Robustness of stage as a prognostic factor for lung cancer: a multiverse survival analysis based on SEER data

ABSTRACT Aims The TNM system is one of the most important parameters for the staging of a tumor, as it allows the clinical course of a patient to be determined, and stage has long been indicated as a fundamental prognostic factor for evaluating patient survival. However, due to the different methods used to determine stage and the lack or incompleteness of this information in many patient databases, it has not been possible to verify whether this assumption holds true in every situation. The aim of this study was therefore to explore the real prognostic value of stage by examining all plausible analytic scenarios. Methods and Results Using data from the SEER database, an analysis was structured with the multiverse package to explore the various possible analytic pathways and to verify the value of stage as a prognostic factor for survival through Cox regression models. Several covariates were considered in association with the main variable, stage, which was evaluated through progressive comparisons between the different levels, from the lowest to the highest. From the results obtained, it appears that stage indeed maintains its prognostic and clinical value, both in terms of the Hazard Ratio results and in terms of statistical significance, with p-values <0.05. Conclusion This represents the first step of a study that will be further expanded, potentially with the addition of interactions and additional covariates. Nevertheless, the current findings already indicate that stage consistently retains its prognostic value across the different analytic choices explored.