Normalization of semantic categorization deficit in first-episode schizophrenia patients following symptomatic recovery: a three-year prospective longitudinal study

Abstract

BACKGROUND: Semantic memory dysfunction has been suggested in schizophrenia using the categorization task. However, data were largely inconsistent and the longitudinal trajectory of the deficit was unknown. Current study aimed to explore the 3-year longitudinal course of semantic memory deficits in first-episode schizophrenia patients. METHODS: First-episode schizophrenia patients (DSM-IV) were assessed prospectively at the point of first contact (A1), after clinical stabilization (A2), and each year for the following 3 years (Y1, Y2, Y3) using the categorization task. Patients are required to make decision (yes or no) on whether a 'word' belonged to a 'category'. 40 pairs of words were divided into 4 categories: fruits, furniture, drinks and clothing. Each category (e.g., furniture) was subdivided into five degrees of semantic relatedness: (1) typical word of the category (e.g., table), (2) atypical of the category (e.g., bookcase), (3) borderline word (e.g., clock), (4) related but outside the category (e.g., painting) and (5) unrelated and outside the category (e.g., sun). Data on both reaction time and proportion of yes response were analyzed. Normal participants were assessed once. RESULTS: 37 first-episode schizophrenia patients and 37 normal were recruited (matched for gender, age and education). Five ANOVAs were carried out to detect the difference in the five semantic relatedness conditions between patients and control at each of the 5 timepoints. In the first ANOVA which compared patients at A1 and control, reaction times in all conditions were slower in patients compared with control (F(1, 72)=7.83, p=.007). Significant main effect of semantic relatedness condition (F(4, 288)=12.30, p<.001) and interaction effect were also found (F(4, 288)=4.88, p=.001). Post-hoc pairwise comparisons found the two groups were different with regard to typical (p=.006), related (p=.036), and unrelated condition (p<.001). Interestingly, the remaining ANOVAs produced the same results. Significant main effect of condition (p=.015, p<.001, p<.001, p<.001, respectively) was observed in patients at A2, Y1, Y2 and Y3 as compared with control. Besides, pairwise comparisons had suggested significant difference in all conditions (all with p<.001). Likewise, ANOVA was carried out in the yes response data and similar results were identified. Main effect of group was found at A1 only (F(1, 72)=7.10, P=.009) but not the other timepoints. DISCUSSION: The data clearly show semantic memory abnormalities (slower reaction time and more error, i.e., more yes response in the outside the category condition) in first-episode schizophrenia as compared with normal. Intriguingly, they largely normalized following symptomatic recovery and remained stable for up to the first three years of the disorder, suggesting a state effect rather than a trait effect.