Background. Several clinical trials have proved that concurrent chemoradiotherapy is more efficacious than radiotherapy alone among high-risk patients with head and neck squamous cell carcinoma (HNSCC) who undergo surgery. A risk-group classification defined according to a recursive partitioning analysis (RPA) for these patients has been recently proposed. The objective of the present study was to carry out an external validation of this RPA-derived classification system. Methods. A retrospective study of 442 HNSCC patients treated with surgery and postoperative radiotherapy was conducted. The external validity of the RPA-derived classification system was assessed, and its ability to stage patients and to predict locoregional control of the disease was compared with the TNM system. Results. The RPA-derived classification system succeeded in obtaining a monotonic prognosis gradient in locoregional control of the disease with increasing stage, and achieved greater differences in survival between stages than the TNM and pTNM classifications. Besides, the RPA method had a better homogeneity of the categories included in each stage, and in the heterogeneity between stages. Conclusions. The RPA-derived classification system allowed for the clear definition of prognostic groups in surgically treated HNSCC patients, improving the prognostic capacity of the TNM and pTNM classifications. The RPA-derived classification system is a useful tool in the definition of patients who, given a poor prognosis, should be considered candidates to adjuvant chemoradiotherapy.