Decision Tool Reduces Antidepressant Discontinuation in Depression

The Trial-and-Error Challenge of Antidepressant Selection

Prescribing antidepressants for major depressive disorder frequently involves a frustrating process of trial and error, leading many patients to abandon therapy due to intolerable side effects or a lack of efficacy. To address this challenge, researchers have increasingly explored pharmacogenetic tests and artificial intelligence-driven clinical decision-support systems (software algorithms that analyze patient data to recommend specific treatments) to help personalize medication selection and improve remission rates [1, 2]. While early iterations of these digital tools have demonstrated an ability to accelerate symptom improvement in outpatient settings [3], matching the right patient to the right drug on the first attempt remains a complex clinical hurdle. Additionally, the risk of adverse events and challenging withdrawal symptoms further complicates long-term medication adherence [4, 5]. A newly published multicenter randomized clinical trial evaluating a web-based personalization tool called PETRUSHKA now offers fresh insights into using data-driven methods to keep patients on their prescribed therapies and achieve long-term symptom control [6].

Deploying the PETRUSHKA Tool in Clinical Practice

Patients receiving antidepressants for moderate to severe major depressive disorder often discontinue their medication prematurely because the initially prescribed drug is not the most appropriate clinical match. To address this gap and evaluate the efficacy of a web-based tool designed to personalize antidepressant treatment, researchers conducted a multicenter, randomized clinical trial (registry identifier NCT05608330). The trial included individuals between the ages of 18 and 74 years diagnosed with major depressive disorder across 47 sites in Brazil, Canada, and the United Kingdom. Between November 29, 2022, and January 15, 2025, a total of 540 participants were randomized in a 1:1 ratio. Specifically, 271 participants were assigned to an evidence-based clinical decision-support system called the PETRUSHKA tool, while the remaining 269 participants were assigned to usual care.

Following randomization, 520 individuals were deemed eligible, and 493 were included in the primary analysis. This primary analysis cohort had a median age of 35 years (interquartile range, 25 to 48 years) and was 58% female. Baseline symptom severity indicated a highly symptomatic population. The cohort presented with a baseline 9-item Patient Health Questionnaire (PHQ-9) mean score of 16.6 (standard deviation, 5.1), reflecting moderately severe to severe depression, alongside a 7-item Generalized Anxiety Disorder (GAD-7) mean score of 11.5 (standard deviation, 4.1), indicating moderate baseline anxiety. For practicing physicians, this demographic represents the typical, complex outpatient presentation where initial treatment selection is critical for long-term success.

Reducing Premature Discontinuation at 8 Weeks

The researchers established the primary outcome as treatment discontinuation due to any cause at 8 weeks, a critical window for establishing medication tolerance and adherence in psychiatric care. The findings demonstrated that at 8 weeks, 41 of 241 participants (17%) in the PETRUSHKA group discontinued the prescribed antidepressant due to any cause, compared to 69 of 252 participants (27%) in the usual care group. This difference yielded an adjusted relative risk for all-cause discontinuation at 8 weeks of 0.62 (95% CI, 0.44 to 0.88; P = .007) for the PETRUSHKA group compared to usual care.

Beyond overall adherence, the secondary outcomes included treatment discontinuation due to adverse events. Looking specifically at the early treatment phase, at 8 weeks, 22 of 241 participants (9%) in the PETRUSHKA group discontinued the prescribed antidepressant due to adverse events, which was significantly lower than the 39 of 252 participants (16%) in the usual care group. The analysis showed an adjusted relative risk for discontinuation due to adverse events at 8 weeks of 0.59 (95% CI, 0.36 to 0.97; P = .04) for the PETRUSHKA group. For clinicians managing major depressive disorder, these data highlight the value of personalized prescribing in the vulnerable early weeks of therapy. Ultimately, compared with usual care, use of the PETRUSHKA tool increased the number of patients still taking their antidepressant at 8 weeks. By reducing the likelihood of early dropout driven by side effects or lack of efficacy, the decision-support system helps bridge the gap between initial prescription and long-term symptom remission.

Long-Term Symptom Control and Study Limitations

To evaluate the long-term clinical efficacy of the decision-support tool, the researchers assessed secondary outcomes that included changes in depressive symptoms measured with the PHQ-9, a standard clinical scale ranging from 0 to 27 where higher scores indicate more severe depression. At 24 weeks, the mean PHQ-9 score was 7.1 (SD, 5.4) in the PETRUSHKA group (n = 129) versus 9.2 (SD, 6.5) in the usual care group (n = 129). This sustained reduction in symptom severity resulted in an adjusted between-group mean difference for PHQ-9 scores at 24 weeks of -1.92 (95% CI, -3.06 to -0.78; P < .001). For practicing physicians, this indicates that personalizing the initial antidepressant choice not only keeps patients on therapy early on but also translates into measurable improvements in mood over a six-month period.

Because anxiety frequently complicates the management of major depressive disorder, the secondary outcomes also included changes in anxiety symptoms measured with the GAD-7 questionnaire, which ranges from 0 to 21. At 24 weeks, the mean GAD-7 score was 4.6 (SD, 4.1) in the PETRUSHKA group (n = 133) versus 5.8 (SD, 4.9) in the usual care group (n = 126). The analysis demonstrated an adjusted between-group mean difference for GAD-7 scores at 24 weeks of -1.39 (95% CI, -2.26 to -0.52; P = .002). Despite these clinical benefits, the authors caution that the validity of the results is limited by the lack of a double-blind design and a large amount of missing data. Physicians should weigh these methodological constraints when considering the integration of such web-based decision tools into routine psychiatric care, recognizing that while the data support improved long-term symptom control, real-world adherence and reporting gaps remain a challenge.

References

1. Bousman CA, Arandjelovic K, Mancuso SG, Eyre HA, Dunlop BW. Pharmacogenetic tests and depressive symptom remission: a meta-analysis of randomized controlled trials.. Pharmacogenomics. 2019. doi:10.2217/pgs-2018-0142

2. Benrimoh D, Whitmore K, Richard M, et al. Artificial Intelligence in Depression-Medication Enhancement (AID-ME): A Cluster Randomized Trial of a Deep-Learning-Enabled Clinical Decision Support System for Personalized Depression Treatment Selection and Management.. Journal of Clinical Psychiatry. 2025. doi:10.4088/jcp.24m15634

3. Benrimoh D, Whitmore K, Richard M, et al. Artificial Intelligence in Depression-Medication Enhancement (AID-ME): A Cluster Randomized Trial of a Deep-Learning-Enabled Clinical Decision Support System for Personalized Depression Treatment Selection and Management.. The Journal of clinical psychiatry. 2025. doi:10.4088/JCP.24m15634

4. Fergusson D, Doucette S, Glass KC, et al. Association between suicide attempts and selective serotonin reuptake inhibitors: systematic review of randomised controlled trials. BMJ. 2005. doi:10.1136/bmj.330.7488.396

5. Davies J, Read J. A systematic review into the incidence, severity and duration of antidepressant withdrawal effects: Are guidelines evidence-based?. Addictive Behaviors. 2018. doi:10.1016/j.addbeh.2018.08.027

6. 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. doi:10.1001/jama.2026.1327