Predictive Markers and Cancer Prevention
Human papillomavirus positivity predicts favourable outcome for squamous carcinoma of the tonsil
Article first published online: 3 JUN 2003
Copyright © 2003 Wiley-Liss, Inc.
International Journal of Cancer
Volume 106, Issue 4, pages 553–558, 10 September 2003
How to Cite
Li, W., Thompson, C. H., O'Brien, C. J., McNeil, E. B., Scolyer, R. A., Cossart, Y. E., Veness, M. J., Walker, D. M., Morgan, G. J. and Rose, B. R. (2003), Human papillomavirus positivity predicts favourable outcome for squamous carcinoma of the tonsil. Int. J. Cancer, 106: 553–558. doi: 10.1002/ijc.11261
- Issue published online: 1 JUL 2003
- Article first published online: 3 JUN 2003
- Manuscript Accepted: 18 FEB 2003
- Manuscript Revised: 9 JAN 2003
- Manuscript Received: 18 SEP 2002
- tonsil cancer;
- cyclin D1;
Mutations in the p53 and retinoblastoma (pRb) pathways associated with the use of tobacco and alcohol are common in squamous cell carcinoma (SCC) of the head and neck. Cell cycle proteins are also affected by human papillomavirus (HPV), which may also have an aetiological role in cancers at particular sites, most notably the tonsil. Attempts to identify prognostic molecular markers in head and neck cancers have met with conflicting results, but few studies have been undertaken with tumours of known HPV status at a single anatomic site. In our study 86 tonsil cancers were analysed for HPV status by sequence analysis of polymerase chain reaction products and for the expression of cell cycle proteins (p53, p21CIP1/WAF1, pRb, p16INK4A, cyclin D1 and p27KIP1) by immunohistochemistry. The HPV status could be established in 67 of the tumours. Thirty-one (46%) of these were HPV-positive, predominantly (28/31) for HPV16. Findings were related to tumour recurrence and patient survival. None of the cell cycle proteins independently predicted recurrence or survival. Patients with HPV-positive tumours, however, were significantly less likely (p < 0.05) to have recurrence or to die of disease than those with HPV-negative tumours, after adjusting for the effects of the cell cycle proteins, clinical stage, pathological node status, tumour grade, age, gender and treatment. These findings support the concept that HPV-positive tonsil cancers may be a distinct biological group with less aggressive characteristics. Screening of tonsil cancers for HPV DNA may help optimise treatment and provide more accurate prognostic information. © 2003 Wiley-Liss, Inc.
Worldwide, squamous cell carcinoma (SCC) of the head and neck is the eighth most common malignancy, with about half a million new cases each year.1 The disease burden falls most heavily on the poorest countries, but even in a developed country like Australia the incidence in the state of New South Wales in 2000 was 28 per 100,000 for males and 7.7 per 100,000 for females.2 These tumours have diverse clinical features and distinctive natural histories, and their management involves surgery, radiotherapy or chemotherapy alone or in combination. Prognosis has not really improved over the last 2 decades despite improvements in our understanding of the patterns of spread of invasive tumours, the significant advances that have occurred in surgical reconstruction and the increased use of combined modality treatment. Clinical staging currently guides treatment but remains an uncertain predictor of outcome.
Technological advances over the last decade have provided insight into molecular pathways associated with the development of head and neck cancer, raising prospects that molecular markers might assist early diagnosis and predict patient outcome including response to treatment. The tumour suppressor proteins p53 and the product of the retinoblastoma gene (pRb) are well recognised targets of carcinogens associated with smoking and alcohol consumption, the 2 most important risk factors for head and neck cancer.3 More recently there have been reports of abnormal expression or inactivation of other genes operating within the p53 and pRb pathways at G1/S transition, such as cyclin D1 and the cyclin dependent kinase inhibitors (cdkI) p21CIP1/WAF1(p21), p16INK4A(p16), and p27KIP1(p27).4, 5, 6 Assessments of specific genetic changes as biological markers in aerodigestive malignancies have generated conflicting data,7, 8 however, and no particular marker has emerged as a reliable guide to clinical outcome.
The cell cycle proteins p53, pRb, p16, p21, p27 and cyclin D1 are also affected by the oncogenes of human papillomavirus (HPV) either directly or indirectly, and there is increasing evidence that oncogenic HPVs have an aetiological role in some head and neck malignancies.9, 10, 11, 12, 13 The association is strongest in tonsil cancer where positivity rates have sometimes exceeded 50%, with HPV type 16 invariably predominant. The influence of HPV on clinical outcome remains uncertain. There is no consensus on the prognostic significance of HPV in head and neck cancers generally, but some recent studies have shown an association between improved clinical outcome and HPV-positivity in tonsil cancer.10, 14
We recently examined associations between the expression of p53, pRb, p16, p21, p27, cyclin D1 and HPV-positivity/negativity in a series of 50 tonsil cancers (Head & Neck, in press). The results suggested that HPV-positive tonsil cancers represent a distinct molecular and pathologic entity. Research has now been extended to investigate the impact of these markers on clinical outcome in a larger cohort of patients. We describe investigations into the effects of HPV DNA status and level of expression of key cell cycle proteins, p53, pRb, p16, p21, p27 and cyclin D1, on the risk of recurrence and survival in 86 patients. The effects of clinical stage, pathological node status, tumour grade, patient age, gender and treatment were taken into account in our analyses.
MATERIAL AND METHODS
Investigations were carried out on 86 patients with SCC of the tonsil treated at Royal Prince Alfred Hospital (RPAH) (71) or Westmead Hospital (15) Sydney, Australia, between 1987 and 2000. Fifty of the 71 RPAH patients were subjects of an earlier study determining the relationships between HPV and cell cycle protein expression (Heat & Neck, in press). The study group of 86 represented all patients for whom tumour material and clinicopathological data were available and who had received definitive treatment with a minimum follow up of 2 years. No patient had distant disease beyond the head and neck at diagnosis. The median age of the group was 57 years (range 30–81) and the male to female ratio was 66:20. The distribution of the tumours according to T and N stage classification of the 1997 American Joint Committee on Cancer Staging criteria15 is presented in Table I. Twenty-six (30%) of the tumours were well-differentiated, 51 (59%) moderately well-differentiated and 9 (11%) poorly-differentiated according to the World Health Organization criteria for grading of squamous cancers of the oral mucosa.16 Fifty (58%) of the patients were treated with surgery and post-operative radiotherapy. Surgery was the only form of treatment for 17 (20%) patients, whereas 19 (22%) were treated with radiotherapy only. Indications for postoperative adjuvant therapy included large deeply invasive tumours, close or positive surgical margins and lymphovascular or perineural invasion. Time to recurrence and disease-specific survival were measured from the date of diagnosis.
|N stage||T stage||Total|
Detection and typing of HPV by sequence analysis of PCR products
PCR was carried out on crude extracts of 5μm sections of paraffin-embedded tumour tissues prepared as described.17 All lysates were tested using a HPV 16 type-specific PCR that amplified a 109 bp region of the HPV 16 E6 gene and also by a general primer-mediated PCR, GP5+/6+, targeting the L1 region of the virus, and capable of detecting a broad spectrum of mucosal HPV types.18, 19 Tumours testing HPV-negative were analysed subsequently by nested PCRs using degenerate primers A10/A5-A6/A8 and CP65/70ct-CP66/69ct designed to detect mucosal HPVs and Group B1/cutaneous epidermodysplasia verruciformis (EV) HPVs respectively.20, 21 The presence of amplifiable DNA in HPV-negative samples was confirmed using a PCR for the β-globin gene.17 The presence of tumour cells was confirmed in H&E stained sections of all HPV-negative samples. The positive and negative controls and the precautions taken to minimise the chances of cross contamination were as described previously.22 Sequence analysis was carried out on pooled PCR products purified by polyethylene glycol precipitation using ABI Prism DyeDeoxyTerminator cycle sequencing chemistry (Applied Biosystems, Foster City, CA). HPV type was determined on the basis of >90% homology with HPV sequences deposited in GenBank using BLAST software.
Analysis of the expression of cell cycle markers was carried out by performing immunohistochemistry on 5 μm sections of paraffin-embedded tumour on silane-coated slides (Sigma-Aldrich, St. Louis, MO). Single representative tumour blocks were selected for examination from each patient. If multiple blocks were available, selection was on the basis of substantial amounts of viable tumour as well as some histologically-normal epithelium for comparison; otherwise, a block of normal tonsillar epithelium from the corresponding patient was used. In the very few cases where only tumour tissue was available, normal tissues from multiple patients having tonsillectomy for reasons unrelated to malignancy were used as reference.
The following monoclonal antibodies were used: p53, p21, p27, pRb and cyclin D1 (DO7/M7001, S×118, S×53G8, Rb1 and DCS-6: all from DAKO, Carpinteria, CA) and two p16 antibodies (F-12 Santa Cruz Biotechnology, Santa Cruz, CA and Ab-4, Clone 16P04 Neomarkers, Fremont, CA). Antigen retrieval was carried out by heating the slides at 90°C for 12 min in 10× buffer (13.4 mM EDTA, 20.6 mM Tris, 10.9 mM tri-sodium citrate pH 8.0 [pRb, p16 Santa Cruz, p53, p21, p27]); or Target Retrieval Solution (high pH) (DAKO) (cyclin D1, or 0.01M citrate buffer PH 6.0 [p16 Neomarkers]). After extensive trials to optimise the protocols, all subsequent steps were carried out at room temperature. Endogenous peroxidase was blocked using 3% hydrogen peroxide in PBS for 5 min. The sections were incubated with unlabeled primary antibodies for 30 min. Processing was then carried out using the DAKO LSAB and chromogen kits. Biotinylated anti-mouse, anti-rabbit and anti-goat immunoglobulins were applied for 15 min. After 2 washes in phosphate buffered saline, streptavidin-peroxidase was added for 15 min then 3,3′-diaminobenzidine and hydrogen peroxide for 5 min. The sections were counterstained with Mayer's haematoxylin. Negative controls included omitting the primary antibody and substitution of the primary antibody with normal serum. A positive control comprising a section from a tonsil cancer block shown previously to be positive for the particular marker was incorporated in each run.
Immunostaining was evaluated semiquantitatively by at least 3 of the 4 independent observers (W.L., R.S., B.R., C.T.) without knowledge of the clinical data or HPV status. At least 1,000 tumour cells were evaluated for each section. For all cell cycle markers except p16 (Neomarkers) the evaluation method was similar to that described by Erber et al.4 taking account of the intensity of staining (graded + to +++) and the proportion of stained cells (graded + to +++) using normal epithelium as reference. Tumours scored as downregulated showed >20% reduction in numbers of cells staining and/or reduced intensity of staining compared to normal epithelium, those scored as unchanged had up to a 20% difference from normal; whereas those designated as upregulated showed >20% increase in numbers and/or intensity of staining. The p53 and p16 (Neomarkers) antibodies rarely stained normal epithelium; tumour cells were graded according to the proportion of cells stained as used by Sano et al.23 Tumours with >10% staining were regarded as positive for p53 and those with <10% staining as negative. For p16, tumours with >5% staining were taken as upregulated and <5% as downregulated. Results were validated by repeat staining of sequential sections cut from the same block on at least one occasion. Interobserver variations occurred infrequently and were reconciled over a double-headed microscope.
Associations between clinical outcome (tumour recurrence and disease-specific survival) and the HPV status of the tumours, expression of cell cycle markers (p53, p21, pRb, p16, p27, cyclin D1) and clinicopathological variables (clinical stage, pathological node status, tumour grade, age [as a continuous variable], gender and treatment [surgery plus adjuvant radiotherapy, surgery alone or radiotherapy alone]) were analysed by the Mantel-Haenszel method.24 All tests were 2-tailed and only p-values of <0.05 were considered significant. The effect of any variable was expressed as a relative risk, which is the risk of recurrence or survival at any particular time for those with that variable compared to those without it. Cox proportional hazards regression25 was then used to examine the independent effects of each significant variable. The statistical software (S Plus, Math Soft Inc.) was used for all analyses.
Recurrence and survival
Among the 86 patients there were 32 (37%) recurrences: 7 of the 17 patients treated by surgery alone; 16 of the 50 patients who received surgery and adjuvant radiotherapy and 9 of the 19 treated by radiotherapy alone. Disease recurred in the tonsil in 18 patients, the neck (with local disease controlled at the primary site) in 7 and in distant sites in 6. The site of recurrence for 1 patient was unknown. The median period of follow-up was 4 years; (range 2–11). The 5-year disease-specific survival rate was 72%.
Prevalence and type distribution of HPV DNA
The HPV status of 19 tumours could not be determined due to lack of amplifiable DNA as shown by consistently negative results in the beta-globin PCR. Of the remaining 67 tumours, 31 (46%) were HPV-positive, whereas 36 (54%) were negative for both mucosal and cutaneous HPVs. Twenty-eight of the 31 HPV-positive samples were confirmed by sequence analysis as HPV type 16; the remaining 3 yielded negative results with the HPV 16-specific PCR but were positive using the consensus GP5+/6+ combination. In these 3 samples the presence of HPV was confirmed on sequence analysis, but the type could not be determined because the percentage homology with existing HPVs was <90%, indicating either the presence of multiple types or a novel type(s). The proportion of samples for which HPV status was unable to be determined was higher than experienced in our earlier surveys of HPV status in paraffin-embedded cervical cancers and may be attributable to length of time in fixative. Nonetheless, there were no significant differences in clinicopathological variables (including recurrence and survival) between the tumours of known and unknown HPV status. The relationship between HPV-positivity and -negativity and clinicopathological variables in the 67 tumours is presented in Table II.
|Pathological node status|
|Unknown or not applicable||5||12|
|Moderately well differentiated||15||22|
Expression of cell cycle proteins
For all markers except p16 (Neomarkers), staining of tumour cells was invariably nuclear but variable in intensity and distribution when compared to the expression in normal tonsil epithelium. p16 (Neomarkers) stained both nucleus and cytoplasm of tumour cells but rarely stained normal cells. In the case of p53, weak staining was noted in normal cells of a small number of tumour-containing sections, and in almost all of these cases the tumour cells were p53 negative. In two cases normal and tumour cells both showed staining, and the p53 result for the tumour was recorded as negative. By these criteria, 58 (67%) of the tumours were positive for p53 and 28 (33%) were negative. The expression of pRb, p16 (Santa Cruz and Neomarkers) p21, p27 and cyclin D1 in the 86 tumours was up-regulated in 30 (35%), 35 (41%) 37 (43%), 44 (52%), 14 (16%) and 32 (38%) respectively; and down-regulated in 19 (22%), 12 (14%), 49 (57%), 13 (15%), 27 (31%) and 14 (16%) respectively.
Evaluation of the immunostaining for p16 (Santa Cruz), pRb, p21 and p27 using the proportion of tumour cells showing positive staining as sole criterion of up- or down-regulation according to other published criteria for head and neck cancers23, 26, 27 yielded almost identical results (data not shown).
Recurrence and survival
Among the 86 patients none of the clinico-pathological variables significantly predicted tumour recurrence on univariate analysis, although there were strong trends for tumours with pathological node involvement to recur locally (p = 0.09) and tumours in females tended to recur less frequently than in males (p = 0.09). In terms of 5-year disease-specific survival, only stage and gender were significant variables on univariate analysis (p = 0.009 and p = 0.037 respectively). Patients with Stage IV disease were more likely to die of disease than those with earlier stage tumours, and females had less risk of death than males. Among the clinico-pathological variables, only stage remained significant when the other variables were taken into account.
HPV and cell cycle proteins.
Univariate analysis of the data for the 67 patients for whom HPV status was known showed that the HPV status of the tumours significantly affected the risk of recurrence. Patients with HPV-positive cancers had a 3 times lower risk of recurrence than those whose cancers were HPV-negative (p = 0.012). The Kaplan-Meier recurrence curve28 for HPV-positive and -negative cases is shown in Figure 1.
There was also a statistically significant association between HPV status and survival. The relative risk of patients with HPV-positive cancers dying of disease was more than 5 times lower than for those whose tumours were HPV-negative (p = 0.013). The 3-year survival rate for the HPV-positive group was 95% compared to 65% in the HPV-negative group, and at 5 years, the survival rate was 89% in the HPV-positive group compared to 65% in the HPV-negative group. The Kaplan-Meier survival curve for HPV-positive and negative cases is shown in Figure 2. Univariate analysis showed that for patients with Stage IV disease survival was significantly greater for those with HPV-positive tumours compared to those with HPV-negative tumours (p = 0.016). The Kaplan-Meier survival curve for HPV-positive and -negative cases is shown in Figure 3. A statistically significant survival benefit for those with Stage I–III/HPV-positive tumours was not demonstrated, but only 7 patients with Stage I–III tumours died of their disease and only 4 of these had tumours of known HPV status.
Initial assessments of the effects of the expression of cell cycle proteins on prognosis were carried out on data from all 86 patients. p16, as assessed using the Neomarkers antibody, was the only significant predictor of recurrence. Patients with cancers showing upregulated p16 had a lower risk of recurrence than those whose cancers were HPV-negative (p < 0.05). In terms of survival, p16 (using the Neomarkers antibody) and pRb were both significant on univariate analysis (p < 0.05 and p = 0.035 respectively); patients whose tumours had overexpression of p16 or reduced expression of pRb had a better chance of survival than those with tumours had not. There were also strong trends toward associations between survival and cyclin D1 and p16 expression using the Santa Cruz antibody (p = 0.06 and p = 0.052 respectively). Patients whose tumours had reduced expression or loss of cyclin D1 or overexpression of p16 had better survival prospects than those whose tumours had not.
Cox regression analysis was carried out on the 67 tumours of known HPV status. We established previously a strong statistical relationship between HPV-positivity and downregulated expression of pRb and cyclin D1 and upregulated expression of p16 (using the Neomarkers antibody) that is highly likely to be causal (MS submitted). pRb, cyclin D1 and p16 (Neomarkers) were therefore excluded from the model assessing the significance of HPV, and conversely HPV was excluded from the model assessing the significance of these variables. HPV-positivity remained a statistically significant marker of recurrence (p < 0.05) after p21, p27, p53 and stage, pathological node status, gender, age, tumour grade and treatment had been taken into account, whereas p16 remained significant when all variables except HPV were included in the model (p < 0.05). Cox regression analysis for survival showed that the benefit for those with HPV-positive cancers was independent of stage, pathological node status, tumour grade, gender, age, treatment, p53, p21 and p27 expression (p < 0.05). p16 and pRb remained significant when HPV was excluded from the model.
Our study represents the most comprehensive assessment of HPV and cell cycle proteins operating at G1/S transition as predictors of recurrence and survival at a single anatomical site within the head and neck region. The HPV-positivity rate of almost 50%, and the overwhelming predominance of HPV 16, were consistent with data published previously on tonsil SCCs.9, 11, 12, 13 In agreement with the report of Mellin et al.14 and a small survey carried out by Andl et al.,10 we have identified HPV-positivity as a favourable prognostic indicator in tonsil cancer. In this survey, patients with HPV-positive tumours were 3 times less likely to have recurrence and 5 times less likely to die of the disease than those whose tumours were HPV-negative, and the risk was independent of cell cycle protein expression and clinicopathological variables. Our study also confirms earlier reports of a survival benefit for patients with Stage IV tumours that are HPV-positive.29, 30 Interestingly, we were unable to find any association between HPV-positivity and Stage IV disease as reported by Paz et al.31 Further studies will be needed to determine whether the prognostic significance of HPV seen in tonsil SCCs applies to head and neck cancers generally. Although Gillison et al.12 and Lindel et al.32 reported the association in oropharyngeal cancers, many of these were tonsil cancers and several other studies found no relationship.29, 31, 33 Paradoxically, an association between HPV-positivity, overexpressed cyclin D1 and increased tumour aggression has been reported recently in laryngeal SCCs.34 These inconsistencies are likely to reflect study design features such as sample size, technical issues including sensitivity and specificity of the methods used for HPV detection, or methods of data analysis. Biological differences relating to anatomical site may, however, have been a contributing factor.
The mechanisms underpinning the apparently better prognosis for patients with HPV-positive cancers are not yet clear. An association between HPV-positivity and radiosensitivity of oropharyngeal tumours has been demonstrated32 suggesting that the effect of HPV may be treatment dependent. This theory is consistent with the survival advantage observed for those with Stage IV disease who are primarily treated with radiotherapy and chemotherapy. Support for this theory has also come from in vitro studies indicating that interaction between the HPV E6 and the p53 proteins does not inactivate p53 completely.35 If this is true, the presence of E6 may not correspond functionally to mutations in the p53 gene. Along the same lines, it has been hypothesised that radiation and other toxic agents may decrease the capacity of HPV E6/E7 to interfere with p53/pRb and other host proteins, rendering the cancers more susceptible to therapy.14 In either case the availability of functional p53 may minimise chromosomal aberrations allowing tumours to retain a radiosensitive and chemosensitive phenotype. It is noteworthy that all patients in the Mellin study14 received radiotherapy, 45% as primary therapy, whereas in our study only 80% of the patients received radiotherapy 50% of these postoperatively. We found no evidence of a relationship between treatment and outcome. It is recognised that over the lengthy study period there has been some shift in the treatment emphasis, but overall techniques have not changed significantly.
Interactions with HPV E6/E7 oncoproteins have clearly complicated assessments of cell cycle proteins, such as p53, p27 p21 and cyclin D1, as prognostic markers in head and neck cancers, and undoubtedly account in part for the inconsistencies in published data. There have been few investigations of the biological significance of cell cycle protein expression in head and neck tumours of known HPV status apart from a handful of studies that focussed on p53 alone.30
The prognostic significance of p53, and its down-stream target p21, in head and neck tumours remains unresolved. On balance, there is evidence suggesting an adverse effect of p53 abnormality on local control and survival for patients with cancers of the head and neck,8 but interpretation has been complicated by differences in treatment and methods of detecting abnormal p53. We have been unable to confirm the inverse relationship between HPV-positivity and p53 overexpression reported by others,10 and there was no association between p53 overexpression and prognosis. Overexpression of p53 does not necessarily indicate mutation, however, and stabilisation of the p53 protein may have occurred through alternative pathways. There have been few studies of the prognostic effects of p21 in oropharyngeal malignancies. Although the HPV E7 oncoprotein has been shown to inactivate p21 by uncoupling cdk activity from cdkI,36 we have not confirmed a relationship between p21 expression and HPV-positivity and there was no association between p21 and prognosis in our study.
Progression through the G1 phase of the cell cycle is mediated by the release of the E2 cellular transcription factor after phosphorylation of pRb. The phosphorylation of pRb is regulated positively by cyclin D1 and negatively by p16. Inactivation of the pRb/p16/cyclin D1 pathway seems of critical importance in carcinogenesis of the head and neck.7, 37 Evidence is emerging that the disruption may occur along 2 main molecular pathways with distinctive biological implications. One commonly reported scenario involves overexpression or amplification of cyclin D1 with concomitant loss of p16 expression, and this has been consistently associated with increased recurrence and reduced survival.8 The other pathway involves an autoregulatory feedback loop mechanism and is characterised by reduced expression of cyclin D1 (and pRb) and upregulated p16. Evidence for the latter pathway has come from analysis of a small series of HPV-positive tonsil cancers in which pRb (and cyclin D1) was inactivated by viral oncogenes such as HPV E7,10 and from investigations of diverse cell lines in which pRb was inactivated by mutation.38, 39 In the small tonsillar series, HPV-positivity with associated disruption to pRb and cyclin D1 was associated with a better prognosis.10 The results of our present study indicate that expression of reduced pRb may serve as a surrogate for HPV as a marker of survival. Interestingly, the significance of p16 overexpression as substitute for HPV as a favourable prognostic marker was shown to be antibody-dependent. Cyclin D1 did not quite reach statistical significance as a prognostic indicator even on univariate analysis, despite its strong relationship with HPV. This highlights the complexities of the interacting pathways governing the cell cycle. The finding that HPV-positivity is a favourable prognostic indicator is entirely consistent with the reported relationship between overexpression or amplification of cyclin D1 and poor prognosis.8 Further large studies will be needed to establish whether overexpression or amplification of cyclin D1 is a predictor of poor outcome independently of HPV.
The p27 protein is a putative tumour suppressor gene, a regulator of drug resistance in solid tumours and promoter of apoptosis. Reduced expression of p27 has been reported as an independent adverse prognostic indicator in various human cancers including breast, colon and prostate adenocarcinomas and there is also evidence of its prognostic significance in head and neck SCCs.26, 40 To our knowledge this is the first time that the prognostic significance of this protein has been analysed specifically in tonsil cancers. Despite reports of in vitro evidence of interactions between HPV E7 and p27,41 there was no association between p27 expression and HPV-positivity in this series, and no evidence that p27 influenced clinical outcome.
There are general expectations within the medico-scientific community that genetic analyses will eventually become a routine part of cancer detection and management. Large-scale prospective studies will be needed to establish the practical value of HPV typing in the clinical setting. Further investigation is needed into the mechanisms underlying the natural history of HPV related tonsil cancers.
We are grateful for the assistance of Ms. B. Cakir in data collection, and to Mr. Lawrence Young for advice on imunohistochemistry.
- 2Cancer in New South Wales: incidence and mortality 2000 featuring projections for selected sites. NSW Cancer Registry, May 2002., .
- 15American Joint Committee on Cancer Staging manual. 5th ed. Philadelphia: Lippincott-Raven, 1997. 34–5., , , , , .
- 16Histological typing of cancer and precancer of the oral mucosa. 2nd ed. Berlin, New York: Springer, 1997. 11–2., , , .
- 25Regression models and life tables. J R Stat Soc 1972; 34: 187–220..