¹⁸F-FDG PET/CT metabolic parameters-derived prediction model for microsatellite instability in colorectal cancer

To create a non-invasive model for forecasting microsatellite instability (MSI) status in colorectal cancer (CRC) using preoperative ¹⁸F-FDG PET/CT metabolic parameters, guiding personalized treatment.

A total of 156 pathologically confirmed CRC patients who underwent ¹⁸F-FDG PET/CT at Mindong Hospital affiliated to Fujian Medical University (November 2023 - August 2025) were retrospectively enrolled. Twenty-one parameters (demographic, clinical, pathological, PET/CT metabolic and heterogeneity indices) were collected. MSI status was determined by immunohistochemistry (IHC) for four mismatch repair (MMR) proteins (MLH1, MSH2, MSH6, PMS2). After variable selection by LASSO regression and collinearity elimination through correlation analysis, a nomogram model was constructed using multivariate logistic regression. Model performance was evaluated by ROC, calibration, and decision curves, with internal validation using LOOCV, 5-fold cross-validation, and bootstrapping.

Among patients, 28 (17.95%) were microsatellite instability-high (MSI-H, deficient mismatch repair, dMMR) and 128 (82.05%) were microsatellite stable (MSS, proficient mismatch repair, pMMR). Independent predictors included mucinous component, lymph node metastasis, differentiation grade, tumor size, and metabolic tumor volume (MTV). The model's AUC was 0.883 (95% CI: 0.820-0.945), with a sensitivity of 0.789 and a specificity of 0.857. Calibration curves and the Hosmer-Lemeshow test (χ² = 8.641, P = 0.3734) indicated good calibration of the model, and decision curve analysis showed that the model provided high net benefit. The AUCs of internal validation ranged from 0.801 to 0.857, which confirmed the good stability of the model.

The integrated model established in this study is expected to assist in the non-invasive evaluation of MSI in CRC and may provide certain reference value for individualized clinical decision-making.