Long-term clinical recovery and treatment resistance in first-episode psychosis: a 10-year follow-up study

The study addresses the wide variability in long-term outcomes among individuals with first-episode psychosis (FEP), ranging from persistent illness to complete recovery. Early course parameters within the first 1–3 years after onset may critically shape long-term prognosis, yet the impact of early treatment response on long-term outcomes remains insufficiently studied. Prior research often mixes multi-episode cohorts, potentially biasing outcome estimates. Clinical recovery (CR) is commonly defined as sustained symptom remission per RSWG criteria plus functional improvement, with long-term CR rates in FEP reported between 14–38% and a meta-analytic estimate of ~21%. Treatment resistance (TR), per TRRIP consensus, affects about one-third of FEP patients and is associated with poorer clinical and functional outcomes and higher service use. Numerous baseline factors (e.g., younger age at onset, poor premorbid adjustment, longer DUP, higher symptom burden) have been linked to poor outcomes, but few have sufficient specificity for clinical use. This study aims to provide long-term (10-year) prevalence estimates of CR and TR in an unselected FEP cohort and to identify pre-illness, demographic, and early clinical predictors of these outcomes.

The authors review evidence that long-term CR in FEP ranges from 14–38% in long-term studies, with a ~21% pooled estimate across 26 studies. Predictors of CR are inconsistent across literature; large reviews found no robust socio-demographic or clinical predictors, though OPUS reported lower baseline negative symptoms and younger age at first diagnosis predicting CR at 10 years. For TR, TRRIP defines persistent non-remission with functional impairment after two adequate antipsychotic trials or clozapine use; meta-analyses estimate TR in approximately one-third of FEP. TR is linked to poor quality of life, higher healthcare use, and societal costs. Suggested predictors of poor outcomes include younger onset, poorer premorbid adjustment, male gender, longer DUP, higher baseline positive/negative symptoms, substance use, poor initial treatment response, and schizophrenia diagnosis; disorganization has been implicated as a potential predictor of TR. Clozapine is the only evidence-based treatment for TR but is underused and often delayed, with indications of a critical window for initiation. The literature highlights the need for comprehensive, long-term FEP cohorts using consensus outcome definitions to clarify predictors.

Design and setting: Prospective longitudinal observational study using the Thematically Organized Psychosis Research (TOP) cohort at NORMENT, Norway. Recruitment from inpatient and outpatient psychiatric units across major hospitals in Oslo (catchment ~660,000; 88% of Oslo population), 2002–2019. Participants: Aged 18–65; DSM-IV schizophrenia spectrum diagnoses at baseline (schizophrenia, schizophreniform, schizoaffective, psychotic disorder NOS); within first year of treatment; no prior adequate psychosis treatment (no hospitalization for treatment at psychosis unit and no adequate antipsychotic treatment for ≥12 weeks or until remission). Substance-induced psychosis, severe brain injury, non-Scandinavian language, or IQ <70 were excluded. Sample and follow-up: 307 completed baseline; 14 deceased; 102 completed 10-year follow-up (~35% retention of living participants). A subsample (n=56–57) had one-year follow-up data. No significant baseline differences between 10-year completers and non-completers; registry-based specialized care contacts were similar. Ethics: Conducted per Helsinki Declaration; informed consent at baseline and follow-ups; approved by REK (#22265 & #2009-2485). Assessments: - Diagnoses: SCID-I (DSM-IV), modules A–E, at all time points; high inter-rater reliability (kappa 0.92–0.99). - Symptoms: SCI-PANSS; symptom dimensions per Wallwork five-factor model (positive, negative, disorganized, excited, depressive). - Premorbid functioning: Premorbid Adjustment Scale (PAS) social and academic subscales for childhood (0–11 years). - DUP: Weeks from onset of psychosis (PANSS P1, P3, P5, P6, or G9 ≥4 persisting >1 week) to adequate treatment. - Mode of onset: Insidious vs acute (evolution to full psychosis over ≥6 months vs <6 months). - Functioning: Global Functioning Scale (GFS; split version); functional recovery defined as GFS ≥61 at 1-year or 10-year follow-up. Outcome definitions at 10 years: - Treatment resistance (TR): Per TRRIP: PANSS items >3 indicating at least moderate symptom severity plus GFS ≤60 for ≥12 weeks, with at least two adequate antipsychotic trials (each >6 weeks at effective dosage) or clozapine use. In this study, adequate trial operationalized per national guidelines as ≥1 defined daily dose for ≥2 weeks or clozapine. - Clinical recovery (CR): Per RSWG remission for >24 weeks plus functional recovery (GFS ≥61). - Heterogeneous middle (HM): Participants not meeting TR or CR. One-year outcome (subsample): Early TR defined as lack of remission for ≥12 weeks with GFS <61; early CR defined as remission with GFS ≥61 (stability criterion 12 weeks). Treatment data: Extracted from medical charts, blood samples, education/occupation/marital status; antipsychotic trials counted; clozapine use noted; hospitalization durations recorded. Statistical analysis: Data management in secure TSD. Analyses in SPSS v29 and R. Two-tailed alpha 0.05. Normality and variance checks; log-transformation for non-normal variables. Group comparisons (CR, HM, TR) via Chi-square (categorical), ANOVA with Tukey post hoc (continuous), Welch ANOVA where variance unequal, and Kruskal–Wallis for non-parametric data. Multivariate logistic regression (binary preferred over multinomial due to HM heterogeneity) to identify predictors for TR vs no-TR and CR vs no-CR. Checked multicollinearity; selected PAS subscale and PANSS dimension with strongest contributions (disorganized symptoms). Model fit assessed via -2LL and AIC; proportional odds assumptions tested. Sub-analyses restricted to those with baseline schizophrenia diagnosis.

- Outcome prevalence at 10 years (n=102): CR 29 (28–29%); TR 32 (31%); HM 41 (40%). - Baseline group differences (Table 2): CR showed better premorbid academic adjustment than HM and TR; shorter DUP than HM and TR; higher baseline GFS than TR; fewer baseline negative and disorganized symptoms than TR; more often in a relationship than TR. Age at baseline was lower in CR vs HM. - Predictors of TR (10-year) in multivariable logistic regression (Table 4a): • Insidious onset: OR 4.16 (95% CI 1.26–13.76), p=0.020. • Baseline disorganized symptoms: OR 2.96 (95% CI 1.30–6.72), p=0.010. • Baseline schizophrenia diagnosis: OR 6.04 (95% CI 2.04–17.86), p=0.001. Model fit: McFadden pseudo r²=0.295; Nagelkerke r²=0.433. In those with baseline schizophrenia, insidious onset remained predictive of TR (OR 4.65). - Predictors of CR (10-year) in multivariable logistic regression (Table 4b): • Acute onset: OR 3.40 (95% CI 1.22–9.46), p=0.019. • Better premorbid academic adjustment: OR 1.60 (95% CI 1.02–2.52), p=0.043. • No baseline schizophrenia diagnosis: OR 5.02 (95% CI 1.56–16.13), p=0.007. Model fit: McFadden pseudo r²=0.153; Nagelkerke r²=0.295. In the baseline schizophrenia-only subsample, predictors did not remain significant (likely limited power; only 5 met CR). - Heterogeneous middle group characterization: 17% continuously psychotic; 27% met clinical TRRIP criteria but lacked full criteria (insufficient adequate trials); 68% were in RSWG remission at 10 years (of these, 17.5% failed time criterion, 52.5% lacked adequate functioning); 2.5% had adequate functioning without remission. - Treatment and hospitalization (Table 5): • Clozapine use at 10 years: 7/102 (7%); all were in TR (22% of TR group). • Antipsychotic use at 10 years: TR 78% vs CR 48% and HM 57% (p=0.049). • Adequate antipsychotic trials: TR 100% had ≥2 trials; CR 55%; HM 44% (p<0.001). Higher counts of trials in TR (more with 3–4 trials). • TR had significantly longer total hospitalization duration than CR and HM (p≤0.006). - One-year to ten-year stability (Table 6; n=57): 50% of early CR remained CR at 10 years; 53% of early TR remained TR at 10 years; chi-square=10.69, p=0.030. - Additional observations: Clozapine use increased from ~2% at 1 year to ~7% at 10 years; 53% of TR participants and 17% of HM participants who were continuously psychotic did not receive a clozapine trial, indicating underutilization.

Using consensus definitions, the study quantifies long-term outcomes in an unselected FEP cohort and demonstrates that about one-third achieve CR, one-third develop TR, and a substantial middle group has partial or mixed outcomes. Baseline disorganized symptoms and insidious onset independently predict long-term TR, beyond the influence of a baseline schizophrenia diagnosis, underscoring clinically recognizable features that can flag high TR risk early. Conversely, acute onset, better premorbid academic functioning, and absence of baseline schizophrenia diagnosis predict CR, highlighting the importance of premorbid and early-course characteristics. Approximately half of early CR and TR statuses persist at 10 years, supporting the prognostic value of early outcomes. The middle group’s composition shows that some patients fail to receive sufficient adequate antipsychotic trials, suggesting that strict adherence to TRRIP criteria may underestimate TR prevalence in real-world settings and that treatment pathways may not align with guidelines. The underuse and delayed initiation of clozapine likely contribute to poorer long-term outcomes. Functional criteria for recovery may be stringent relative to general population norms, and a need for functionally focused interventions is suggested. Male gender, despite higher proportion in TR, did not emerge as an independent predictor at 10 years, consistent with literature indicating similar long-term functional outcomes by sex. Overall, findings reinforce focusing on early identification and tailored treatment strategies, including timely clozapine and psychosocial interventions, to improve long-term trajectories.

The study identifies three long-term outcome groups in FEP at 10 years—clinical recovery (28–29%), treatment resistance (~31%), and a heterogeneous middle group (~40%)—using consensus criteria. Readily assessable early-course features—mode of onset, baseline disorganized symptoms, premorbid academic adjustment, and baseline diagnosis—predict long-term outcomes and can guide early risk stratification. The results also indicate gaps in adherence to treatment guidelines, particularly underutilization of clozapine, which may hinder optimal outcomes. Future work should: optimize early identification of patients at TR risk; implement guideline-concordant, timely clozapine initiation; evaluate psychosocial and tertiary care models; refine functional recovery criteria; and investigate strategies to improve functional outcomes in those achieving symptomatic remission but with persistent disability.

- Attrition: Only ~35% of living baseline participants completed 10-year follow-up, potentially reducing power and risking selection bias; however, no baseline differences or registry-based care utilization differences were found between completers and non-completers. - Generalizability: Conducted within a Norwegian public health system; findings may not fully generalize to different healthcare contexts. - TRRIP applicability: Requirement of two adequate antipsychotic trials may underestimate TR prevalence in observational cohorts where treatment may be suboptimal or delayed. - Underuse of clozapine: Low clozapine exposure limits assessment of its potential impact on outcomes. - Subgroup power: In analyses restricted to baseline schizophrenia diagnosis, predictors of CR were non-significant likely due to small numbers (only 5 met CR). - Observational design: Residual confounding cannot be excluded; medication adherence and detailed psychosocial treatment exposure were not fully quantified.