A predictive model for fusion gene positivity in treatment-naive advanced non-small cell lung cancer

Fusion gene-positive non-small cell lung cancer (NSCLC) represents a distinct molecular subtype with unique clinical and biological features. Although next-generation sequencing (NGS) is the gold standard for detecting ALK, ROS1, and RET fusions, it is costly, time-consuming, and occasionally limited by insufficient tissue samples. Therefore, a simple bedside predictive tool is urgently needed to identify patients likely to harbor fusion genes and prioritize testing. This study aims to establish a predictive model for fusion gene positivity and scoring system for fusion genes, assisting clinicians in early identification of patients likely to harbor fusion genes before rapid disease progression occurs, thereby preventing patients from missing optimal treatment opportunities due to rapid disease deterioration. The scoring system can also help prioritize fusion gene testing when tissue samples are insufficient.

We retrospectively analyzed data from 448 NSCLC patients admitted to The First Affiliated Hospital of Guangzhou Medical University between June 2015 and October 2024, meeting the study's inclusion criteria. These patients were categorized into the fusion gene-positive group (130 cases) and the driver-negative group (318 cases). Baseline data between the two groups were compared to identify factors potentially associated with fusion gene-positive. Univariate logistic regression analysis was utilized for screening these factors, which were then included in a multivariate analysis to construct a predictive model for fusion gene-positive. The model's discrimination and calibration were subsequently assessed. Each factor was assigned scores based on the regression coefficient β value in the model, leading to the transformation of the model into a scoring system.

Analysis shows that specific factors independently associated with fusion gene positivity include age, history of thrombosis, carbohydrate antigen 153 (CA153) levels, and vertebral body metastasis. Utilizing these four indicators, we developed a predictive scoring system that effectively evaluates the likelihood of patients testing positive for fusion genes [area under the curve (AUC) =0.797].

The novel scoring model can assess the positivity of the fusion gene. Our research indicates that this score incorporates age, the presence or absence of thrombosis, CA153 levels, and the presence or absence of vertebral metastasis.