ABSTRACT
OUTCOME STUDIES IN FIRST-EPISODE SCHIZOPHRENIA The past two decades have witnessed the emergence of prospective follow-up studies focusing upon first-episode psychosis (8-10) and first-episode schizophrenia cohorts (11-15). These studies provide the opportunity to examine the characteristics of psychotic disorders and to assess prognosis among a cohort, at the same phase of illness, free from the confounding effects of previous treatment interventions or secondary disability (16, 17). However, despite the obvious advantages of this approach, few large, prospective, longitudinal studies of illness course and outcome in first-episode psychosis have been conducted, reflecting the complexities and challenges of recruiting and retaining such a cohort. Studies to date have typically involved either large but heterogeneous, nonrepresentative samples (15, 18, 19) or, conversely, comparatively smaller cohorts that approximate epidemiological samples (8, 20, 21). Other
limitations of first-episode follow-up studies are lack of standardized sampling methods, entry criteria and information regarding the completeness and representativeness of recruited samples. Few studies have conducted longitudinal assessments beyond 24 months; these have focused on patients with schizophrenia (13-15, 22-25). This body of work provides a resource for examining the characteristics of first-episode psychosis and schizophrenia, particularly when conducted in representative samples. However, in a recent review of the methodology, in particular the sampling procedures, of longitudinal first-episode studies published from 1996 (26), all but two studies were confidently considered representative for first-episode psychosis (8, 27). Sampling biases falling into three categories were observed: (1) the absence of data collected or reported on individuals who refused to participate, and comparisons of this group with study participants (28-38); (2) limited diagnostic range with a sampling bias toward nonaffective psychosis (schizophrenia spectrum disorder) (31-34, 36, 37); and (3) convenience samples with inevitable biases, for example, inpatients only, tertiary referrals to private hospitals, mixed first-and multiepisode samples, male predominance, and exclusion of participants without family involvement (18, 30, 34, 35, 37-40). The findings are summarized in Table 5.1. While this body of research has sought to elucidate the correlates of outcome in psychosis, the limitations of nonrepresentative first-episode psychosis studies with sampling biases, small samples, and short follow-up periods impede generalizability and validity of the results.
