2.1 Tumor material and patient data

The study population consisted of 368 patients with TSCC diagnosed between 1987 and 2011 at the Maastricht University Medical Centre. That population was an expansion of our study group on which results were published in 2009 (n = 81).³⁵ Their formaldehyde-fixed, paraffin-embedded archival biopsy and resection materials were classified by histopathology at the Department of Pathology, University Hospital Maastricht, the Netherlands. The materials were analyzed for the presence of oncogenic HPV16 DNA by means of PCR and/or FISH as well as p16^4INKA immunostaining in 335 available specimens.² Data on age, sex, TNM classification, tumor differentiation grade, tobacco and alcohol consumption, treatment modality, and follow-up (5 years after treatment) were collected from the head and neck tumor database of our institute and from reviewing clinical, pathological, radiological, and surgical reports. All tumors were reclassified according to the 7th and 8th edition of the UICC's TNM classification.

Classification of smoking and alcohol consumption conformed to Hafkamp et al.: patients were classified as daily smokers (≥1 cigarette, pipe and/or cigar per day), nonsmokers (never smokers), or former smokers (those who had stopped smoking more than 10 years before the diagnosis of TSCC).⁶ During multidisciplinary counseling, treatment plans were based on tumor size, neck staging, presence of distant metastases, tumor histology and cytology of metastases, feasibility of surgery, clinical condition, comorbidities, and histopathology of resection specimens in case of surgery. Elective treatment of the neck was performed routinely—also in the N0 neck—because of the high incidence of occult metastases in TSCC.^{15, 19-21}

The investigation was conducted in accordance with the declaration of the 18th meeting of the world medical association in Helsinki 1964 and subsequent revisions. Approval for the study protocol was granted by the institutional ethical committee. Written informed consent was obtained from all included patients.

2.2 Statistical analysis

Variables considered were age at time of diagnosis, sex, TNM classification, tumor differentiation grade, smoking and alcohol consumption, T, N, and M classification, and therapy.

The Youden Index was calculated to determine the optimal cut-off point in the ROC displaying the relation between sensitivity and specificity for the range of age value.⁴⁶

Survival analysis was performed; disease-free survival (DFS) and overall survival (OS) were calculated using the Kaplan–Meier method for each of the considered independent variables. Five-year survival (OS) was calculated from date of diagnosis until death or until discharge from follow-up. DFS was calculated from date of diagnosis until date of recurrence (local, regional or distant). Patients without recurrence were censored at date of last follow-up or death. Level of significance was determined at p ≤ 0.05. Four patients who initially presented with distant metastases were excluded from the survival analysis. Multivariate analyses were performed using the Cox proportional hazards model. Variables remained in the model if p-values were below 0.10.² IBM SPSS Statistics version 20 was used for the statistical analysis.

3.1 Demographic data and outcome related to HPV association

In total, 368 patients with TSCC were included with an average age of 60.3 years (range 39–87). HPV association could be tested in 335 patient samples, of which 32.8% were HPV-positive (n = 110/335). An increasing prevalence of HPV-associated TSCCs during the last decade (since 2002) was observed (Figure 1).

HPV-positive TSCCs were significantly related with a non- or former smoking status, with alcohol use of less than 2 units/day and a poorly differentiated tumor status. There was an equal distribution among age and sex groups regardless of HPV association.

No statistical differences were noted in the development of second primary tumors regardless of HPV association. Also, no differences were seen in prior history of head and neck cancer or cancer at other sites in both groups. For HPV-positive and HPV-negative tumors, an equal number of patients presented with distant metastases. HPV status relative to patient characteristics is summarized in Table 2. The presented small numbers on failure of control of disease in HPV-positive patients did not allow an adequate uni- or multivariate analysis, in particularly on smoking status and age in this group.

3.2 Analysis of the 7th edition UICC tumor classification system for staging TSCCs

Compared to HPV-negative TSCCs, HPV-positive tumors were associated with significantly smaller primary tumors (Table 3). The extension of neck disease did not differ between both groups. As a consequence, HPV-positive and HPV-negative TSCCs were equally distributed over UICC tumor stages I–IVc in the 7th edition. Within each tumor stage of the 7th edition, HPV-positive TSCCs were associated with a smaller T classification. For TSCCs staged Iva, for example, 38 out of the 61 HPV-positive tumors were classified T1–T2, compared to 31 out of 108 HPV-negative TSCCs (χ², p < 0.001).

In HPV-positive TSCCs, only N3 (affected lymph node >6 cm) was associated with a worse survival (Cox regression, p < 0.001; OR 8.233; 95%CI 2.68–25.29). No significant differences in survival were noted between N stages N0 to N2c, nor between the different stages when we classified tumors according to the 7th edition. Correspondingly, only TSCCs staged IVb or higher showed a worse survival (Cox regression, p < 0.001), and no prognostic differences were noted between stages I and IVa, which altogether cover 90% of the HPV-associated TSCCs.

Among the largest subgroups, stage IVa, for example (n = 181), an overall survival of 85% was found in HPV-positive TSCCs, a significantly higher rate than the 42.1% found in HPV-negative TSCCs (log rank, p < 0.001). This difference may possibly reflect the association of HPV and smaller primary tumor size, given that in stage IVa the 5-year OS for T1-2 and T3-4 were, respectively, 62.3% versus 28.7%, independent of HPV status (log rank, p < 0.001).

In contrast, in the HPV-negative population, a worse survival was associated with increasing T classification, N classification, and UICC tumor stages (Cox regression, p < 0.001).

Thus, in the 7th edition, HPV-positive tumors were mainly staged higher as a consequence of their nodal classification. However, this higher stage seemed to have no prognostic implications.

3.3 Analysis of the 8th edition UICC tumor classification system for staging TSCCs

All HPV-positive tumors were restaged in our study. In the 8th edition, a distinction is made between clinical and pathological staging. Because 38 out of our 110 HPV-positive TSCCs were treated surgically, pathological staging was possible in these 38 tumors (Table 4).

Using the 8th edition of the UICC system for clinical staging, restaging of the majority of HPV-positive tumors resulted in classifying only 3% of patients as having stage IV tumors, compared to 67% in the previous classification. Figure 2 depicts the shift between stages. Using the 8th edition, we found a decreasing survival with increasing clinical UICC tumor stage in HPV-positive tumors (p < 0.001) (Table 4). In the group of HPV-positive TSCCs which were treated surgically and thus staged pathologically (n = 38), we found no decrease in survival with increasing tumor stage.

3.4 Univariate analysis of survival of T, N, and M classification (8th edition UICC TNM classification) in patients with HPV-positive TSCC

Only nodes larger than 6 cm (cN3 status) were associated with a worse survival (p < 0.001), and no differences in survival were found for necks that were clinically staged cN0, cN1, and cN2 (Table 5). Remarkably, cN1 (unilateral nodes with a diameter of no more than 6 cm) seemed to be associated with a better survival than when no lymph nodes were affected. The number of contralateral or bilateral involved necks in TSCCs was very low in this study.

Also, no correlation with survival was seen when N classification was defined pathologically, and no difference in survival was found between HPV-positive pathologically staged necks pN1 and pN2.

Altogether, only lymph nodes that were both larger than 6 cm (cN3) or TSCCs that were seen with distant metastases (M1) correlated with a poor prognosis in HPV-positive TSCCs.

Table 6 presents the odds ratios (and 95% confidence intervals) for each considered explanatory variable obtained from Cox regression modeling of OS.

3.5 Univariate analysis of survival of nonanatomical characteristics of patients with HPV-positive TSCC

Smoking, age >65 years, cN3 status and M classification were correlated with OS. Treatment category had no influence on OS, with the exception of palliative care for which a significantly lower OS was found

The rationale for choosing age 65 years as the cut-off point is as follows. The Youden Index was calculated to determine the optimal cut-off point in the ROC curve that displays the relation between sensitivity and specificity for the range of age value.⁴⁶ The optimal value was 2.22 for both the total data set and for the HPV+ cases. The Youden Index corresponded to the age of 63.5 for the total data set and 66.5 for the patients with HPV+. Age >65 years showed a significant correlation with a worse prognosis (p < 0.001).

In the HPV-positive group, the 5-year OS in the non- or former smokers group (n = 41/110) was 95%, compared to 66% in the smoking group (n = 62/110) (p < 0.001). The number of pack-years had no further influence on prognosis. There was no difference in survival between never-smokers and former smokers (those having quit more than 10 years before diagnosis). Moreover, non- or former smokers were equally distributed over the different stages in the UICC 8th edition (Figure 3). Interestingly, being a non- or former smoker had a favorable prognosis independent of tumor stage. In stage I the overall survival of nonsmokers was 95%, in stage II 100%, and in stage III 90%.

An inverse relationship was found between age and smoking behavior in patients with HPV-positive TSCCs (χ², p = 0.036): nonsmokers were more often older than 65 years, whereas smokers developed a HPV-positive TSCC more often at or under age 65. Smokers older than 65 had a worse overall survival (log rank, p = 0.049). Smoking and age were not correlated to T classification, N classification, and/or UICC tumor staging (8th edition).

3.6 Multiple regression analysis of survival including T, N, and M classification and nonanatomical characteristics of patients with HPV-positive TSCC

The multiple regression analysis resulted in a model with M classification, smoking category, and age 65 years or older (Table 7).

After analysis of the above-mentioned prognostic factors (smoking behavior, age, and N3 status), we developed a predictive model of four groups of patients with HPV-positive TSCCs for patients without hematogenous metastases (Figure 4). M classification was not included as the 5-year OS was 0%. Group I (n = 41) included non- or former smokers with a 5-year overall survival rate of 95.1%, regardless of age, number of pack-years or T and/or N classification, with the exception of cN3 classification. Group II (n = 45) included smokers of age 65 or younger with a 5-year OS of 75.6.1% (HR 5.649, 95%CI 1251–25498). The third group (n = 13) consisted of smokers older than 65 years of age with a 5-year OS of 46.2% (HR 14.165, 95%CI 2935–68368). Survival in group II and III was determined regardless of TNM staging. The fourth group contains patients staged with a N3 staged neck This group showed a 5-year OS of 33% (n = 6). Interestingly, the two patients who survived were nonsmokers who were seen with a cN3-staged neck.

We then considered other prognostic classification models which utilize nonanatomical parameters, as introduced in the literature (Ang et al., O'Sullivan et al. [ICON-S], and Huang et al.). When applying these to our HPV-positive tonsillar carcinoma patients, no relation between the proposed staging and outcome was found (Figure 4).^{38, 40, 42} The model proposed by Ang et al. had been tested on the stage III/IV tumors treated with concomitant chemotherapy, and also on the total population.

4.1 Role of nonanatomical characteristics

The 8th edition of the clinical classification system is the result of a study published by O'Sullivan et al. (ICON-S).⁴² The predictive staging model adopted in that study included only anatomical features of the HPV-associated tumors. The earlier recursive partitioning analysis RTOG (2010) by Ang et al. included nonanatomical parameters such as age and tobacco smoking for the first time, and reported that tobacco smoking had important prognostic value.³⁸ Results of the study by Ang et al. were validated by others.^{48, 49} Regarding smoking, Marur et al. noticed that treatment failures in a de-escalating regime of combining cetuximab with radiotherapy were seen in smokers (>10 pack-years).⁵⁰ Although, in the study of Hawkin et al., number of pack-years were not correlated with survival, authors found a significant correlation between overall survival and a 1+ pack-year smoking history. Interestingly, 1+ pack-year smoking was not correlated with disease-free survival.⁵¹ Further study by Rietbergen et al., on the other hand, showed no differences in outcome regarding smokers versus nonsmokers.⁵² In that study, smoking status was defined based on the number of pack-years and no separate classification was performed for former-smoker status, which might have influenced the results for the smoking group.⁵² Broughman et al. postulated to leave the 10 pack-year rule, proposed by Ang et al. as a stratification tool in HPV-positive OPSCCs, because of the favorable prognosis of former-smoking status in regard of the number of pack-years in their recent study.^{38, 53} Haigentz et al. emphasized that including smoking in a predictive model has great limitations because of the lack of validated, prospective data, and the subjectivity of data collection on tobacco use.⁵⁴ In our predictive model, non- or former smokers had a very favorable prognosis of more than 95% 5-year OS, even in more advanced tumor stages. It need to be emphasized that this groups of non- or former smokers include also patients that quitted smoking for more than 10 years despite having smoked numerous pack-years. Moreover, the number of pack-years was probably therefore not a predictor in prognosis in our patient population. Our study is the first to present that a time period of 10 years since cessation to smoking has a similar favorable outcome as nonsmokers with HPV-positive OPSCCs regardless of number of pack-years. This stresses the importance of adequate history taking regarding smoking status and period since cessation in the work-up of patients with HPV-positive OPSCCs.

Concerning the role of age, both Ang et al. and Huang et al. found that age influenced prognosis in OPSCCs in their studies^{38, 40} In our data, we observed a prognostic role for age with a cut-off point of 65 years.

In the present study, outcome of HPV-positive TSCCs was not predominantly TNM classification dependent, even when using the 8th edition. The most significant prognostic factors in HPV-positive TSCCs were smoking, age, N3 classification, and the presence of distant metastases. Therefore, we developed a new prognostic model comprising four groups. The first group consisted of non- or former smokers (patients who had quit smoking more than 10 years prior to the diagnosis of TSCC), and was associated with a 5-year OS of 95.1%, even in more advanced tumor stages and in former smokers who quitted smoking longer than 10 years ago the number of pack-years had no influence on prognosis. Group 2 included smokers aged 65 or younger with an associated overall survival rate of 75.6%. In group 3 patients who smoked and were older than 65 had a 5-year OS of 46.2%. Patients with N3 and M1 classification comprised group 4. Interestingly, the two patients who survived in this latter group were the only nonsmokers with N3-staged necks. Within the different groups, survival was not differentiated by T and/or N classification. Therefore, this proposed model appears to provide a simple additional tool for predicting outcome in the clinical setting. The current authors are aware of the relative small sample sizes on which this model is based. Therefore, an attempt to validate our predictive model was undertaken. A previous published database of almost 1000 patients with OPSCCs of which HPV-status and the concerning variables were collected and obtained with data transfer agreement. Nevertheless, smoking history was noted by pack-years only and times of cessation were not as such obtained in the dataset nor could be obtained out of original clinical records. This finding hampered validation of the in this study present predictive model. Nonetheless by presenting the model, the authors would like to emphasize the crucial role of smoking behavior including former-smoking, and age in the prediction of survival in HPV-related TSCCs, as both nonanatomical variables act almost independently of the UICC staging system (8th edition) with the exception of the N3-neck and the presence of distant metastases. The model highlights the role of adequate smoking history taking in the ongoing discussion whether or not smoking behavior is of influence on survival in HPV-positive TSCCs.

The role of nonanatomical parameters is not taken into account in the 8th version of the UICC classification system. Results of our research group on a prognostic model for OPSCCs was previously presented by Rios et al.⁵⁵ These results were validated by Rietbergen et al. in a larger cohort, showing the large impact of performance status on outcome in the whole patient group.⁵² However, within the HPV-positive subgroup no further differentiation in risk profiles was provided. The present study is the first in which both age and smoking behavior play a role in classifying HPV-associated tumors in a way that is almost without reference to TNM classification. Our findings suggest that outcome in HPV-positive TSCCs is influenced by patient characteristics including senescence and intoxication (factors such as tobacco use) regardless of tumor stage. Furthermore, as a low number of locoregional and distant failures were noted, it appears that mechanisms through which senescence and intoxication influence overall survival are not necessarily tumor-related, as also Hawkins suggested.⁵¹ Nevertheless, these factors have proven to be unavoidable when considering response to therapy and rate of survival and when taking into account which patients might benefit from de-escalating therapeutic strategies.

4.2 How to address the N0 neck

In our patient population, a clinically negative neck classification was not associated with a better prognosis than for N1. The cut-off point for prognostic value in our study of tonsillar carcinomas was N3, which meant that very few bilaterally involved (and thus N2) neck stages were diagnosed. The prognosis for the N0 neck was even worse than for the N1 neck. As already mentioned regarding the study by Cramer et al., in which the 8th edition was validated in a population of more than 15000 HPV-positive oropharyngeal tumors, cN1 classification was associated with a significantly better survival than N0. Moreover, the bilaterally involved neck (cN2 classification) was not associated with a significantly worse survival than the clinically negative neck after adjustment for age, sex and race. Only cN3 classification was significantly associated with a worse survival.⁴⁴ In previous research, it was noticed that patients with HPV-positive carcinomas more often had a lymph node as presenting symptom when compared to their HPV-negative counterparts. Presenting with that symptom may have led to an earlier discovery of the primary tumor.³⁵ It is therefore possible that these presenting “alarming” nodes are associated with a better prognosis, as we found in our study. Also Fritsch et al. and Ang et al. found that patients with an HPV-positive single neck node between 3 and 6 cm in size (N2a, 7th edition) had a better outcome than patients without lymph node metastases.^{37, 38} Fritsch et al. compared outcome based on N classification between HPV-dominant (tonsillar fossa and base of tongue) and non-HPV-dominant oropharyngeal subsites in more than 15 000 oropharyngeal tumors.³⁷ In their HPV-dominant population, cN2a (7th edition) had a better survival rate than the N0/1 patients. In the total population, no differences in outcome were noted as long as lymph node metastases were unilateral (<cN2c). Our study only included patients with TSCC; in this strongly HPV-dominant subgroup, similar results were found and a clinically negative neck (cN0) in HPV-positive TSCCs was not associated with a better survival than necks with lymph nodes smaller than 6 cm in diameter (i.e., cN1 and cN2 classification).

4.3 Role of oropharyngeal subsite

The influence of HPV on prognosis is often analyzed for all oropharyngeal subsites, without even discriminating between HPV-dominant and nondominant subsites. For the present study, we selected a group of 110 patients with squamous cell carcinomas of the tonsil, the most HPV-dominant oropharyngeal subsite. According to Sood et al., a bilaterally involved neck classification is predominantly seen in tumors at the base of the tongue, indicating that lymph node dissemination patterns differ even within the HPV-dominant oropharyngeal sites.⁴⁴ This may explain why only a few bilaterally involved necks are seen in our population of tonsillar carcinomas. It may also explain the lack of significance of the cN2 neck in our study where we classify bilateral neck involvement in accordance with the 8th edition. The location of the tumor in the different subsites of the oropharynx therefore likely plays a prominent role in the development of advanced (N-) tumor stages in HPV-positive OPSCCs.

4.4 HPV-detection

P16 immunohistochemistry is a widely available, low-cost test, unlike the more complex HPV-in situ hybridization. In our study, p16 was combined with HPV-DNA PCR and/or FISH. We found that 15 out of 124 p16-positive patients were not HPV-positive (12%).

Nauta et al. raised the issue of consensus on the exact definition of HPV-associated OPSCC.⁵⁶ HPV infection alone is not sufficient to classify an instance of OPSCC as HPV-related since the presence of HPV-DNA could merely reflect a transient infection. Detection of p16 alone is not specific for HPV-activity. Smeets et al. described an algorithm in which p16-IHC was combined with HPV 16 DNA PCR.⁵⁷ Taberna et al. recently investigated the outcome of HPV-positive OPSCC in relation to the definition for HPV-positivity.⁵⁸ They concluded that definitions of HPV positivity have an impact on TNM classification and patients' survival. Bussu et al. recently confirmed that p16-IHC alone may not be specific enough to become the diagnostic standard from the perspective of treatment de-intensification. Standardization of clinical use and of detection methods for HPV as a marker for molecular characterization in head and neck oncology is warranted.⁵⁹