# Development of a prognostic prediction model for non-smoking lung adenocarcinoma based on pathological information and laboratory hematologic indicators: a multicenter study

## TL;DR

This study creates a model to predict survival for non-smoking lung cancer patients using tumor data and blood markers.

## Contribution

A novel prognostic model combining tumor and blood indicators for non-smoking lung adenocarcinoma patients.

## Key findings

- The model achieved C-index values of 0.811 in training and 0.786 in testing for survival prediction.
- High-risk groups showed significantly worse outcomes compared to low-risk groups (P < 0.001).
- The model effectively distinguished survival outcomes across tumor stages (P < 0.0001).

## Abstract

To develop a simple and practical model to predict the prognostic survival of non-smoking patients with lung adenocarcinoma by combining general pathological information with laboratory hematologic indicators.

Cox univariate and multivariate analyses were used to identify the variable indicators. A Cox proportional hazards model was constructed based on the selected variables to compare survival outcomes between the high-and low-risk groups of non-smoking patients with lung adenocarcinoma and to validate the model’s performance. Subsequently, a nomogram model was established to systematically evaluate the impact of selected variables on prognosis.

Data of non-smoking patients with lung adenocarcinoma from four hospitals were retrospectively collected. We enrolled 1,172 patients, this includes 372 external validation data. Multivariate analysis identified six significant variables (P < 0.05): tumor TNM stage, tumor size, white blood cell count, neutrophil percentage, lymphocyte percentage, and hemoglobin level. We combined these six variables to build a model. The C-index of the training set is 0.811 (0.780–0.842), this value is 0.786 (0.737–0.835) in,test set and 0.810 (0.772–0.847) in validation set. The area under the curve (AUC) results of the predicted 3-years overall survival (OS) of the three data sets were 0.850, 0.819, and 0.860, respectively. These values for 5-years were 0.811, 0.771, and 0.849. Stratified analysis based on tumor staging showed that the model effectively distinguished outcomes (P < 0.0001). High-risk groups demonstrated significantly poorer prognosis compared to low-risk groups (P < 0.001).

The prognostic model based on tumor TNM stage, tumor size, white blood cell count, neutrophil percentage, lymphocyte percentage, and hemoglobin levels effectively predicted the prognosis of non-smoking patients with lung adenocarcinoma. Compared with the more studied blood markers at present, the indicators of our model do not need conversion, Our model provides a useful reference for personalized diagnosis and treatment in clinical practice.