Preoperative magnetic resonance imaging-based nomogram for predicting overall survival in patients with glioma: a...

Glioma prognostication relies on histopathological and molecular assessment requiring tissue sampling. Accessible tools for noninvasive preoperative prognostication remain limited.

This multicenter retrospective study included 1992 patients (1207 in the training dataset [TD] and 785 in the external validation dataset [EVD]). A preoperative clinicoradiological nomogram (PCR nomogram) was developed using Cox regression with age and visual magnetic resonance imaging features. A histopathology-based model incorporated tumor type and grade according to the 2021 World Health Organization (WHO) classification, and a clinicoradiological-histopathological (CRH) model combined these histopathological variables with the PCR nomogram variables. Performance was evaluated using the C-index and time-dependent area under the receiver operating characteristic curve (tdAUROC). Maximally selected rank statistics were used to define a risk-score cutoff for high- and low-risk groups. Overall survival (OS) was compared using Kaplan-Meier curves and the log-rank test.

In the EVD, the PCR nomogram achieved a C-index of 0.736, compared with 0.701 for the histopathology-based model (P < 0.001) and 0.759 for the CRH model (P < 0.001). The PCR nomogram's 1-, 3-, and 5-year tdAUROCs were 0.764, 0.850, and 0.900, compared with 0.725, 0.822, and 0.918 for the histopathology-based model and 0.796, 0.866, and 0.930 for the CRH model, respectively. High-risk patients showed shorter OS than low-risk patients (hazard ratio, 5.485; 95% confidence interval, 4.261-7.062; P < 0.0001).

The PCR nomogram may provide a readily accessible and interpretable tool for noninvasive preoperative OS prediction and prognostic stratification in adults with diffuse glioma.