A treatment can look promising on paper and still be difficult to fit into a real life. A person considering an antidepressant might worry about feeling sleepy during a workday, for example, or about a possible change in sexual function. Another person might have different concerns entirely.
Could bringing those concerns into the choice of medication make treatment easier to continue? A 2026 randomized trial offers a cautiously encouraging answer. Adults whose clinicians used a personalized decision-support tool were less likely to have stopped their initially prescribed antidepressant after eight weeks than adults receiving usual care.
The important wrinkle is that the tool did more than ask about preferences. It also used personal and clinical information to predict how different medicines might work. The trial cannot tell us whether discussing side effects alone made the difference.
A choice shaped around a person#
The researchers studied a web-based tool called PETRUSHKA, designed to support a conversation between a person with major depressive disorder and their clinician. Rather than naming one universally best antidepressant, it combined predictions about different medicines with the person's preferences about which side effects they most wanted to avoid.
The predictions drew on characteristics such as age, symptoms, previous antidepressant use and other medical conditions. The tool produced a ranked shortlist of three medications to inform the final choice. Its developers worked with people who had lived experience of depression.
This was not a machine diagnosing people or guaranteeing a side-effect-free medicine. Predictions are estimates, and the researchers noted that it was not always easy to explain exactly how the tool combined an individual's characteristics and preferences into its recommendations. The version tested also did not include every available antidepressant.

How the researchers tested it#
The trial took place at 47 sites in Brazil, Canada and the United Kingdom. It enrolled adults ages 18 to 74 who had major depressive disorder, had a reason to consider medication and were willing to take an antidepressant on its own. Most participants were treated in primary care.
Researchers randomly assigned 540 participants to either the tool group or usual care. In the usual-care group, clinicians prescribed the medicine they had chosen before learning the person's group assignment. In the tool group, the decision was informed by PETRUSHKA's personalized shortlist.
Random assignment is a strength because it helps balance differences between groups that could otherwise affect the results. After assignment, 20 people were found not to meet the eligibility criteria and were excluded; 18 were in the tool group and two were in usual care. Of the 520 eligible participants, 493 had the information needed for the main eight-week analysis.
Most eligible participants were White, and about 70% were in the United Kingdom. The study did not include people with treatment-resistant depression or those who needed urgent mental health care or hospital admission. Its results may not apply equally to people in those circumstances or to every community.
What happened after eight weeks#
The main question was whether people were still taking the antidepressant prescribed when they joined the study. After eight weeks, 41 of 241 people in the tool group had stopped it, compared with 69 of 252 in usual care. That's 17% versus 27%: a difference of about 10 percentage points, or roughly 10 fewer people stopping for every 100 treated.
The researchers also looked specifically at stopping because of adverse events. That happened to 22 people in the tool group, or 9%, and 39 in usual care, or 16%. This finding is consistent with the idea that tolerability matters, but it still does not isolate the effect of discussing side-effect preferences.
There is a useful distinction hiding inside the word stopped. The researchers counted someone as having discontinued if they switched to a different antidepressant as well as if they stopped medication altogether. Some participants did switch. A change of treatment can be an appropriate decision, not a sign that someone has failed.
Continuing the same prescription is also not the same thing as feeling better. It can reflect several factors, including benefit, side effects and a person's circumstances. To understand symptoms, the researchers used separate questionnaires and assessments.
A less certain longer-term picture#
At 24 weeks, people who completed self-report questionnaires gave lower average scores for depression and anxiety symptoms in the tool group than in usual care. That is worth noting: the person's own account of how they feel matters.
But the longer-term findings need care. Only 129 people in each group completed the 24-week depression questionnaire, roughly half of the 520 eligible participants overall. If the people whose answers were missing differed from those who responded, the comparison could be misleading.
Assessments made by observers who did not know the participants' group assignments did not show a clear difference in depression or anxiety at 24 weeks. Nor was there a clear difference between groups in the proportion who had stopped their original antidepressant by that point. Those findings do not prove the groups were identical; they mean the longer-term advantage remains uncertain.
What the trial can and cannot answer#
Participants and treating clinicians knew whether they were using the tool. That could have influenced expectations, conversations and decisions to continue or change a medicine. For example, someone who volunteered for a trial of personalized treatment might have felt disappointed to receive usual care. The researchers kept symptom assessors and statisticians unaware of group assignments to reduce some sources of bias.
The tool group also received a package of changes: a structured discussion, individual predictions and potentially different medication choices. The study was not designed to separate those ingredients. Perhaps the recommendations were a better fit, perhaps involvement in the decision mattered, or perhaps both contributed.
The University of Oxford sponsored the study, which was funded by the UK's National Institute for Health and Care Research. The researchers reported that the funder and sponsor had no role in the study's design, analysis or publication decisions. Some underlying data used for the tool came from pharmaceutical companies.
The practical takeaway is modest. When discussing an antidepressant with a clinician, it can be useful to mention the possible side effects that would matter most in daily life, along with previous experiences and other health concerns. That conversation is worthwhile even though this trial cannot promise that mentioning preferences alone will make a particular treatment easier to continue.
If a medicine is difficult to take or seems unsuitable, discussing it with the prescribing clinician can help clarify the options. Staying on one medication is not the only measure of good care; finding an approach that fits the person is the larger goal.

