Background Antidepressants’ effects are established in randomised controlled trials (RCTs), but not in the real world. Aims To investigate real-world comparative effects of antidepressants for depression and compare them with RCTs. Method We performed a cohort study based on the QResearch database. We included people with a newly recorded diagnosis of depression, exposed to licensed antidepressants in the UK. We assessed all-cause dropouts (acceptability), dropouts for adverse events (tolerability), occurrence of at least one adverse event (safety), and response and remission on the Patient Health Questionnaire (PHQ)-9 (effectiveness) at 2 and 12 months. Logistic regressions were used to compute adjusted-odds ratio (aOR) with 99% CIs, assessing the associations between exposure to each antidepressant against fluoxetine (comparator) and outcomes of interest. We compared estimates from the real world with RCTs using ratio-of-odds ratio (ROR) with 95% CI. Results A total of 673 177 depressed people were studied: females 57.1%, mean age 42.8 (s.d. 17.7) years, mean baseline PHQ-9 17.1 (s.d. 5.0) (moderately severe depression). At 2 months, antidepressant acceptability was 61.4%, tolerability 94.4%, safety 54.5%, PHQ-9 decreased to 12.3 (s.d. 6.5). At 12 months, acceptability was 12.3%, tolerability 87.5%, safety 28.8%, PHQ-9 12.9 (s.d. 6.8). In the short and long term, tricyclics, mirtazapine and trazodone were worse than fluoxetine for most outcomes; citalopram had better acceptability than fluoxetine (aOR 0.95; 99% CI 0.92, 0.97), sertraline had lower tolerability (aOR 1.12; 99% CI 1.06, 1.18), and both citalopram and sertraline had lower safety (aOR 1.17 and 1.25, respectively). In the long term, citalopram had better acceptability (aOR 0.78; 99% CI 0.76, 0.81) and effectiveness (aOR 1.12 for both response and remission), but worse tolerability (aOR 1.09; 99% CI 1.06, 1.13) and safety (aOR 1.12; 99% CI 1.08, 1.16). Observational and randomised data were similar for citalopram and sertraline, while there was some difference for drugs less prescribed in the real world. Conclusions Antidepressants showed low acceptability, moderate-to-high tolerability and safety, and small-to-moderate effectiveness in the real world. Real-world and RCT estimates showed similar findings only when the analyses were carried out using large datasets; otherwise, the results diverged.

Real-world effects of antidepressants for depressive disorder in primary care: population-based cohort study