Risperidone, which is available free of charge in the public health system, is usually the first treatment option for psychosis (image: Miguel Á. Padriñán/Pixabay)
Published on 08/05/2025
By Emilio Sant’Anna | Agência FAPESP – In Brazil, risperidone is typically the first treatment option for psychosis because the drug is free in the public health system. However, a significant proportion of patients do not respond to the therapy, which may be related to clinical and environmental factors, as well as the patient’s genetic profile.
In a study supported by FAPESP and published in the Brazilian Journal of Psychiatry, researchers from the Federal University of São Paulo (UNIFESP) and their collaborators evaluated the effectiveness of different models in predicting which individuals affected by the disorder could benefit from the use of the drug.
The clinical trial included 141 patients experiencing their first psychotic episode who had never taken antipsychotics. The patients were evaluated before and after ten weeks of risperidone treatment. The results showed that 72 patients (51%) responded well to the treatment, while 69 patients showed no significant improvement in symptoms.
The researchers used machine learning algorithms (artificial intelligence) to predict patients’ level of response to treatment. They created different models using three categories of data: clinical data alone, genetic data alone, and a combination of the two (hybrid models). The latter performed best, correctly predicting responses in 72.9% of cases. Models that used only clinical data achieved a success rate of 63.3%.
According to Síntia Belangero, a geneticist, professor at UNIFESP, and one of the authors of the article, these results reinforce the potential of precision psychiatry, integrating genetic and clinical information to optimize the treatment of patients with psychosis.
“Because it’s a disease with a complex etiology and genetic and environmental effects, ideally, we should take into account both genetic and clinical-environmental factors to improve the prediction of clinical outcomes, response to treatment, and, who knows, be able to carry out prevention at least at a secondary level [when the disease is already established and the goal is to prevent it from worsening],” says Belangero.
Among the clinical factors analyzed, duration of untreated psychosis (DUP), the period between the onset of psychotic symptoms and treatment with medication, was the most relevant for predicting therapeutic response. A previous study by the same group had already shown that a longer DUP resulted in worse symptoms and poorer overall functioning in psychotic patients.
“DUP is associated with worse outcomes after treatment, but doesn’t change the baseline clinical profile of patients, which reinforces the importance of early intervention,” emphasizes the researcher.
Another environmental factor that proved important in hybrid models was cannabis use. “Like DUP, it’s a modifiable factor that can alter the course of progression and outcome. Although controversial, other studies have shown that its use is related to a higher number of hospital admissions and impaired treatment, as patients who use cannabis are at greater risk of non-adherence to pharmacological treatment,” she explains.
In addition to working with thousands of genetic variants and trying to integrate them with other clinical variables, the researchers often encountered complicated situations. “One of the main challenges was accessing patients and their families at such a critical time as the first psychotic episode, still in the emergency room and before the start of medication,” says the researcher.
Belangero points out that the research results may guide the development of new treatment methods in the future. Doctors need to analyze several variables, including which antidepressant or antipsychotic to choose, its safety and efficacy, and how the drug will be metabolized.
“In theory, these problems will be solved through precision psychiatry, which will allow us to know, from the first consultation, which treatments are most beneficial and tolerable for each individual. But we aren’t yet at the point of effective precision psychiatry; we’re only moving toward it,” says the UNIFESP researcher.
The team, which includes researchers Giovany Oliveira, Vanessa Ota, Ary Gadelha, Cristiano Noto, and Diego Mazzotti, now intends to validate these predictive models in larger samples and with groups with different genetic ancestries. “Our goal is to test these predictors in larger populations of different ethnicities to verify the robustness and applicability of the results,” says Belangero.
The article “A hybrid model for predicting response to risperidone after first episode of psychosis” is available at www.bjp.org.br/details/3541/en-US/a-hybrid-model-for-predicting-response-to-risperidone-after-first-episode-of-psychosis.
Source: https://agencia.fapesp.br/55518
