Prognostic value of diabetes-related clinical indices in patients with non-small cell lung cancer and comorbid type 2 diabetes receiving immune checkpoint inhibitor therapy.

BACKGROUND: Immune checkpoint inhibitors (ICIs) targeting programmed cell death protein 1 (PD-1) or programmed death-ligand 1 (PD-L1) have transformed the treatment of non-small cell lung cancer (NSCLC). However, diabetes-induced immune dysfunction may attenuate ICI efficacy. Clinical evidence further suggests poorer outcomes in patients with NSCLC and comorbid type 2 diabetes mellitus (T2DM). Reliable prognostic tools for this subgroup remain lacking. METHODS: We conducted a three-part study. First, a two-sample Mendelian randomization (MR) analysis was performed to assess the causal relationship between T2DM and PD-L1. Second, Cox regression was applied to evaluate associations between diabetes-related clinical indices and progression-free survival (PFS) and overall survival (OS). Third, machine learning models were developed to predict prognosis using diabetes-related features. RESULTS: MR analysis supported a causal association between T2DM and PD-L1. Cox regression identified HbA1c as significantly associated with PFS. Among the machine learning models, the support vector machine (SVM) achieved the best performance for predicting PFS (AUC = 0.935, 95% CI: 0.872-0.997; C-index: 0.763). For OS prediction, the best-performing model was a random forest model based on HbA1c (AUC = 0.863, 95% CI: 0.736-0.990; C-index: 0.618). CONCLUSION: Diabetes-related indices-particularly HbA1c-are meaningful prognostic markers in patients with NSCLC and comorbid T2DM receiving ICI therapy. The proposed machine learning models may support individualized outcome prediction and clinical decision-making.