Data Sheet 1_An explainable nomogram incorporating systemic inflammatory indicators and radiation esophagitis for risk assessment of radiation pneumonitis in lung cancer.pdf

Finally, SHAP analysis was performed to interpret the final Cox model through beeswarm and waterfall plots.</p>Results<p>This study retrospectively collected data on 726 patients with lung cancer who received radiotherapy. The primary endpoint was grade ≥2 RP, with non-RP-related death treated as a competing risk. Multivariable Cox regression, validated by Fine-Gray competing risk analysis, identified alcohol history, radiotherapy modality, baseline lactate dehydrogenase (LDH), systemic inflammation score (SIS), and acute radiation esophagitis (RE) as independent factors (all P < 0.05).

Among the three sequential models, the integrated model (Model 3) achieved the highest C-index (0.688) and 6-month AUC (0.713). Compared to the baseline model (Model 1), Model 3 improved risk reclassification (continuous NRI = 0.274, IDI = 0.037; both P < 0.05). Model 3 demonstrated optimal calibration and yielded the highest net clinical benefit within the 10%-15% threshold probability range in DCA.

Furthermore, SHAP analysis enhanced model interpretability, indicating a potential association between a baseline systemic inflammation score of zero (SIS = 0) and higher RP risk.</p>Conclusion<p>This study developed and validated a nomogram that integrates fundamental clinical factors with inflammatory indicators (LDH and SIS) and early acute radiation esophagitis (RE). This model serves as a multidimensional, auxiliary screening tool for early risk assessment of radiation pneumonitis (RP) in lung cancer patients.</p>