Subclinical and attenuated negative symptoms in the prediction of psychosis in children and adolescents: sample composition matters.

In children and adolescents, subjective and attenuated negative symptoms were hypothesized to best differentiate clinical-high risk for psychosis (CHR-P) from other conditions, and to be most predictive of psychosis. To examine this, we analysed data from 8-17-year-old CHR-P in-/outpatients (n = 183), and inpatients not clinically suspected of developing psychosis (n = 277) based on cross-sectional and two-year naturalistic follow-up data from a Swiss-German study. CHR-P criteria and other symptoms were assessed using the Schizophrenia Proneness Instrument, Child & Youth version, and the Structured Interview for Psychosis-Risk Syndromes. Pairwise comparisons, two-step cluster analyses and Cox regression models were applied to identify cluster profiles and the best predictors of psychosis. In pairwise comparisons, cognitive and perceptive basic symptoms, and attenuated positive symptoms better distinguished CHR-P patients from non-CHR-P patients and converters from non-converters than the negative symptom subscales "Adynamia" and "Negative symptoms. In the stable two-cluster solution of patients scoring high (n = 338) and low (n = 122) on all scales, "Adynamia" was most important (importance = 1.0) whereby the high-scorer cluster included most of CHR-P (86.9%) and converters (84.6%), and was predictive of psychosis. Cox models of subscales differed in the selected predictors depending on the analysed samples (total sample, CHR-P and ultra-high risk only (UHR)), whereby the earlier report of "Negative symptoms" predicting conversion in UHR was replicated when only SIPS subscales were considered. While our findings partially support the significance of subjective and attenuated negative symptoms in CHR-P and psychosis prediction, they also suggest that the role of negative symptoms is highly dependent on sample composition.