Analysis of specific survival and local control through a recursive partitioning analysis in patients with head and neck carcinoma.

INTRODUCTION AND OBJECTIVES: Recursive partitioning analysis (RPA) is a technique that allows prognostic classification in oncological patients. The aim of the present study is to analyse by means of an RPA a cohort of patients with squamous carcinomas of the head and neck (SCHN). METHODS: 5,226 SCHN were retrospectively analysed with an RPA, considering the specific survival and local control of the disease as dependent variables. A cohort of patients was used for the creation of the classification model, and another cohort was used to carry out its internal validation. RESULTS: Considering specific survival as a dependent variable we obtained a classification tree with 14 terminal nodes that were grouped into 5 categories, including as partition variables the local and regional extent of the tumour, and the location of the tumour. When considering the local control of the disease as a dependent variable we obtained a classification tree with 10 terminal nodes that were grouped into 4 categories, including as partition variables the local extension and location of the tumour, the type of treatment performed, the age of the patient, and if it was a first tumour or a subsequent neoplasm. The validation study confirmed the prognostic capacity of the models developed with the RPA. One of the advantages of the RPA is that it allows the identification of groups of patients with specific behaviour. CONCLUSION: RPA is shown to be an effective technique for the prognostic classification of patients with a SCHN.