Refining Uicc Tnm Stage And Prognostic Groups For Hpv-Related Oropharyngeal Carcinomas

The current TNM staging for oropharyngeal cancer (OPC) was designed empirically for HPV-unrelated [HPV(-)] disease. Emerging evidence suggests it is unsuited for HPV-related [HPV(+)] OPC. This study refines stage grouping for HPV(+) OPC patients and proposes additional prognostic risk groups within the guidelines of the UICC/AJCC TNM framework. We retrospectively analyzed a prospectively assembled OPC cohort treated with primary radiation therapy with or without chemotherapy from 2000-2010. Overall survival (OS) was compared among the current TNM stages (I-IV) for HPV(+) and HPV(-) patients separately. Recursive partitioning analysis (RPA) with ordinal T and N elements derived new RPA-stages objectively. Cox regression calculated relative mortality risk (RMR) to derive additional RMR-stages. The performance of RPA- and RMR-stages was assessed against current UICC stages in predicting OS based on 4 widely accepted criteria: hazard consistency within each stage level; hazard discrimination between stage levels; outcome prediction, and sample size balance. Prognostic risk groups were further derived by RPA combining T-, N-classification, age, and smoking pack-years (PY). A total of 810 HPV ascertained (by p16 staining) non-metastatic OPCs were identified, including 573 HPV(+) (UICC stage I: 8; II: 25; III: 79; IV: 461) and 237 HPV(-) (I: 8; II: 31; III: 38; IV: 160) OPC. Median follow-up was 5.1 years. Reduced 3-year OS with higher UICC TNM stage was evident for HPV(-) (88, 67, 62, and 39% respectively, p = 0.003). However, OS was similar within HPV(+) (88, 87, 81, and 80% respectively, p = 0.712). RPA and RMR methods were applied to the HPV(+) cohort to refine current UICC stage groupings. RPA divided non-metastatic HPV(+) into RPA-I (T1-3N0-2b), RPA-II (T1-3N2c), and RPA-III (T4 or N3) with corresponding 3-year OS of 88, 81, and 63%, respectively (p < 0.001). M1 disease (20% OS at 3-years) was classified as RPA-stage IV. RMR also provided a valid stage grouping scheme (not shown) but was more cumbersome compared to RPA-stage. RPA-stage and RMR-stage were the two best stage groupings while UICC stages performed least well. Prognostic risk grouping by RPA sub-divided all HPV(+) into: group I (T1-3N0-N2c ≤ 20 PY), group IIA (T1-3N0-N2c ≥ 20 PY), group IIB (T4 or N3, age ≤ 70), group III (T4 or N3, age > 70), and group IV (M1 disease), with corresponding 3-year OS of 93, 74, 67, 44, and 20%, respectively. This large cohort study confirms that current UICC TNM stage is unsuited for HPV(+) OPC although acceptable for HPV(-). A refined RPA-based TNM stage grouping significantly improved survival prediction performance for HPV(+) OPC. Prognostic risk groupings are further enhanced by incorporating non-anatomical factors. The result should be validated in an independent dataset.