Novel nutritional indicator as predictors among subtypes of lung cancer in diagnosis

Abstract

Introduction: Lung cancer is a serious global health concern, and its subtypes are closely linked to lifestyle and dietary habits. Recent research has suggested that malnutrition, over-nutrition, electrolytes, and granulocytes have an effect on the development of cancer. This study investigated the impact of combining patient nutritional indicators, electrolytes, and granulocytes as comprehensive predictors for lung cancer treatment outcomes, and applied a machine learning algorithm to predict lung cancer.

Methods: 6,336 blood samples were collected from lung cancer patients classified as lung squamous cell carcinoma (LUSC), lung adenocarcinoma (LUAD), and small cell lung cancer (SCLC). 2,191 healthy individuals were used as controls to compare the differences in nutritional indicators, electrolytes and granulocytes among different subtypes of lung cancer, respectively.

Results: Our results demonstrated significant differences between men and women in healthy people and NSCLC, but no significant difference between men and women in SCLC patients. The relationship between indicators is basically that the range of indicators for cancer patients is wider, including healthy population indicators. In the process of predicting lung cancer through nutritional indicators by machine learning, the AUC of the random forest model was as high as 93.5%, with a sensitivity of 75.9% and specificity of 96.5%.

Discussion: This study supports the feasibility and accuracy of nutritional indicators in predicting lung cancer through the random forest model. The successful implementation of this novel prediction method could guide clinicians in providing both effective diagnostics and treatment of lung cancers.

Keywords: cancer prediction; lung cancer subtypes; machine learning; nutritional indicators; tumor nutrition.