The clinical course of first-episode psychosis (FEP) is highly variable, ranging from full recovery to chronic illness. Understanding early predictors of long-term outcomes is crucial for improving treatment. While early-course symptoms significantly impact prognosis, the long-term effects of early treatment response remain understudied. Previous research often included multi-episode patients, potentially skewing the understanding of prognosis. Clinical recovery (CR) is typically defined by remission and functional improvement, with prevalence rates varying widely across studies. Treatment resistance (TR) is defined by a lack of response to adequate antipsychotic (AP) trials, leading to poor outcomes and high healthcare costs. Early intervention for TR may improve long-term outcomes. Previous studies identified several potential predictors of poor outcomes, including younger age at onset, poor premorbid adjustment, longer duration of untreated psychosis (DUP), and baseline symptom severity. However, comprehensive, long-term follow-up studies of unselected FEP samples are lacking. This study aimed to investigate the prevalence of CR and TR at 10 years and identify predictors of these long-term outcomes in a comprehensive FEP sample.

The literature on long-term outcomes in FEP reveals considerable variability, highlighting the need for early identification of patients with differing treatment needs. While the first 2-3 years after illness onset are crucial, the long-term effects of early treatment response remain under-researched. Existing studies show inconsistent CR rates (14-38%), partly due to varying definitions and methodologies. Meta-analyses have failed to identify consistent sociodemographic or clinical predictors of CR. Similarly, studies on TR in FEP report varying prevalence rates (approximately one-third), and the clinical utility of the TRRIP criteria in everyday practice is unclear. Previous research has suggested several predictors of poor outcomes, but their specificity and clinical usefulness remain limited, suggesting distinct predictors may exist for different outcome dimensions.

This prospective longitudinal observational study used data from the Thematically Organized Psychosis Research study (TOP) at the Norwegian Centre for Mental Disorders Research (NORMENT). Participants (18-65 years) meeting DSM-IV criteria for schizophrenia spectrum disorders were recruited within their first year of treatment. Exclusion criteria included substance-induced psychotic disorders, severe brain injury, non-Scandinavian language speakers, and significant intellectual impairment. A total of 102 participants completed 10-year follow-up assessments (retention rate ~35% of living participants). Clinical assessments at baseline, one-year, and 10-year follow-up included the SCID-I for diagnosis and the SCI-PANSS for symptom assessment (using the Wallwork five-factor model). Premorbid adjustment was measured with the PAS, and DUP was calculated in weeks. Mode of onset was categorized as acute or insidious. Functional assessment used the GFS. Outcome groups (CR, TR, heterogeneous middle group [HM]) were defined using TRRIP and RSWG criteria. Statistical analyses included Chi-Square tests, ANOVAs, Kruskal-Wallis tests, and logistic regression analyses to identify predictors of CR and TR.

At the 10-year follow-up, 29 (28%) participants were classified as CR, 32 (31%) as TR, and 41 (40%) as HM. Statistically significant differences between outcome groups were found at baseline for premorbid academic adjustment, DUP, age, global functioning, relationship status, negative and disorganized symptoms, and alcohol use. Multivariate logistic regression analysis identified insidious onset (OR=4.16), baseline disorganized symptoms (OR=2.96), and baseline schizophrenia diagnosis (OR=6.04) as independent predictors of TR. Acute onset (OR=3.40), good premorbid academic adjustment (OR=1.60), and not having a baseline schizophrenia diagnosis (OR=5.02) predicted CR. Only 7% of participants had tried clozapine at 10 years, all of whom were classified as TR. Subsample analysis of participants with one-year data showed that 50% of those categorized as CR and 53% of those categorized as TR at 10 years were already classified as such at one year.

This study confirms the heterogeneity of long-term outcomes in FEP, with considerable variation in CR and TR rates. The identified predictors (insidious onset, disorganized symptoms for TR; acute onset, good premorbid adjustment for CR) are readily identifiable in clinical practice and may aid in early risk stratification. The strong association between baseline schizophrenia diagnosis and TR highlights the need to investigate specific aspects of this diagnosis contributing to treatment resistance. The low rate of clozapine use is concerning, suggesting underutilization despite potential benefits, particularly within a potential critical time window. The heterogeneous middle group emphasizes the complexity of FEP outcomes and the need for diverse treatment approaches beyond those tested in this study. Findings also suggest potential limitations of the TRRIP criteria in everyday clinical practice.

This study underscores the importance of early-course parameters in predicting long-term outcomes in FEP. Insidious onset, baseline disorganized symptoms, premorbid adjustment, and baseline diagnosis are significant predictors. These readily identifiable characteristics may serve as early warning signs for poor outcomes. However, the findings also highlight the need for improved adherence to treatment guidelines and the exploration of broader treatment options for patients not achieving CR or TR status.

The study's main limitation is the relatively high attrition rate (approximately 65%), although analyses showed no significant differences in demographic or baseline clinical characteristics between completers and non-completers. Further, data on health service utilization for all initial participants was collected to mitigate attrition bias. The follow-up period partly overlapped with the COVID-19 pandemic, which may have affected the results. The use of DSM-IV criteria may also limit the generalizability to studies using later DSM versions. Finally, the study is limited by its observational design and cannot establish causal relationships.