Inflammation-nutrition biomarker model for survival prediction in lung cancer patients with concurrent tuberculosis

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Objectives: To explore the prognostic value of eight inflammation-nutrition biomarkers in patients with lung cancer and tuberculosis as no multidimensional prognostic models for this comorbid population are available currently.Methodology: A retrospective study included 100 patients with lung cancer and tuberculosis admitted to a tertiary hospital from October 2019 to October 2024. Eight inflammation-nutrition markers (NLR, PLR, SII, LMR, PNI, HALP, HRR, ALB/GLB) were chosen as predictors while overall survival (OS) was the major event. Feature selection was implemented by LASSO regression; a Cox proportional hazards model was established afterwards. The nomogram’s performance was assessed by ROC curve and C-index as well as the calibration using bootstrap resampling. The statistical power was calculated by PowerSurvEpi and sensitivity analyses were implemented to test the robustness of the model.Results: There were six predictors remaining in the final model including diabetes, ECOG PS, NLR, PNI, HRR and RDW. Among them, ECOG PS was an independent prognostic factor (HR = 1.76, p = 0.04). The nomogram achieved a good performance (C-index = 0.71), an AUC of 0.693 for 3-year OS as well as an excellent calibration (Bootstrap P > 0.05). In the high-risk subgroup with ECOG PS ≥ 2 and NLR>8, the 5-year survival rate was close to zero. The model achieved an adequate statistical power (83%, α = 0.05). Sensitivity analysis revealed an significant interaction between ECOG PS and NLR (p = 0.032) and NLR>8 was the most robust threshold for this interaction.Conclusion: This is the first study to establish and validate a combined inflammation-nutrition prognostic model for patients with lung cancer and tuberculosis. Our model provides a quantitative tool to stratify individual risk and offers evidence for the usage of nutritional interventions in high-risk patients.

This publication has 20 references indexed in Scilit: