Antidepressant Treatment Personalized by Preference, Not Pharmacogenomics

A web-based clinical decision support system facilitated selection of a well tolerated, and possibly more effective antidepressant for primary care patients with major depressive disorder (MDD) in a recent clinical trial comparing the system with treatment as usual (TAU) prescribing.¹

"As part of the Personalizing Antidepressant Treatment for Unipolar Depression Combining Individual Choices, Risks and Big Data (PETRUSHKA) project, we developed an online, evidence-based, decision-support system for patients and clinicians to enable personalized antidepressant treatment decision-making for patients with MDD treated in routine clinical practice," shared Andrea Cipriani, MD, PhD, Department of Psychiatry, Warneford Hospital, University of Oxford, Oxford, UK, and colleagues.

The PETRUSHKA tool, they explain, utilizes statistical and machine learning methods to predict antidepressant effectiveness for a patient based on demographics, medical history, and presentation; and further narrows the choice of agents by accounting for the patient's preferred adverse effect profile.

This approach diverges from efforts to personalize antidepressant treatment based on biomarkers, observed Gregory Simon, MD, MPH, of the Kaiser Permanente Washington Health Research Institute, in Seattle, WA, in an accompanying editorial.² "Early research to personalize depression treatment yielded decades of disappointment,” he remarked.

"Subtyping depression based on clinical characteristics did not predict patient response to specific medications, reflecting both the heterogeneity of depression and the wide range of clinical and nonclinical factors affecting outcomes. Pharmacogenomic, imaging, and other biomarker research failed to identify reproducible predictors of specific medication responses," Simon recounted.

Comparing PETRUSHKA With TAU Prescribing

Cipriani et al recruited 520 adults diagnosed with MDD at 47 sites in 3 countries to participate in the trial; with the diagnosis confirmed on the Clinical Interview Schedule, revised version. Each participant had an indication for pharmacological treatment and consented to receive an antidepressant as monotherapy. Participants were randomly assigned 1:1 to either an antidepressant selected with support of the PETRUSHKA tool or TAU prescribing.

The primary outcome was discontinuation of the antidepressant for any cause by 8 weeks. Secondary outcomes included discontinuation of treatment due to adverse events; and self-rated change in depressive symptoms on the 9-item Patient Health Questionnaire [PHQ-9]) and of anxiety symptoms on the 7-item Generalized Anxiety Disorder questionnaire [GAD-7]) at 8 and 24 weeks. Other assessments included observer-rated measures of depressive and anxiety symptoms (eg, with the 17-item Hamilton Depression Rating Scale [HAM-D] and Hamilton Anxiety Rating Scale [HAM-A]), suicidality, health-related quality of life, and adverse events.

Of the 520 eligible participants, 493 continued through the trial to be included in the primary analysis. The severity of their baseline depressive symptoms corresponded to PHQ-9 mean score of 16.6 (SD, 5.1), and 16.3 (SD, 5.2) on the HAM-D. Mean baseline severity of anxiety symptoms were 11.5 (SD, 4.1) on the GAD-7, and 18.9 (SD, 7.4) on the HAM-A.

The investigators found that the top choices in the PETRUSHKA group were mirtazapine, escitalopram, and vortioxetine; and that sertraline, citalopram, and fluoxetine were most frequently prescribed in the TAU group. At 8 weeks, 15 in the Petrushka group and 27 in the TAU had switched to another antidepressant.

Cipriani et al noted, however, that treatment discontinuation at 8 weeks due to any cause was not statistically different between the PETRUSHKA and TAU patients treated by psychiatrists. The larger effect in favor of the PETRUSHKA tool was among patients treated in primary care (36/226 [16%] discontinuing the initial antidepressant prescribed with PETRUSHKA vs 66/234 [28%] with TAU).

Simon remarked on the apparent greater tolerability or acceptability of the antidepressant selected with the PETRUSHKA system for patients in primary care, and suggested, as did Cipriani et al, that the system is likely to be of less utility for psychiatrists.

"A psychiatrist has more experience with antidepressant prescribing, especially with newer medications, that may blunt any benefits of personalized medication recommendations," Simon observed.

Simon also suggested that the disparate results may arise in the differences in expectations of patients treated in specialty and primary care settings. "Higher motivation and higher likelihood of continuing an antidepressant medication among patients seeking care from specialists may blunt any benefits of the personalized medication recommendations," he posited.

There were no statistical differences in measures of antidepressant efficacy between the groups at 8 weeks. The investigators found, however, that the PETRUSHKA group evidenced greater improvement in both depressive and anxiety symptoms at 24 weeks. The PHQ-9 mean score at 24 weeks was 7.1 (SD, 5.4) in the PETRUSHKA group vs 9.2 (SD, 6.5) in the TAU group. Similarly, at 24 weeks the GAD-7 anxiety score was mean 4.6 (SD, 4.1) in the PETRUSHKA group vs 5.8 (SD, 4.9) with the TAU group.

Cipriani et al also reported that at 24 weeks, but not 8 weeks, there was "strong evidence" favoring the PETRUSHKA group antidepressant in improving health-related quality of life (assessed on the EuroQol VAS). There were no other differences found between groups on other secondary measures.

Simon pointed out several methodological issues that limit generalizing from the study results, but he supports the approach to selecting an antidepressant, "given that personalization based on patient preference involves little added cost or delay in starting treatment."

Cipriani et al also acknowledge study limitations but suggest that a tool such as PETRUSHKA is aligned with a growing trend toward incorporating the patient perspective into clinical management and health care decision-making, "which may improve care and reduce costs."

Dr Bender reports on medical innovations and advances in practice and edits presentations for news and professional education publications. He previously taught and mentored pharmacy and medical students, and he provided and managed pharmacy care and drug information services.

References

1. Cipriani A, Fernandes KBP, Mulsant BH, et al. A decision-support system to personalize antidepressant treatment in major depressive disorder: a randomized clinical trial. JAMA. 2026;335(14):1219-1231.

2. Simon GE. Picking antidepressants—patient preferences beat biomarkers. JAMA. 2026;335(14):1212-1219.