Individualized treatment rule improves treatment success rate of first-episode schizophrenia

An individualized treatment rule may increase the treatment success rate of patients with first-episode schizophrenia, according to study results published in JAMA Network Open. However, researchers noted the necessity of a clinical trial and replication of these results with an expanded predictor set before clinical implementation.

“Numerous clinical trials and observational studies have investigated comparative treatment effects and risk profiles of antipsychotic drugs,” Chi-Shin Wu, MD, PhD, of the department of psychiatry at National Taiwan University, and colleagues wrote. “Others have searched for reliable prognostic predictors of antipsychotic treatment response (ie, factors associated with treatment response regardless of treatment type) to help guide the decisions about adjunctive psychosocial treatment and psychoeducation, while a small number of recent studies have searched for reliable prescriptive predictors of antipsychotic treatment response (ie, predictors of which antipsychotic medications are best for which patients).”

The researchers emphasized that current prescriptive predictors have been too weak to be of individual clinical value, which has led to a focus on composite models to develop individualized treatment rules. None of the rules proposed thus far have been practical, meaning treatment decisions rely on considerations of trial and error investigations of tolerability, as well as pharmacodynamics and pharmacokinetics, they wrote.

In the present study, Wu and colleagues developed a preliminary individualized treatment rule for patients with first-episode schizophrenia. To do so, they collected data of 32,277 patients in this population who were included in Taiwan’s National Health Insurance Research Database. Specifically, they looked at patients with prescribed antipsychotic medications, ambulatory claims or discharge diagnoses of a schizophrenic disorder. They developed an individualized treatment rule by applying a targeted minimum loss-based ensemble machine learning method to predict treatment success from demographic and baseline clinical data in a 70% training sample, and then validated the model in the remaining 30% of the sample. For each medication for each patient under the model, the researchers estimated treatment success probability, with treatment success defined as not switching medication and not being hospitalized for 12 months.

Treatment success rate in the validation sample was 51.7% under the individualized treatment rule and was 44.5% in the observed population. If all patients were given a prescription for one medication, estimated treatment success was significantly lower for all 13 medications than under the individualized treatment rule. The rule most often recommended aripiprazole (n = 3,088) and amisulpride (n = 2,920), but only 1,054 patients (10.9%) received medications recommended by the rule. Although lower than the success rate under the individualized treatment rule, the observed treatment success rate was significantly higher than if medications had been randomized, the researchers noted.

“The finding that aripiprazole and amisulpride were the most commonly [individualized treatment rule]-recommended medications confirmed the broad clinical plausibility of the [rule], given that a network meta-analysis of clinical trials has shown that amisulpride is among the most effective agents for reducing symptom severity and has one of the lowest all-cause discontinuation rates and that aripiprazole is one of the best tolerated drugs overall,” they wrote. “We also found that estimated increases in treatment success rates associated with nonrecommended prescribing of aripiprazole and amisulpride were comparatively small, suggesting that even when prescribed but not recommended, these medications’ treatment success rates were close to those of recommended medications.” – by Joe Gramigna

Disclosures: Wu reports grants from Taiwan Ministry of Science and Technology. Please see the study for all other authors’ relevant financial disclosures.