Cerebello-Thalamo-Cortical Hyperconnectivity Classifies Patients and Predicts Long-Term Treatment Outcome in First-Episode Schizophrenia

Abstract

It has previously been shown that cerebello-thalamo-cortical (CTC) hyperconnectivity is likely a state-independent neural signature for psychosis. However, the potential clinical utility of this change has not yet been evaluated. Here, using fMRI and clinical data acquired from 214 untreated first-episode patients with schizophrenia (62 of whom were clinically followed-up at least once at the 12th and 24th months after treatment initiation) and 179 healthy controls, we investigated whether CTC hyperconnectivity would serve as an individualized biomarker for diagnostic classification and prediction of long-term treatment outcome. Cross-validated LASSO regression was conducted to estimate the accuracy of baseline CTC connectivity for patient-control classification, with the generalizability of classification performance tested in an independent sample including 42 untreated first-episode patients and 65 controls. Associations between baseline CTC connectivity and clinical outcomes were evaluated using linear mixed model and leave-one-out cross validation. We found significantly increased baseline CTC connectivity in patients (P = .01), which remained stable after treatment. Measures of CTC connectivity discriminated patients from controls with moderate classification accuracy (AUC = 0.68, P < .001), and the classification model had good generalizability in the independent sample (AUC = 0.70, P < .001). Higher CTC connectivity at baseline significantly predicted poorer long-term symptom reduction in negative symptoms (R = 0.31, P = .01) but not positive or general symptoms. These findings provide initial evidence for the putative "CTC hyperconnectivity" anomaly as an individualized diagnostic and prognostic biomarker for schizophrenia, and highlight the potential of this measure in precision psychiatry.

Keywords: cerebellum; diagnostic biomarker; functional connectivity; prognostic biomarker/schizophrenia; thalamus.

Fig. 1.

Mean CTC connectivity at baseline and follow-ups in the main sample. (A) Significantly higher connectivity was observed in untreated first-episode patients with schizophrenia compared with healthy controls at baseline. The error bars indicate standard error. (B) Individual trajectories and group trajectory for CTC connectivity during follow-ups. No significant change was found for CTC connectivity after treatment.

Fig. 2.

Classification accuracy of CTC connectivity in discriminating untreated first-episode patients from controls. (A) The selected CTC connections showing highest predictability for patients from cross-validated LASSO regression in the main dataset. See supplementary table S3 for details of these connections. (B) The receiver operating characteristic (ROC) curves for the main and generalization datasets.

Fig. 3.

Long-term clinical outcome after treatment in the follow-up sample of the main dataset. (A) Individual trajectories and group trajectories for positive symptoms, negative symptoms, and general symptoms. The slope coefficients for all symptoms were highly significant. (B) The mean CTC connectivity at baseline significantly predicted changes of negative symptoms during follow-up, where patients with higher baseline CTC connectivity had worse outcome. (C) The predicted slope coefficients of negative symptoms were significantly correlated with the actual coefficients, as revealed by leave-one-out cross validation.