Development of a Prognostic Survival Risk Score for Lung Cancer Patients With Type 2 Diabetes Mellitus: A Territory‐Wide Retrospective Cohort Study

Vollständiger Abstract

Worum geht es in dieser Arbeit?

ABSTRACT Background Lung cancer is a leading cause of cancer‐related mortality, and its prognosis is often affected by comorbidities such as type 2 diabetes mellitus (T2DM). This study aimed to identify survival risk factors for lung cancer patients with T2DM, evaluate the predictive models, and develop a risk score system for survival prediction. Study Design and Methods We analyzed data from 5491 lung cancer patients with T2DM from the Hong Kong Hospital Authority Data Collaboration Laboratory (HADCL) (2000–2020). Prognostic factors were evaluated using Cox proportional hazards regression. Four algorithms were used to construct survival analysis models: Cox proportional hazards regression, LASSO Cox regression, survival tree, and random survival forests (RSF). An interpretable risk scoring system was subsequently derived using the selected predictors. Results Older age at cancer diagnosis, male sex, longer duration between T2DM diagnosis and lung cancer diagnosis (T2DM duration), smoking, alcohol consumption, history of stroke, and higher HbA1c were associated with increased mortality risk, whereas hypertension, coronary heart disease, insulin usage, anti‐lipid usage, and anti‐diabetic usage were associated with reduced mortality risk. Among the evaluated models, the RSF model demonstrated the best predictive performance, as indicated by the C‐index (0.71) and time‐dependent AUC (0.883). The developed risk score system included the following criteria: age at cancer diagnosis; T2DM duration; smoking status; HbA1c; HDL‐C; serum potassium (K) levels; LDL‐C. A score ≥ 75 classified 47.33% of patients as high‐risk, with a corresponding five‐year survival probability of 5.51%. Conclusions Among the evaluated models, the RSF model demonstrated the best predictive performance. The developed risk scoring system may support risk stratification and identification of high‐risk patient subgroups, potentially facilitating personalized prognostic assessment and clinical management.