Deciphering reward-based decision-making in schizophrenia : a meta-analysis and behavioral modeling of the Iowa Gambling Task

Background: Patients with schizophrenia (SZP) have been reported to exhibit impairments in reward-based decision-making, but results are heterogeneous with multiple potential confounds such as age, intelligence level, clinical symptoms or medication, making it difficult to evaluate the robustness of these impairments. Methods: We conducted a meta-analysis of studies comparing the performance of SZP and healthy controls (HC) in the Iowa Gambling Task (IGT) as well as comprehensive analyses based on subject-level data (n = 303 SZP, n = 188 HC) to investigate reward-based decision-making in SZP. To quantify differences in the influence of individual deck features (immediate gain, gain frequency, net loss) between SZP and HC, we additionally employed a least-squares model. Results: SZP showed statistically significant suboptimal decisions as indicated by disadvantageous deck choices (d from 0.51 to 120.62) and lower net scores (d from 120.35 to 121.03) in a meta-analysis of k = 29 samples (n = 1127 SZP, n = 1149 HC) and these results were confirmed in a complementary subject-level analysis. Moreover, decision-making in SZP was characterized by a relative overweighting of immediate gain and net losses and an underweighting of gain frequency. Moderator analyses revealed that in part, decision-making in the IGT was moderated by intelligence level, medication and general symptom scores. Conclusion: Our results indicate robust impairments in reward-based decision-making in SZP and suggest that decreased cognitive resources, such as working memory, may contribute to these alterations