For nasopharyngeal cancer the AJCC TNM staging system is shown in Tables 2 and 3. Besides T category (local invasion), tumour volume is of prognostic significance, especially when attempting to kill cancer cells by radio- and chemotherapy as opposed to complete surgical resection (cell number, increasing likelihood of hypoxia in larger tumours) (Sze et al. 2004; Chen et al. 2009; Guo et al. 2012). Other staging systems have been developed by different groups. In general, T classification predominantly affects local control, while N classification predicts neck and distant control. Epstein-Barr virus DNA (less than versus equal to or greater than 1,500 copies/ml) is a prognostic biomarker that predicts overall and relapse-free survival. Other markers such as serum lactate dehydrogenase (LDH) (Zhou et al. 2012; Wan et al. 2013), vascular endothelial growth factor (VEGF) and osteopontin have been suggested and should be studied further (Lv et al. 2011).

For cancer of the oral cavity and oropharynx the AJCC TNM staging system is shown in Tables 4 and 5. In these diseases, male gender, cigarette smoking and poor performance status are adverse prognostic factors. Patients who are p16 positive have improved survival. After surgical resection for oral cavity carcinoma, adjuvant radiotherapy may be recommended for patients at higher risk for loco-regional recurrence, but it can be difficult to predict whether a particular patient will benefit. Wang et al. constructed several types of survival models using a set of 979 patients with oral cavity squamous cell carcinoma treated in the United States and Brazil (Wang et al. 2013). Covariates were age, sex, tobacco use, stage, grade, margins, and subsite. The best performing model was externally validated on a set of 431 patients from Princess Margaret Hospital, Toronto, Canada. The primary outcome measure of interest was loco-regional recurrence-free survival. An online nomogram was built from this model that estimates loco-regional failure-free survival with and without postoperative radiotherapy. However, none of the patients had received postoperative radiochemotherapy. Information on extracapsular extension, lymphovascular space invasion, and perineural invasion were not available in this data set, also limiting its general use. The prediction model indicated that patients with positive margins and/or N2–N3 disease will derive the most benefit from adjuvant radiotherapy, although the relative magnitude of this benefit will vary depending on the other patient and tumour characteristics entered into the model. Histologic grade was also an independent predictor of prognosis in stage I or II oral cavity carcinoma in the Surveillance, Epidemiology, and End Results (SEER) database of the National Cancer Institute (Thomas et al. 2013). Among patients age 20–65 with AJCC stage I or II cancer, the adjusted risk of death was 2.7 times greater (95 % CI 1.7 to 4.1) if the tumour was poorly differentiated or undifferentiated than it was if the tumour was well differentiated. Among patients age 66–94, the risk of death was 3.0 (95 % CI 2.0–4.5) times greater. For those over age 65, moderately differentiated tumours also conferred an estimated 42 % increased risk of death, which was borderline significant (p = 0.05).

It is hypothesised that molecular biomarkers might improve current decision models. Protein 53 (p53), insulin-like growth factor II mRNA-binding protein 3 (IMP3), cyclooxygenase 2 (COX2), and HuR were analysed by immunohistochemistry in 96 patients with primary oral squamous cell cancer who underwent surgical resection at the Yonsei Dental Hospital in Seoul, Korea (Kim et al. 2012). On univariate and multivariate analysis, the expression of IMP3 was significantly associated with the risk of death. P53 was also significantly associated with survival in the case of negative IMP3 and the prediction accuracy was improved by including these 2 factors in the prediction model. The latter nomogram included age, sex, presence of lymph node metastases, T stage, tumour site, p53 and IMP3. In general, such tools need external validation in sufficiently large databases before widespread implementation.

For cancer of the larynx and hypopharynx the AJCC TNM staging system is shown in Table 6. Egelmeer et al. performed a population-based cohort study on 994 laryngeal carcinoma patients, treated with radiotherapy from 1977 until 2008 (Egelmeer et al. 2011). Two nomograms were developed and validated. Performance of the models was expressed as the area under the curve (AUC). Unfavourable prognostic factors for overall survival were low haemoglobin level, male sex, high T status, nodal involvement, older age, lower EQD(2T) (total radiation dose corrected for fraction dose and overall treatment time), and non-glottic tumour. All factors except tumour location were predictive for local control. The AUCs were 0.73 for overall survival and 0.67 for local control. External validation of the survival model yielded AUCs of 0.68, 0.74, 0.76 and 0.71 for cohorts from Leuven (n = 109), VU Amsterdam (n = 178), Manchester (n = 403) and Amsterdam (n = 205), respectively, while the validation procedure for the local control model resulted in AUCs of 0.70, 0.71, 0.72 and 0.62. The resulting nomograms were made available on the website www.​predictcancer.​org. Further evidence from multivariate analyses suggests that tumour volume also influences local control and survival in T2 tumours (Rutkowski et al. 2013) and T3-4 tumours (Hoebers et al. 2013).

Wang et al. (2013) analyzed the impact of the lymph node ratio (LNR, ratio of metastatic to examined nodes) on the prognosis of hypopharyngeal cancer patients. SEER registered hypopharyngeal cancer patients with lymph node metastasis were evaluated using multivariate Cox regression analysis to identify the prognostic role of the LNR. The categorical LNR was compared with the continuous LNR and pN classifications to predict cause-specific and overall survival (n = 916). T classification, N classification, M classification, the number of regional lymph nodes examined, the continuous LNR (Hazard ratio 2.4, 95 % CI 1.7–3.4, p < 0.001) were among prognostic variables that were associated with cause-specific survival. The categorical LNR showed a higher C-index and lower Akaike information criterion (AIC) value than the continuous LNR. When patients were classified into four risk groups according to LNR, R0 (LNR = 0), R1 (LNR ≤ 0.05), R2 (LNR 0.05–0.3) and R3 (LNR > 0.3), the Cox regression model for both endpoints using the R classification had a higher C-index value and lower AIC value than the model using the pN classification. In conclusion, using the cutoff points 0.05/0.3, the R classification was more accurate than the pN classification in predicting survival.