Psychologic Treatment of Depression Compared With Pharmacotherapy and Combined Treatment in Primary Care

A Network Meta-Analysis

Pim Cuijpers, PhD; Matthijs Oud, MA; Eirini Karyotaki, PhD; Hisashi Noma, PhD; Soledad Quero, PhD; Andrea Cipriani, MD, PhD; Bruce Arroll, PhD; Toshi A. Furukawa, MD, PhD


Ann Fam Med. 2021;19(3):262-270. 

In This Article


Selection and Inclusion of Studies

We examined 21,976 abstracts (16,701 after removal of duplicates) and retrieved 2,553 full-text papers. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart describing the inclusion process[43] is presented in Supplemental Appendix 3 ( A total of 58 studies met the inclusion criteria (9,301 patients). Selected characteristics of the included studies are summarized in Supplemental Appendix 2, and references in Supplemental Appendix 4 ( In 4 studies, 2 types of psychotherapy were examined as separate arms (including 1 in which the treatment was provided by either a nurse or a GP). In total, 62 comparisons were available for the NMA (psychotherapy: 56; pharmacotherapy: 16; combined: 5; care as usual [CAU]: 39; waitlist: 6; pill placebo: 3).

Characteristics of Included Studies

The aggregated characteristics of the included studies are presented in Table 1. In 19 studies, patients were recruited via systematic screening. Twelve trials were aimed at specific target groups, 32 used CBT as therapy, 39 used an individual treatment format, and the therapy was adequate in 28. A selective serotonin reuptake inhibitor was applied in 7 of the 16 studies examining pharmacotherapy, and in 9 of these the pharmacotherapy was adequate.

A total of 44 studies reported an adequate sequence generation, 44 reported allocation to conditions by an independent party, 25 reported masking of outcome assessors, and 30 used only self-report outcomes. In 44 studies, intention-to-treat analyses were conducted. A total of 28 studies met all 4 RoB criteria, 14 met 3 criteria, and 20 met ≤2 criteria.

Pairwise Meta-analyses

Supplemental Appendix 5 ( shows the results of the pairwise meta-analyses. Forest plots for response rates with ≥5 comparisons are shown in Supplemental Appendices 6–9 ( Heterogeneity was low to moderate for most comparisons. For the comparisons on response, only heterogeneity of combined treatment vs pharmacotherapy was high (84%). The Egger test was significant only for psychotherapy vs care as usual (P = .01) and for psychotherapy vs waitlist (P = .01).

Network Plot

The network for response is presented graphically in Figure 1. The number of studies for each comparison is listed in Supplemental Appendix 5. The most examined nodes were psychotherapy, CAU, and pharmacotherapy. A small number of comparisons included combined treatment, pill placebo, or waitlist. Several nodes were not well connected to the network. Waitlist had a small connection to psychotherapy and none to any other node. The control conditions were not connected to each other, and combined treatment was not connected to any of the control conditions. The contribution plot, showing the percentages of contributions from the direct comparisons separately for the mixed and indirect estimates, is presented in Supplemental Appendix 10 (

Figure 1.

Network plot.

Network Meta-analysis

The main results of the NMA for response, remission, acceptability, and SMD are presented in Table 2. The outcomes are presented graphically in Figure 2. No significant difference was found between psychotherapy and pharmacotherapy for response (RR = 1.03; 95% CI, 0.88–1.22). Combined treatment was significantly more effective than psychotherapy (RR = 1.35; 95% CI, 1.00–1.81) but not than pharmacotherapy (RR = 1.30; 95% CI, 0.98–1.73), although this might have been caused by low power. Psychotherapy, pharmacotherapy, and combined treatment were all more effective than CAU (RRs = 1.60, 1.65, 2.15, respectively) and waitlist (RRs = 2.35, 2.43, 3.16, respectively). The number of studies including pill placebo was too small to result in meaningful outcomes.

Figure 2.

Ranked forest plots.
CAU = care as usual; RR = relative risk; SMD = standardized mean difference.

The outcomes for remission were comparable to those for response, with the exception that combined treatment was not significantly different from psychotherapy. The SMDs for CAU ranged from 0.70 (95% CI, 0.35–1.05) for combined treatment to 0.44 (95% CI, 0.31–0.57) for psychotherapy and 0.41 (95% CI, 0.18–0.64) for pharmacotherapy. None of the outcomes for acceptability were significant.

The distribution of potential effect modifiers for the 4 comparisons with ≥5 studies is presented in Table 1. Visual inspection of the distribution across comparisons indicated that the potential effect modifiers were similarly distributed across the comparisons. This suggested no significant evidence against the transitivity assumption. However, this must be considered with caution because of the small sample sizes in some cells.

Examination of consistency with the loop-specific approach (Supplemental Appendix 11, indicated that no loop was significantly inconsistent. However, this cannot be considered as evidence for the absence of inconsistency because of the small or zero number of comparisons in several loops. The design-by-treatment interaction model did not indicate global inconsistency in the network (χ 2 = 8.02; df = 5; P for the null hypothesis of consistency in the network = .16).

Ranking of Treatments

The results for surface under the cumulative ranking curve are summarized in Table 3. Combined treatment ranked clearly best for response, remission, and SMD. There were no large differences between psychotherapy and pharmacotherapy for response, remission, or SMD. No clear directions were found for acceptability.

Heterogeneity and Metaregression

The common τ2 estimates were 0.06 for response, 0.12 for remission, 0.10 for SMD, and 0.09 for acceptability. Compared with the empirically predicted distribution for semiobjective outcomes in drug vs placebo comparisons (median 0.049; 95% CI, 0.001–1.83),[41] the heterogeneity variance estimates would be moderate.

The results of the multivariate metaregression analysis that was conducted to examine possible sources of heterogeneity are shown in Supplemental Appendix 12 ( For several comparisons, a considerable number of outcomes were not available because there were not enough studies in the comparisons or because of collinearity. Only 1 predictor (screening vs other recruitment of patients) was found to be significant in the comparison between psychotherapy and CAU (P = .03).

Sensitivity Analyses and Follow-up

The results of the sensitivity analyses in which we included only studies with low RoB (Supplemental Appendix 13, resulted in outcomes comparable to the main analyses. The only exception was that the difference between psychotherapy and combined treatment, as well as the difference between pharmacotherapy and combined treatment, was now significant. In the analyses in which we included only trials on CBT (Supplemental Appendix 14,, we also found that combined treatment was significantly more effective than either CBT or pharmacotherapy alone. The results of the other sensitivity analyses that were conducted are shown in Supplemental Appendices 15–19 ( Overall, these analyses supported the main findings of the study.

A total of 27 studies reported outcomes at ≥6 months follow-up, but the follow-up periods differed considerably, and because of the small number of studies for each of the different follow-up periods, we decided not to analyze those data.