Probabilistic reinforcement learning impairment in patients with first episode Schizophrenia

Abstract

BACKGROUND: Accumulating evidence indicates that schizophrenia patients exhibited reinforcement learning (RL) impairment. Nonetheless, most previous studies focused on chronically ill patients who had longstanding exposure to antipsychotic medication treatment which may interfere with RL. In this study, we aimed to investigate both rapid and gradual RL in patients with first-episode schizophrenia-spectrum disorder. METHODS: A total of 31 patients with first-episode DSM-IV schizophrenia, schizophreniform disorder or schizoaffective disorder (treated with antipsychotic for 3-6 months) and 33 healthy controls matched with age, sex, and educational level were recruited. Each subject completed 2 computerised probabilistic RL experiments, namely Go/NoGo (GNG) and Gain vs. Loss-Avoidance (GLA) tasks which have previously been examined in chronic schizophrenia patients. A battery of cognitive assessments was administered to patients and controls. Patients’ symptom severity was assessed with PANSS (Positive and Negative Symptom Scale), SANS (Scale for the Assessment of Negative Symptoms), and CDS (Calgary Depression Scale). Data on antipsychotic medication were obtained. Between-group differences on rapid reversal learning, and gradual positive and negative RL measures were investigated. Correlation analyses were conducted to examine relationships between RL parameters and symptom dimensions. RESULTS: There were no significant between-group differences in age, sex, and educational level. Controls had better overall cognitive functions than patients. In GNG task, patients exhibited significantly longer Go-response latency and lower accuracy toward positive stimuli than controls, indicating poorer gradual positive RL. In GLA task, patients displayed significantly lower accuracy toward both gain-seeking (positive reinforcement) and loss-avoidance (negative reinforcement) stimulus-pairs than controls. Patients performed worse than controls in rapid reversal learning as evidenced by their significantly lower win-stay rate. Negative symptom level was correlated with rapid reversal and gradual positive RL measures. There were no significant correlations between any of the RL measures and positive symptom rating, CDS score, cognitive composite score, and chlorpromazine equivalent. CONCLUSION: Our findings indicated that both rapid and gradual RL deficits (especially positive reinforcement dysfunction) were found in the initial stage of schizophrenia and was associated with negative symptoms. Further prospective research is required to clarify the longitudinal course of RL impairment and its relationship with symptom and functional outcomes.