Combined analysis of HPV-DNA, p16 and EGFR expression to predict prognosis in oropharyngeal cancer



Molecular prognostic indicators for oropharyngeal squamous cell carcinoma (OSCC), including HPV-DNA detection, epidermal growth factor receptor (EGFR) and p16 expression, have been suggested in the literature, but none of these are currently used in clinical practice. To compare these predictors, 106 newly diagnosed OSCC for the presence of HPV-DNA and expression of p16 and EGFR were analyzed. The 5-year disease-free survival (DFS) and overall survival (OS) were calculated in relation to these markers and a multivariate Cox analysis was performed. Twenty-eight percent of the cases contained oncogenic HPV-DNA and 30% were positive for p16. The p16 expression was highly correlated with the presence of HPV-DNA (p < 0.001). Univariate analysis of the 5-year DFS revealed a significantly better outcome for patients with p16-positive tumors (84% vs. 49%, p = 0.009). EGFR-negative tumors showed a tendency toward a better prognosis in DFS (74% vs. 47%, p = 0.084) and OS (70% vs. 45%, p = 0.100). Remarkable and highly significant was the combination of p16 and EGFR expression status, leading to 5-year DFS of 93% for p16+/EGFR− tumors vs. 39% for p16−/EGFR+ tumors (p = 0.003) and to a 5-year OS of 79% vs. 38%, respectively (p = 0.010). In multivariate analysis p16 remained a highly significant prognostic marker for DFS (p = 0.030) showing a 7.5-fold increased risk for relapse in patients with p16-negative tumors. Our data indicate that p16 expression is the most reliable prognostic marker for OSCC and further might be a surrogate marker for HPV-positive OSCC. HPV+/p16+ tumors tended to have decreased EGFR expression, but using both immunohistological markers has significant prognostic implications. © 2007 Wiley-Liss, Inc.

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide, with more than 300,000 cases occurring in the oral cavity and pharynx every year.1 Despite a decreasing occurrence of most tumors of the upper aerodigestive tract in developed countries, reflecting a reduced exposure to alcohol and tobacco, the incidence of oropharyngeal squamous cell carcinoma (OSCC) has been found to be on the rise.2, 3 Overall prognosis of HNSCC has not been improved in recent years. Only the implementation of multimodal treatment regimes has resulted in a slight favorable prognosis for the oropharyngeal subset in the last decade.4

Insight into genetic and molecular alterations in the development of head and neck cancer research has grown explosively in recent years. As a consequence potential molecular and prognostic markers and new targets for cancer treatment have been proposed, and there is hope that individualized cancer treatment based on these molecular targets will improve prognosis and decrease side effects.

Over the last 5 years it has become clear that a subset of OSCC is associated with oncogenic human papillomavirus (HPV), in particular HPV type 16.5, 6, 7 HPV-positive OSCC furthermore differ from HPV-negative tumors with respect to tumor differentiation,8, 9 risk factors10 and genetic changes,6, 11 indicating that HPV-positive OSCC represent a separate tumor entity12 and that at least 2 molecular pathways leading to OSCC exist.1 On basis of HPV-DNA detection by PCR, the presence of HPV has been associated with a favorable prognosis in many studies,13, 14, 15 although some studies did not find this relation.16, 17 This discrepancy may be due to the fact that not all HPV-DNA-positive OSCC show viral oncogene expression, which is believed to be essential in HPV related carcinogenesis and tumor progression. Therefore, the biological meaning of HPV-DNA detection is doubtful in these cases.18, 19


ABC, avidin-biotinylated-peroxidase complex; DFS, disease-free survival; EGFR, epidermal growth factor receptor; HNSCC, Head and neck squamous cell carcinoma; HPV, human papillomavirus; OS, overall survival, OSCC, oropharyngeal squamous cell carcinoma.

The p16 protein inactivates the function of cdk4- and cdk6-cyclin D complexes. One of the critical substrates of the G1-specific cdk complexes is the release of E2F through phosphorylation of the retinoblastoma (pRb)-E2F protein complex.20 Hence, p16 negatively regulates cell proliferation by suppression of hyperphosphorylation of pRb.21, 22 On the other hand, pRb acts as a negative regulator of p16 expression.23 Functional loss of p16 has been demonstrated in a wide variety of tumors.24 In contrast, in HPV associated cancer p16 inactivation is rare.25 Moreover, p16 overexpression has been observed in squamous cell carcinomas and dysplasia of the cervix.26 This is in line with the fact that the viral oncoprotein E7 inactivates the pRb protein, which would otherwise inhibit p16 transcription, and with recent reports and our previous results showing highly increased p16 expression in HPV-positive OSCC.6, 9, 18, 27, 28, 29, 30 In contrast, frequent p16 inactivation by transcriptional silencing or genetic changes has been reported for HPV-independent HNSCC,31 and chromosome 9p21 abnormalities were reported as early events in the development of HNSCC.32 In parallel to that are some studies reporting a decreased survival for OSCC patients where p16 is inactivated regardless of HPV-DNA detection.33

In addition to p16 as a potential prognostic marker for OSCC the expression of the epidermal growth factor receptor (EGFR) has been in focus as a molecular marker. EGFR is a transmembrane tyrosine kinase receptor of the erbB-family that is normally expressed at low levels on the surface of most normal cells. A variety of studies have shown that EGFR is overexpressed in HNSCC.34, 35 Overexpression of EGFR has been associated with a more aggressive clinical behavior, resistance to treatment and a poor prognosis.36, 37 Recently also quantitative determination of EGFR expression could be correlated to prognosis in OSCC.38

Despite the indicated potence as prognostic markers for OSCC, no studies have been performed to analyze their combination. Here we examine the combination of the most promising predictive markers for OSCC, including HPV-DNA detection, p16 and EGFR expression, in a series of primary OSCC.

Material and methods

Subjects and material

We retrospectively analyzed formalin-fixed, paraffin-embedded tissue from 106 consecutive patients with newly diagnosed and histologically confirmed squamous cell carcinoma of the oropharynx treated at the hospital of the University of Cologne between January 1997 and December 2002. Twenty-five of these cases were the subject of a previous publication.30 Patients' ages ranged from 43 to 80 years (median age 59 years). Eighty-three patients were males (78%) and twenty-three were females (22%). Tumor staging was assessed according to the 2002 American Joint Committee on Cancer staging criteria.39 Three percent of the patients presented at stage I, 14% at stage II, 18% at stage III, 54% at stage IVa and 9% at stage IVb. Two patients (2%) had distant metastases (stage IVc) at the time of diagnosis.

Tumor specimens of all cases were obtained during surgery, immediately frozen in liquid nitrogen, and stored at −80°C. In addition, tissue was fixed in 4% buffered formalin and embedded in paraffin by routine procedures in the Department of Pathology of the University of Cologne. Slides from all blocks were reviewed by one pathologist (H.U.K.) to select the most representative areas of the tumors for further processing and for immunohistochemistry. The criteria for block selection were to take vital tumor tissue without necrosis and to have a front of invasion. Only blocks with estimates of at least 70% tumor cells were included.

HPV16-positive cervical carcinoma specimens were used as positive controls. Histological grading was performed in a blind manner following the WHO criteria for squamous cell carcinomas of the oral mucosa.40

Sample preparation, polymerase chain reaction and HPV typing

Tissues were processed as previously described.8 After confirming integrity of DNA by β-globin gene PCR,41 HPV sequences were detected by highly sensitive nested PCR protocols with degenerated primers A10/A5-A6/A8 for HPV group A (genital/mucosal) HPVs42 and CP62/70-CP65/69a for group B1 (cutaneous/EV) HPVs.43 PCR products (10 l) were separated in 2% agarose gels and visualized by ethidium bromide staining. HPV typing was performed as described previously.8

Immunohistochemical staining

Ninety-six cases in the study group were eligible for immunohistochemical staining similar to that described previously.18 In each case, 2–3 m sections of formalin-fixed and paraffin-embedded biopsy samples were processed by the avidin–biotinylated–peroxidase complex (ABC) method (Chem Mate Detection kit; Dako Cytomation, Carpinteria, CA). Sections were dewaxed by passage through xylene and rehydrated by a graded series of ethanol, followed by microwave treatment for antigen retrieval, which was for p16 staining 2 times 7 minutes at 600 Watts in 10 mM citrate buffer (pH 6.0) and for EGFR staining 20 minutes in Pepsin at room temperature (Pepsin solution Digest-All3, Zymed). After a brief rinse with tris-buffer (BUF1), they were incubated with the primary antibody for 25 minutes (p16: Ab-4, clone 16P04 Neo Markers, Fremont, CA; EGFR: EGFR, clone 31G7 Zytomed). After another brief rinse with tris-buffer (BUF1), the tissue was incubated with the biotinylated secondary antibody for another 25 minutes. After a brief rinse with tris-buffer (BUF2), the endogenous peroxidase was inactivated with peroxidase-blocking solution, 3 times for 2½ minutes each. After rinsing, the tissues were incubated for another 25 minutes with Streptavidin conjugated with horseradish-peroxidase. Visualization was performed with 3-Amino-9-Ethylcarbazol (AEC, Dako Cytomation, Carpinteria, CA) 3 times for 5 minutes, and sections were counterstained with hematoxylin. Nondysplastic, peritonsillar squamous cell epithelium was used as a negative control.

Strong nuclear staining as well as strong cytoplasmic staining was considered positive for p16 expression. P16 immunostaining was regarded as overexpression if it was strong and diffuse and more than 60% of the tumor cells were p16-positive. EGFR expression was regarded as positive when more than 50% of the cell membranes of the tumor cells were completely stained. Cytoplasmatic staining was not considered for this stratification.

Statistical analysis

P16, EGFR and HPV status were analyzed using cross-tabulations and χ2 test with the SPSS Base System, version 11.0 (SPSS, Chicago, IL). Disease-free survival and overall survival rates were estimated using the Kaplan–Meier algorithm for incomplete observations. Twenty-six patients were excluded from survival analysis because of the presence of stage IVc disease, unresectable tumor, incomplete curative treatment or R2 resection margin, or because they never achieved disease free status. The overall survival time was defined as the interval between the date of diagnosis and the last date when the patient was known to be alive (censored) or date of death for any reason (uncensored). The disease-free survival rate was measured as the period of time between the date of diagnosis and the date of the last follow-up examination in which the patient was disease-free (censored), or the date of first recurrence independently if it was a local, regional or distant recurrence (uncensored). The log rank test was used to test for differences between subgroups. A p-value ≤0.05 was considered statistically significant in 2-sided tests.


HPV status and HPV typing

Sequences of high-risk HPV were detected in DNA preparations for 30 (28%) out of 106 oropharyngeal carcinoma cases. Of these positive cases, 29 (97%) carried HPV16 and 1 (3%) HPV33 sequences. In 2 of the cases (7%), additional infection with low-risk HPV was detected. HPV-positive tumors showed significantly poorer differentiation than HPV-negative tumors (p = 0.022, Table I).

Table I. Associations Between Tumor's Biologic Markers and Patient's Clinicopathologic Characteristics [Number/Total Number (%)]
CharacteristicsHPV-positive (%)p16-positive (%)EGFR-positive (%)
  1.  n.a. = not accessible. Of the 106 patients, 96 cases of the study group were eligible for immunohistochemical staining.

 Male21/83 (25)21/74 (28)39/74 (53)
 Female9/23 (39)8/22 (36)7/22 (32)
 p value0.1930.4740.085
Tumor stage   
 Stage I1/4 (25)1/3 (33)2/4 (50)
 Stage II3/15 (20)3/14 (21)7/14 (50)
 Stage III5/18 (28)4/17 (23.5)3/17 (18)
 Stage IV21/69 (30)20/60 (33)34/61 (56)
 p value0.8760.8200.051
Tumor grade   
 Grade I0/5 (0)0/3 (0)1/3 (33)
 Grade II11/55 (20)10/52 (19)28/52 (35)
 Grade III19/46 (41)19/41 (46)17/41 (41)
 p value0.0220.0090.433
Tumor site   
 Tonsils21/63 (33)21/56 (37.5)27/56 (48)
 Base of tongue3/14 (21)3/13 (23)9/13 (69)
 Others6/29 (21)5/27 (18)10/27 (37)
 p value0.5940.3290.384
Lymph node involvement   
 N+24/74 (32)23/66 (35)33/66 (50)
 N−6/32 (19)6/30 (20)13/30 (43)
 p value0.1510.1420.544
Smoking (n.a., n = 7)   
 +14/80 (17.5)13/71 (18)39/72 (55)
 −14/18 (78)14/17 (82)4/17 (23.5)
 p value0.0000.0000.020
Alcohol (n.a., n = 8)   
 +12/72 (17)10/64 (16)33/64 (52)
 −16/25 (64)17/24 (71)9/24 (37.5)
 p value0.0000.0000.239
Therapy (none, n = 2)   
 Only surgery6/21 (29)6/19 (32)6/19 (32)
 Surgery + RT/RCT20/64 (31)19/56 (34)26/56 (46)
 Definitive + RT/RCT4/19 (21)3/19 (16)13/19 (68)
 p value0.6690.4520.152

Expression of p16

There were 3 different immunostaining phenotypes observed for p16 immunoreactivity, i.e., strong and diffuse cytoplasmic and nuclear staining, weak cytoplasmic and nuclear staining and negative staining for all tumor cells (Fig. 1a). In general, surrounding mesenchymal cells showed no p16 immunoreactivity, except for some weak staining found occasionally in lymphocytes and salivary glands. Nondysplastic, squamous cell epithelium was p16-negative and served as an internal negative control. Interestingly, in most cases (29; 30.2%) p16 was either overexpressed in all malignant cells or totally absent in the tumor (58; 60.4%), whereas the group with weak cytoplasmic and nuclear staining of the tumor cells consisted only of 9 (9.4%) cases.

Figure 1.

(a) EGFR expression in oropharyngeal squamous cell carcinoma, demonstrating strong and diffuse membrane and cytoplasmic staining (magnification ×200). (b) P16 expression in oropharyngeal squamous cell carcinoma, demonstrating strong and diffuse cytoplasmic and nuclear staining (magnification ×200). [Color figure can be viewed in the online issue, which is available at]

P16 overexpression was significantly associated with poorer differentiation of the tumor. Of the p16-positive cases, 19 (65.5%) were poorly differentiated, whereas of the p16-negative cases only 33% showed poor differentiation (p = 0.009; Table I).

Expression of EGFR

The EGFR staining was more heterogenous compared with the pattern observed for the p16 staining (Fig. 1b). Membranes and cytoplasm of the tumor cells showed variable staining intensities in different cases. For the EGFR staining, the number of immunopositive cancer cells ranged from 0% to 100%. Using the above described evaluation criteria, 46 (43%) of the OSCC cases showed EGFR overexpression.

Comparison of the tumor stage with EGFR expression suggested a correlation between EGFR expression and advanced tumor stage (p = 0.051, Table I). However, this trend was only the result of the low rate of EGFR-positive tumors in patients with stage III disease (Table I). In comparisons between stage I–III and stage IV the correlation between EGFR expression and advanced tumor stage became significant (p = 0.043).

Correlation of HPV-DNA detection and immunhistochemistry

The correlation between p16 expression and the prevalence of oncogenic HPV-DNA in the tumor cells was highly significant (p < 0.001, Table II). Additionally, the study results showed a trend toward an inverse correlation of the p16 and EGFR expression in terms that p16-positive OSCC showed less EGFR expression (p = 0.083); however, this trend did not reach statistical significance. No significant correlation was revealed between EGFR expression and HPV status (p = 0.125).

Table II. Correlation Between the Presence of P16 Expression and the Presence of the Parameters HPV-DNA and EGFR Expression in OSCC [Number/Total Number (%)]
 P16+ (29/96)P16− (67/96)
HPV+25/29 (86)2/67 (3)
HPV−4/29 (14)65/67 (97)
p value (HPV × P16) <0.001  
EGFR+10/29 (34.5)36/67 (54)
EGFR−19/29 (65.5)31/67 (46)
p value (EGFR × P16) 0.083  

Survival analysis

For survival analysis, 26 patients were excluded as described above, including 7 (27%) HPV-positive and p16-negative cases. HPV-containing tumors showed a tendency toward a better 5-year disease-free survival rate (74%) compared with the HPV-lacking group (53%, p = 0.083; Table III). In an analysis of overall survival, the 5-years' probability of HPV-positive tumors was 70%, compared with 51% for HPV-negative tumors, which was not significant (p = 0.231).

Table III. Univariate 5-Year Overall Survival and Disease-Free Survival Analysis for Clinical Characteristics and Markers
ParametersOverall survival (%) (95% CI)p valueDisease-free survival(%) (95% CI)p value
  1.  Twenty six patients were excluded from survival analysis because of having stage IVc disease, unresectable tumor, incomplete curative treatment, R2 resection margin or never achieving disease free status.

Stage 1–371 (52–90) 64 (43–85) 
Stage 449.5 (35–64)0.10057 (41–72)0.635
N+ limited85 (69–101) 65 (40–90) 
N+ extended42 (26–59)0.00952 (40–74)0.201
 Well50 (−19 to 119)0.33567 (13–120)0.520
 Moderately52 (36–68)0.04553 (35–70.5)0.165
 Poorly66 (50–83) 68 (52–85) 
p16+73 (54–91) 84 (67–101) 
p16−49 (33–65)0.07049 (31–66)0.009
EGFR+45 (27–63) 47 (27–66) 
EGFR−70 (56–85)0.10074 (58–90)0.084
HPV+70 (51–89) 74 (54–94) 
HPV−51 (36–65)0.23153 (37–68)0.083
EGFR+/p16−38 (17–59) 39 (16–62) 
EGFR−/p16+79 (57–100)0.01093 (79–106)0.003

Patients with p16 overexpression in their tumors had a significantly better 5-year disease-free survival rate, namely, 84% compared with 49% for patients with p16-negative tumors (p = 0.009, Fig. 2a; Table III). Although not statistically significant, patients with p16-positive tumors also tended to have better overall survival rate than those with p16-negative tumors (73% vs. 49%, p = 0.070, Fig. 2b). Better disease-free and overall survival rates were also seen for those patients who were treated only by surgery (21 cases), but reached no statistical difference in this small subgroup.

Figure 2.

Univariate survival analysis by p16 tumor status and by the combination of EGFR and p16 expression. Disease-free survival (a and c) and overall survival (b and d) curves based on the Kaplan–Meier method were analyzed.

Patients with EGFR-overexpressing tumors showed a rather poor disease-free survival rate (47%), while cases with absent EGFR-expression tended toward a better prognosis (74%, p = 0.084). Statistically similar results were obtained for the 5-years' probability for the overall survival which was 45% versus 70% (p = 0.100) for the EGFR-negative and positive cases, respectively. However, neither estimate reached statistical significance.

The combination of p16 and EGFR as predictor for survival was tested. Comparing the EGFR-positive and p16-negative patient group with the EGFR-negative and p16-positive group turned out to be significant. Of all patients 43.4% were either p16+/EGFR− or p16−/EGFR+.

For patients with EGFR+/p16− tumors, the 5-year disease-free survival rate was 39%, while the survival rate of patients with EGFR−/p16+ tumors was 93% (p = 0.003, Fig. 2c). The 5-year overall survival rate for the EGFR+/p16− group was poor (38%), compared with the EGFR−/p16+ group, which showed a significantly better outcome (79%, p = 0.010, Fig. 2d).


Using the Cox proportional hazards model, we performed a multivariate analysis to assess the independent predictive value of all markers and the promising combination of p16 and EGFR for overall- and disease-free survival. The following prognostic variables were included in the model: HPV, p16, EGFR and tumor stage (Table IV). P16 expression status was revealed as an independent and significant prognostic factor for disease-free survival in this model (p = 0.030). In terms of overall survival, p16 expression status showed a trend but was not significant (p = 0.059, Fig. 3). All other included variables were not significant (Table IV).

Figure 3.

Disease-free survival (a) and overall survival (b) in multivariate analysis. Disease-free survival and overall survival curves based on the Cox proportional hazard model were analyzed. The survival curves distinguish p16 positive and negative cases with all other parameters being equal for the cohort.

Table IV. Multivariate Overall Survival and Disease-Free Survival for Clinical Characteristics and Markers, Using the Cox Proportional Hazard Model
VariablesHazard ratio95% CIp value
  1.  n.a., not accessible.

Disease-free survival   
 hpv (−/+)0.340.06–1.850.213
 p16 (−/+)7.51.22–46.190.030
 EGFR (−/+)1.120.51–2.420.784
 Stage (1/4)n.a.n.a.0.974
 Stage (2/4)0.530.17–1.600.260
 Stage (3/4)0.650.23–1.800.404
Overall survival   
 hpv (−/+)0.420.10–1.760.235
 p16 (−/+)4.190.95–18.570.059
 EGFR (+/−)1.140.60–2.150.692
 Stage (1/4)0.97n.a.0.971
 Stage (2/4)0.060.13–1.040.060
 Stage (3/4)0.190.21–1.370.194


Oropharyngeal cancers are a heterogenious group of malignancies in terms of etiology, biological behavior and prognosis.44 Although many clinicopathological parameters have been indicated to predict prognosis, no such markers are used in clinical practice today, apart from TNM staging. However, recent insights in molecular changes underlying HNSCC development have identified new potential prognostic indicators, such as p16, EGFR and HPV status. The existing data regarding the prognostic significance of p16 in HNSCC are controversial. In carcinoma of the anterior tongue45 as well as in tonsillar carcinoma30 and OSCC,46 loss of p16 expression has been found to be associated with a worse prognosis. In contrast, other studies have found no prognostic significance of p16 inactivation in HNSCC of different localizations.47, 48 This might be due to differences in tumor biology within the different mucosal regions of the head and neck. In addition, a discrepancy exists with respect to the meaning of EGFR overexpression in OSCC. Some studies have found no correlation between survival and EGFR expression,49, 50 whereas others showed a favorable outcome for the EGFR-negative group of OSCC.51

Our previous studies and others have shown that at least one third of OSCC are infected by oncogenic HPV,6, 7, 8, 52 predominantly HPV type 16. In this study, only 28% of the tumors were HPV-positive, even though others had found a higher prevalence of HPV-positive tumors in OSCC.6, 28, 53, 54 This discrepancy could be the result of varying numbers of exact anatomical stratification of the lesions within the oropharynx and the use of different detection methods. However, since multiple and nested PCR assays were used for HPV-DNA detection in this study, it is unlikely that we missed any HPV-positive cases in our series. Serologic studies have demonstrated that HPV16-positive subjects have an increased risk for the development of oropharyngeal carcinoma.55, 56 Recently, the specific T-cell response to HPV16 E7 epitopes in subjects with HPV16 E7 expressing and p16-positive OSCC was shown.57 Evidence is growing that HPV-DNA detection in OSCC can be used as predictor for prognosis14, 58, 59 which might be attributed to a higher sensitivity to radiation therapy.15

To the best of our knowledge the current study is the first combining the most promising predictive markers HPV-DNA detection, p16 and EGFR expression in a large series of primary OSCC. We found that p16 expression is highly correlated with the presence of HPV-DNA in OSCC, which confirms our results for tonsillar cancer.6, 30 This is in line with the hypothesis that in HPV-positive tumors p16 is upregulated because of the interaction of the HPV16 E7 oncogene product with the pRb protein, which would otherwise inhibit p16 transcription. The interaction of the presence of HPV-DNA in the tumor and clear p16 overexpression, and, respectively, the absence of HPV-DNA and the loss of p16, was seen in 94 of 100 OSCC. Of this set, only 2 tumors were HPV-positive but p16-negative, and both of them occurred in strong smokers consuming more than 22 packs per year. It could be argued that exposure to this additional risk factor imposed, additional genetic or especially epigenetic changes yielding to loss of p16 which overcame the HPV16 E7 effects. Because of the lack of fresh tissue in these cases it was not possible to further analyze this possibility. However, from a clinical point of view both of these cases had a rather bad outcome with tumor recurrence after 7 months for 1 and 12 months for the other. This result would be in line with the predictive potential of p16 overexpression even in HPV-positive cases. Four cases of our batch were HPV-negative but expressed the p16 protein, and two of them did not meet the inclusion criteria for survival analysis. The 2 remaining cases were free of recurrence for 49 and 50 months. Thus, the presence of p16 expression also predicts a rather favorable outcome in these cases.

P16 expression was found to be a strong independent prognostic factor for 5-year disease-free survival and showed a tendency for better 5-year overall survival. In multivariate Cox analysis, p16 remained to be an independent prognostic factor. The prediction for disease-free survival of p16 expression was superior to all clinicopathological parameters normally used for treatment decisions and assessment of prognosis. In the current study population, patients with p16-negative OSCC had a 4-fold increase in their risk of death from any cause, and a 7.5-fold increased risk of recurring cancer, compared with patients with p16-positive tumors. Despite the increased outcome of p16-positive and HPV-positive tumors, they were more likely to be poorly differentiated. Furthermore, no correlation was detected between tumor grading and prognosis. Therefore, the data suggest that the general rule that higher histological grade implicates decreased prognosis is not the case for the p16-positive OSCC.

However, EGFR expression did not reach significant levels in either univariate or multivariate analysis. The EGFR-positive cases tended to have worse 5-year disease-free survival and worse 5-year overall survival rates compared with EGFR-negative cases. In univariate analysis, the comparison of the p16+/EGFR− group with the p16−/EGFR+ group revealed a significant difference for both 5-year disease-free survival (p = 0.003) and 5-year overall survival (p = 0.010) rates.

To our knowledge, this is the first series showing a tendency for an inverse correlation of p16 and EGFR expression in OSCC. The molecular explanation for this interaction remains unclear and is not the subject of this study. Reported interaction of HPV proteins with EGFR has shown that HPV-E5 expression leads to elevated EGFR expression in cell culture models.60 Therefore, it is likely that EGFR overexpression in our series is HPV independent.

In our series of OSCC, the multimodal treatment regime included postoperative or definitive radiation in 80% of the patients. The prognostic differences with the used immunohistochemical markers were also found in the tumors of patients treated only surgically. This subgroup was too small to allow further analysis of the involvement of the therapy chosen. Therefore, it remains unclear whether the favorable outcome of p16-positive cases was attributed to higher radiosensitivity, as recently postulated.61

As a future perspective, it might be possible to improve therapeutic outcome in tumors with inactive p16 by restoration of p16 activity or the use of cdk4 inhibitors. Such inhibitors are already used in combination with chemotherapy in clinical trials and show promising results.62, 63, 64 P16 adenovirus-mediated gene therapy has also successfully been applied against HNSCC in vivo.65, 66, 67 Concerning EGFR, several approaches for targeted therapy have been tried. Approaches with antigen specific monoclonal antibodies and small molecule kinase inhibitors look especially promising. Some of these agents have already been approved for the treatment of epithelial cancer.68 Future studies will have to show whether the marker tested in the current study will be useful to select patients for these adjuvant therapies.

In conclusion, p16 is a strong, independent prognostic indicator in patients with oropharyngeal tumors. P16 immunoreactivity is most likely the result of transcriptionally active HPV infection and therefore p16 expression is highly correlated with HPV status in OSCC. Furthermore, our data indicate that p16 may be a more powerful marker than HPV-DNA detection for predicting prognosis. However, other studies have shown that detection of HPV-DNA by PCR was a significant predictor for disease free survival as well as overall survival.14, 53, 54, 69

For clinicians, combined immunohistochemistry of p16 and EGFR is both easily applicable in a routine pathology laboratory, and it provides important prognostic information for the individual patient with more relevance than staging alone. As it was shown in some studies for OSCC, the outcome of primary radiation with salvage neck dissection was equal to surgery with postoperative radiation.70 Further studies have to show whether, for p16+/EGFR− OSCC, the appropriate treatment of the primary tumor might be radiochemotherapy alone. Therefore, these markers might not only be of prognostic significance but may also become the basis of treatment decisions in the future.