Decision tool improved antidepressant persistence and symptoms
A web-based decision-support tool reduced antidepressant stopping and modestly improved depression and anxiety symptoms.
*Randomized clinical trial; Level 1b (OCEBM).

Citation

Cipriani A, Fernandes KBP, Mulsant BH, et al; for the PETRUSHKA Team. A Decision-Support System to Personalize Antidepressant Treatment in Major Depressive Disorder: A Randomized Clinical Trial. JAMA. 2026;335(14):1219-1231. doi:10.1001/jama.2026.1327

Background

Many people stop antidepressants early, often because the first choice is a poor fit. This trial tested whether a shared, web-based tool that considers patient characteristics and side-effect preferences improves treatment continuation and symptoms.

Patients

Adults aged 18-74 years with major depressive disorder at 47 sites (Brazil, Canada, UK). Key exclusions: treatment-resistant depression, pregnancy, heart rhythm problems, urgent mental health care or hospitalization.

Intervention

PETRUSHKA decision-support tool used to select an antidepressant.

Control

Usual clinician-selected antidepressant.

Outcome

Primary: stopping the prescribed antidepressant for any reason at 8 weeks. Secondary: stopping due to adverse events; depression (Patient Health Questionnaire-9; 0-27) and anxiety (Generalized Anxiety Disorder-7; 0-21) at 24 weeks.

Follow-up Period

24 weeks (primary outcome at 8 weeks).

Results

Primary analysis included 493 participants (median age 35 years; 58% female). Analyses followed assigned groups; some 24-week outcomes had substantial missing data.
Outcome PETRUSHKA Usual care Adjusted effect (95% CI) NNT
Stopped antidepressant by 8 weeks (primary) 17% (41/241) 27% (69/252) Relative risk 0.62 (0.44 to 0.88) 10
Stopped due to adverse events by 8 weeks 9% (22/241) 16% (39/252) Relative risk 0.59 (0.36 to 0.97) 15
Depression symptoms at 24 weeks Mean 7.1 Mean 9.2 1.92 points lower (0.78 to 3.06 lower) —
Anxiety symptoms at 24 weeks Mean 4.6 Mean 5.8 1.39 points lower (0.52 to 2.26 lower) —

Limitations

Patients and clinicians were not blinded, which could influence stopping decisions and self-rated symptoms. Many 24-week outcomes were missing, and improvements in symptom scores were statistically significant but modest.

Funding

UK National Institute for Health and Care Research; University of Oxford sponsor.

Clinical Application

Consider PETRUSHKA-style shared decision tools to reduce early antidepressant stopping, especially in primary care; interpret symptom gains cautiously due to missing data and open-label design.