To develop and internally validate a multivariable model for individualized pseudoprogression risk estimation in high-grade glioma patients after radiotherapy, addressing the challenges of distinguishing pseudoprogression from true progression.
Approach:
Data Collection: Pseudoprogression was adjudicated through multidisciplinary review, with histopathological confirmation for some cases and a requirement of at least 6 months of stable or improved follow-up imaging for non-histologically confirmed cases.
Model Validation: The integrated model's performance was evaluated using AUC metrics in a held-out validation cohort, with AUC values reported alongside 95% confidence intervals.
Key Findings:
The integrated model achieved an AUC of 0.811, combining RadScore and clinical variables.
The clinical-imaging model without RadScore achieved an AUC of 0.744.
RadScore alone achieved an AUC of 0.771.
Interpretation:
The integrated model provides a framework for estimating the risk of pseudoprogression by combining various clinical and imaging data.
Limitations:
The study was conducted at a single center, which may limit generalizability and introduce potential biases.
External validation with standardized imaging protocols is needed.
Conclusion:
The integrated model demonstrated acceptable internal validation performance for estimating pseudoprogression risk in high-grade glioma patients.
