Risk calculators for prediction of conversion of Clinical High-Risk (CHR) individuals to syndromal psychosis have recently been developed and have generated considerable clinical use and research interest. Predictor variables in these calculators have been clinical rather than biological, and our goal was to incorporate a neurochemical imaging measure into this framework and assess its impact on prediction. We combined striatal glutamate 1H MRS data with the SIPS symptoms identified by the Columbia Risk Calculator as having the greatest predictive value in order to develop an imaging-based risk calculator for conversion to psychosis. We evaluated the calculator in 19 CHR individuals, 7 (36.84%) of whom converted to syndromal psychosis during the 2-year follow up. The receiver operating characteristic (ROC) curve for the logistic model including only striatal glutamate and visual perceptual abnormalities showed an AUC = 0.869 (95% CI = [0.667, 1.000]) and AUCoa = 0.823, with sensitivity of 0.714, specificity of 0.917, positive predictive value of 0.833, and negative predictive value of 0.846. These results represent modest improvements over each of the individual ROC curves based on either striatal glutamate or visual perceptual abnormalities alone. The preliminary model building and evaluation presented here in a small CHR sample suggests that the approach of incorporating predictive imaging measures into risk classification is not only feasible but offers the potential of enhancing risk assessment.
Keywords: Clinical high-risk; Conversion to psychosis; Glutamate; Risk calculator.
Copyright © 2019 Elsevier B.V. All rights reserved.