i.e., proportion of pills taken or therapy sessions attended) unrelated to 12 or 24-week depression severity or remission.
A simultaneous regression of significant bivariate demographic and clinical predictors indicated that lower baseline HRSD (Beta=0.48, p=0.010), being White (Beta=4.32, p=0.062), lower CANE scores (Beta=0.90, p=0.007), and higher MMSE scores (Beta=−1.61, p=0.013) predicted lower 12-week HRSD scores (R2=0.57, p<0.001). In a separate regression, higher MMSE scores (Beta=−2.24, p=0.007) and higher functioning as measured by the WHODAS (Beta=0.44, p=0.012) predicted lower 24-week HRSD scores (R2=0.49, p<0.001).