Prognostic prediction using recursive partitioning analysis of patients undergoing salvage surgery for locally recurrent oral squamous cell carcinoma.

Abstract

Background: The prognosis of patients with locally recurrent oral cavity squamous cell carcinoma (LR-OCSCC) remains poor even when salvage surgery can be performed. Therefore, the current study aimed to identify adverse prognostic factors in these patients and use recursive partitioning analysis (RPA), a machine learning statistical method, to develop a prognostic classification based on these factors.

Methods: The clinical data of 75 patients who underwent salvage surgery for LR-OCSCC at the National Cancer Center between 2010 and 2022 were retrospectively reviewed. Prognostic factors were analyzed using Cox proportional hazards regression models. Patients were classified by survival outcomes using RPA. Survival rates were determined using the Kaplan-Meier method.

Results: The 3-year overall survival (OS) and locoregional recurrence rates for all patients who underwent salvage surgery were 53.4% and 32.7%, respectively. The univariate Cox proportional hazards regression analysis identified concurrent regional recurrence, previous history of radiotherapy to the neck, and recurrence within 12 months from initial treatment as adverse prognostic factors. RPA was performed using these variables and patients were classified into low-, intermediate-, and high-risk groups based on a combination of concurrent regional recurrence and time to recurrence. The 3-year OS rates for the low-, intermediate-, and high-risk groups were 73.3%, 53.3%, and 24.4%, respectively.

Conclusion: A novel prognostic classification for LR-OCSCC salvage surgery was developed to facilitate development of treatment strategies and identification of patients that would not benefit from this procedure.