ABSTRACT
IMPORTANCE: The current American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC) staging system, developed for human papillomavirus (HPV)-unrelated disease, discriminates poorly when applied to HPV-related oropharyngeal squamous cell cancer (OPSCC), leading to calls for a new staging system. OBJECTIVE: To compare the prognostic ability of the AJCC/UICC seventh edition staging system; a recently proposed system, the International Collaboration on Oropharyngeal Cancer Network for Staging (ICON-S); and a novel objectively derived system for HPV-related OPSCC using a national database of patients primarily treated with either radiation or surgery. DESIGN, SETTING, AND PARTICIPANTS: In this observational study, patients with HPV-related nonmetastatic OPSCC were identified in the National Cancer Database between 2010 and 2012. Recursive partitioning analysis (RPA) was used to derive the proposed-RPA staging system. The data were analyzed from March to May 2016. MAIN OUTCOMES AND MEASURES: Overall survival was calculated using the Kaplan-Meier method. The performance of the 3 systems was compared using published criteria, and internal validation using bootstrap methods was performed. Survival differences between stage groups were evaluated using the log-rank test. RESULTS: A total of 5626 patients (86.0% male; median [range] age, 58 [21-90] years) were identified. The median (range) follow-up was 28.5 (0.1-58.8) months. A novel staging system (proposed-RPA) consisting of stage IA, T1-2N0-2a; stage IB, T1-2N2b-3; stage II, T3N0-3; stage III, T4a-bN0-3 resulted in 3-year overall survival rates of 91%, 87%, 81%, and 70%, respectively. This system, as well as the ICON-S, significantly prognosticated for survival when either primary surgery or primary radiation subgroups were examined (log-rank P < .001 for all). The AJCC/UICC system, ICON-S, and proposed-RPA all significantly predicted survival outcomes when analyzed globally (log-rank P < .001 for all). The AJCC/UICC system could not differentiate between survival when stages I and IVA were compared, however (log-rank P = .17). On comparative performance evaluation for survival prediction, the proposed-RPA provided superior prognostication compared with the other systems. CONCLUSIONS AND RELEVANCE: We validated the ICON-S staging as prognostic, overall, and in primary radiation therapy and surgery subgroups, but ultimately found that a staging system consisting of stage IA, T1-2N0-2a; stage IB, T1-2N2b-3; stage II, T3N0-3; and stage III, T4a-bN0-3 (with stage IV representing M1 disease) outperformed the others. The proposed-RPA is an alternative staging system that should be evaluated for potential adoption as part of the next AJCC/UICC staging system.
