eTable 1. Using p16 as a Test for HPV in the Hypopharynx and Nasopharynx
eTable 2. Viral Association and WHO Classification in the Nasopharynx
Wilson DD, Crandley EF, Sim A, Stelow EB, Majithia N, Shonka DC, Jameson MJ, Levine PA, Read PW. Prognostic Significance of p16 and Its Relationship With Human Papillomavirus in Pharyngeal Squamous Cell Carcinomas. JAMA Otolaryngol Head Neck Surg. 2014;140(7):647-653. doi:10.1001/jamaoto.2014.821
The prognostic significance of p16 in squamous cell carcinoma (SCC) of the hypopharynx (HP) and nasopharynx (NP) and relationship between human papillomavirus (HPV) and p16 is unclear.
To evaluate the prognostic significance of p16 in pharyngeal subsites (oropharynx [OP], HP, and NP) and assess the relationship between HPV and p16 in the HP and NP.
Design, Setting, and Participants
Retrospective medical record review of 172 patients with SCC of the pharynx treated with definitive radiation therapy from 2002 to 2013 at a university tertiary referral center, with tissue available for immunohistochemical analysis. The median follow-up was 30.1 months.
A total of 118 patients were treated with chemoradiation, and 54 patients were treated with radiation alone. Immunohistochemical analysis for p16 was performed for all tumors. Hypopharynx and NP tumors were tested for HPV using in situ hybridization, and NP tumors were tested for Epstein-Barr virus.
Main Outcomes and Measures
Overall survival, locoregional control, and disease-free survival were analyzed according to p16, HPV, and Epstein-Barr virus status.
Thirty-two patients had HP SCC, 127 had OP SCC, and 13 had NP SCC. p16 Was positive in the HP (34%), OP (66%), and NP (46%). Prevalence of HPV was 14% in the HP and 50% in the NP. As a test for HPV, p16 had a positive predictive value of 38% (HP) and 67% (NP) and a negative predictive value of 100% in HP and NP tumors. p16 Status was a significant predictor of all clinical outcomes for patients with OP SCC (P<.001), but not for patients with HP or NP SCC. Patients with Epstein-Barr virus– or HPV-associated NP SCC had improved clinical outcomes.
Conclusions and Relevance
p16 Was not associated with improved outcomes in patients with HP or NP SCC. The positive predictive value of p16 as a test for HPV is too low for p16 testing alone in the HP and NP. However, p16 negativity is sufficient to rule out HPV. As a research approach, we recommend p16 immunohistochemistry as a screening test for HPV in NP SCC and HP SCC followed by confirmatory HPV in situ hybridization when p16 positive.
Investigators have reported the prognostic significance of multiple biomarkers for patients with head and neck squamous cell carcinoma (SCC). The presence of human papillomavirus (HPV) is a strong and independent predictor of prognosis in oropharyngeal (OP) SCC. A meta-analysis and prospective clinical trials have demonstrated that patients with HPV positive (+) OP SCC have a 28% to 58% reduced risk of death compared with patients with HPV negative (−) OP SCC.1,2 A similar effect on prognosis has been found in OP SCC with p16, a cyclin-dependent kinase inhibitor whose expression in OP SCC has been significantly correlated with HPV status, improved overall survival (OS), and locoregional control (LRC).1- 4 The difference between the response of HPV+/p16+ and HPV−/p16− tumors is so statistically significant that current clinical trials are investigating if treatment of OP SCC can be altered based on HPV-p16 status.
In contrast, little is known about the significance of HPV and p16 in SCC of pharyngeal subsites outside of the OP. A small number of studies have found HPV and p16 to be present in some hypopharyngeal (HP) and nasopharyngeal (NP) SCCs,5- 7 but the prognostic significance of this association remains unclear. Specifically, it has not been established that the predictive value of p16 for HPV status and the improved prognosis for HPV+/p16+ tumors observed for OP SCC can be extrapolated to HP or NP SCC.
The present study evaluates the correlation between p16 and HPV and their prognostic significance for patients with SCC of the pharynx (NP, OP, and HP) treated with definitive intensity-modulated radiation therapy.
Data were retrospectively collected on a protocol approved by the institutional review board of Health Sciences Research. Prior to treatment, patient care was discussed by members of a multidisciplinary head and neck cancer tumor board. All patients in this analysis were treated with definitive intensity-modulated radiation therapy for histologically confirmed pharyngeal SCC. Patients with a history of neck dissection or primary resection before irradiation, distant metastases at diagnosis, or absent biomarker data were excluded from the study.
A total of 172 patients were eligible by clinical criteria and had associated biomarker data and thus were included in the analysis. Of the 172 patients, 118 (69%) received chemotherapy and radiotherapy and 54 (31%) were treated with radiation alone. Most patients who received chemotherapy received either induction chemotherapy with paclitaxel-platinum-fluorouracil (TPF) and/or concurrent carboplatin or cisplatin. Neck dissections were performed on patients with bulky (>4 cm in diameter) nodal disease regardless of response to radiation and on patients with apparent residual nodal disease on neck computed tomography (CT) at 4 to 6 weeks after completion of radiation.
An aquaplast mask was used for immobilization, and CT simulation was performed. The gross (macroscopic) tumor volume (GTV) comprised the primary tumor and pathologically involved lymph nodes based on pretreatment imaging. The clinical target volume (CTV) included the GTV with margin for microscopic spread and the draining cervical lymph node levels at risk for microscopic spread. A 3-mm expansion of the CTV accounted for setup uncertainties to create the planning treatment volume (PTV). This expansion was based on published data and a retrospective institutional analysis of mean head and neck SCC patient setup inaccuracies determined by pretreatment CT guidance.8
Patients received a dose to the primary tumor of 66 to 70 Gy in 30 to 35 fractions with 3 exceptions. One patient with HP SCC did not respond after 50 Gy and underwent immediate surgical salvage. Another patient with NP SCC was treated to 60 Gy in 30 fractions to the NP primary without elective nodal irradiation because she had previously received 70 Gy to the soft palate for an OP SCC. A final patient with a T1N1 tonsil SCC was treated to 64.4 Gy to the primary and node. With the above exceptions, 66 to 70 Gy was delivered to the pathologic lymph nodes. Clinically uninvolved cervical lymph nodes were treated to 45 to 54 Gy. Intensity-modulated radiation therapy plans were optimized so that at least 95.0% of the PTV received the prescribed dose.
Archived formalin-fixed and paraffin-embedded pretreatment primary tumor biopsy specimens were available for all 163 patients. Immunohistochemical (IHC) analysis was performed using a DAKO Autostainer with primary antibodies for p16 (BD Biosciences 550834; titration: 1:100; pressure retrieval). Automated in situ hybridization (ISH) for high-risk HPV and EBV was performed using the Ventana HR HPV III and Ventana EBER1 DNP probes, respectively, on specimens from the NP and HP when sufficient tissue was available, with EBV only tested in the NP. For p16 IHC analysis, a cervical specimen with a high-grade squamous intraepithelial lesion served as the positive control, and uninvolved glandular and squamous epithelium in the cervical specimen served as the negative control. p16 Was considered positive if strong nuclear and cytoplasmic staining was present in more than 60% of tumor cells, as previously described.2 Human papillomavirus ISH was interpreted as positive in the presence of a punctate staining of tumor cell nuclei, and EBV ISH was interpreted as positive in the presence of strong staining of tumor cell nuclei.
The following definitions were used for statistical analyses: OS was the time between the date of diagnosis and the date of death from any cause; LRC was the time between the date of diagnosis and the date of first local or regional recurrence; and disease-free survival (DFS) was the time between the date of diagnosis and the date of first disease recurrence or death from any cause. Patients who were living and were without evidence of recurrence at the time of analysis were censored at the date of last follow-up. The t test and Fisher exact test were used to assess differences in patient and tumor characteristics. The log-rank test was used to compare Kaplan-Meier survival curves. Predictors with a P value <.10 on univariate Kaplan-Meier analysis were included in a multivariate Cox regression model. P < .05 was considered statistically significant. Statistical calculations were performed using SPSS (Version 20.0.0; SPSS Inc) and SAS (Version 9.2; SAS Institute Inc) statistical software.
Characteristics of patients and their tumors are given in Table 1. A total of 172 patients met the inclusion criteria. Of the tumors in the 172 patients, 101 (59%) were p16+, while 71 (41%) were p16−. Only 67% of patients with p16+ tumors had a history of tobacco use, in contrast with 94% of patients with p16− tumors (P < .001). In addition, only 50% of patients with p16+ tumors had a history of alcohol use, compared with 75% of patients with p16− tumors (P < .001). With respect to pharyngeal subsites, 34% of HP SCC tumors were p16+ (11 positive vs 21 negative), while 66% of OP SCC tumors were p16+ (84 positive vs 43 negative) and 46% of NP SCC tumors were p16+ (6 positive vs 7 negative). The distribution of pharyngeal subsites between p16+ and p16− tumors was significantly different (P = .003). In addition, there were significantly more basaloid differentiated tumors in the p16+ group than in the p16− group (28% vs 7%; P = .001). There were no significant differences in sex, cancer stage, tumor site, or age between patients with p16+ and p16− tumors. Median follow-up for the entire cohort was 30.1 months (range, 2.4-129.7 months). For all cases, p16 staining was dichotomous, indicating there were no difficult or borderline cases.
As given in Table 2, there was a statistically significant difference in 3-year LRC (93% vs 79%; P = .01), 3-year DFS (86% vs 61%; P < .001), and 3-year OS (90% vs 62%; P < .001) for patients with p16+ vs p16− tumors, respectively. In addition, primary tumor stage was also significantly predictive of LRC (P = .005), DFS (P = .02), and OS (P = .04). However, neither overall stage nor nodal stage was a significant predictor of LRC, DFS, or OS. Smoking was also significantly predictive of LRC (P = .04) and OS (P = .03) but not DFS. Alcohol use was predictive of reduced LRC, DFS, and OS as well. Patients younger than 70 years had improved DFS, and male patients had improved DFS and OS over female patients. Patients who received chemotherapy also had improved LRC, DFS, and OS over those who did not receive any chemotherapy. On multivariate analysis, age, p16 status, T stage, and chemotherapy use were all independently predictive of DFS, as demonstrated in Table 3.
As given in Table 4, when the clinical outcomes were stratified by pharyngeal subsite, p16 was a significant predictor of outcomes for OP SCC, but p16 did not predict for significantly different outcomes in HP SCC or NP SCC. There were 14 patients (11%) with locoregional recurrences in the OP SCC group, 2 (15%) in the NP SCC group, and 5 (16%) in the HP SCC group.
Of the HP SCC samples, 21 had sufficient tissue available for HPV ISH, of which 3 (14%) were positive for HPV. Of the 13 NP SCC tumors analyzed, 4 were associated with HPV, 5 with EBV, and 4 had no evidence of viral association. The 5 EBV+ tumors were not associated with p16 or HPV. p16 Status did not have prognostic significance in NP SCC, but viral association predicted improved outcomes. As given in Table 5, association with HPV or with EBV in NP SCC resulted in improved outcomes, but these were not statistically significant. However, viral association (with EBV or HPV) was a significant predictor of increased LRC (P = .03), DFS (P = .03), and OS (P = .03). These data are based on small numbers because 2 of the 13 patients with available data had a locoregional failure and subsequent death.
p16 Had different predictive values for HPV depending on the pharyngeal subsite. On the basis of samples with available tissue for HPV ISH, HPV prevalence was 14% in the HP and 50% in the NP. When p16 was used as a test for HPV in HP SCC, p16 had a sensitivity of 100%, a specificity of 72%, a positive predictive value of 38%, and a negative predictive value of 100% (eTable 1 in Supplement). When p16 was used as a test for HPV in NP SCC, p16 had a sensitivity of 100%, a specificity of 50%, a positive predictive value of 67%, and a negative predictive value of 100% (eTable 1 in Supplement). The predictive value of p16 for HPV in OP SCC is well established1,4 and was not examined in this patient series.
Nasopharyngeal SCC tumors characterized according to the World Health Organization classification are presented in eTable 2 in the Supplement and stratified by HPV and EBV association. Tumors associated with EBV were mostly undifferentiated nonkeratinizing SCC (80%), while HPV-associated tumors were mostly differentiated nonkeratinizing SCC (75%).
In the present study, we investigated the correlation between p16 and HPV and the prognostic significance of p16 in patients with SCC of the pharynx, subdivided into HP, OP, and NP, who were treated with definitive intensity-modulated radiation therapy. This study is unique in that it includes a comparison between HPV and p16 in tumors of patients with SCC of non-OP subsites.
The correlation between p16 and HPV in OP SCC has been so close that p16 has been used as a surrogate IHC marker for HPV.1,9 The expression of p16, or lack thereof, has been found to be superior to all other clinical or IHC parameters for predicting prognosis for patients with OP SCC.9,10 We found a similar prognostic effect associated with p16+ OP SCC at our institution and previously reported an approximate 25% absolute improvement in outcomes for patients with p16+ vs p16− OP SCC tumors.3 When compared with patients with p16− tumors, patients with p16+ tumors had a significantly improved 3-year locoregional progression-free survival (97.8% vs 73.5%; P = .006), increased 3-year DFS (88.2% vs 61.4%; P = .004), and increased 3-year OS (87.8% vs 47.6%; P < .001).3
Table 2 demonstrates improved LRC, DFS, and OS in this study for patients with p16+ pharyngeal tumors. This is most likely because most patients in this cohort had OP SCC, for which outcomes are heavily influenced by p16 status. This may also explain why there was no significant difference in outcomes based on overall stage and nodal stage, as p16+ OP cancers frequently have higher nodal stage and overall stage. Similarly, p16+ OP SCC are more likely to have lower primary tumor stages and are likely contributing to the better prognosis with lower primary tumor stage.
The prognostic significance of HPV in subsites outside of the OP is unclear. While data from multiple studies indicate that HPV+ tumors are found in 38% to 64% of OP SCC,1,3,11,12 far lower percentages of HP SCC tumors (7%-29%) and NP SCC tumors (9%-53%) are reported to be HPV+.7,11,13- 17 It is difficult to interpret the prognostic significance of HPV positivity in tumors occurring outside of the OP. Such an analysis could only be based on various heterogeneous studies that have included patients with primary tumors from multiple head and neck cancer subsites and treated patients with various combinations of surgery, radiation, and/or chemoradiation to their primary tumors. We found a very low rate (14%) of HPV+ HP cancers. Our data are consistent with subset analyses from other series that have reported HPV+ rates of 7% to 29% in the HP.7,13,14
We report an HPV prevalence of 50% in NP SCC, which is consistent with previous reports in non-Asian populations.15,17 There was an insufficient number of patients with HPV+ tumors in our series to comment on the prognostic significance of HPV alone in the HP or NP. However, by combining HPV and EBV data in the NP, we found that patients with viral-associated NP SCC (with either EBV or HPV) had significantly improved LRC, DFS, and OS. Notably, this observation is based on a small group of 13 patients with 2 treatment failures; these data must be interpreted cautiously. Of the 2 patients who had locoregional failures and subsequent deaths in the NP SCC cohort, neither patient had a tumor with viral association. One patient had keratinizing SCC and the other had differentiated nonkeratinizing SCC. These histologic subtypes both have a poorer prognosis than the undifferentiated nonkeratinizing SCC that is closely associated with EBV. Therefore, we cannot distinguish whether the unfavorable histologic subtype, lack of a viral association, or both explain the poorer survival in the NP SCC cohort without a viral association. However, our results are supported by findings from Dogan et al,5 who also demonstrated that patients with either HPV+ or EBV+ NP SCC have improved survival over patients with HPV−/EBV− NP SCC.
One potential anatomic mechanism of the association between HOV and HP SCC could be tumor initiation in the lymphoid tissue of the inferior pole of a tonsil or base of tongue followed by tumor progression inferiorly along the pharyngeal wall into the HP. The association between HPV and NP SCC could come from direct extension from the superior tonsillar pole superiorly into the NP or from the adenoids directly.
In the present analysis, 84 of 127 patients (66%) with OP SCC had p16+ tumors, which is consistent with subset analyses from other series that demonstrated p16+ rates in the oropharynx of 38% to 64%.1,2,4,9 In addition, 11 of 32 patients (34%) with HP SCC had p16+ tumors, which is consistent with subset analyses from other series that demonstrated p16+ rates in the HP of 7% to 28%.7,18,19 Lastly, 6 of 13 patients (46%) with NP SCC had p16 + tumors. The prevalence of p16+ tumors in the NP varies widely from 9% to 80%15,20 and may vary based on ethnicity. Most importantly, patients in this series with p16+ tumors in the HP and the NP did not have better outcomes than patients with p16− tumors in these subsites; thus, the significant differences in outcomes based on p16 status observed in patients with OP SCC were not seen in patients with HP SCC and NP SCC. Wendt et al7 recently published the largest series to date analyzing p16 expression in the HP, examining 109 patients with HP SCC treated by various modalities, and similarly found p16 to have no prognostic significance in this subsite. Finally, the present series found primary tumor stage and tobacco use to be predictive of recurrence and survival, which is consistent with other studies.21- 24
When used as a test for HPV in the OP, p16 has a sensitivity of nearly 100% and a specificity of approximately 79%.25 As seen in eTable 1 in the Supplement, we found p16 to have a sensitivity of 100% and a specificity of 72% when used as a test for HPV in the HP and a sensitivity of 100% and specificity of 50% in the NP. Though p16 has a similar sensitivity and specificity in these subsites to that of the OP, the prevalence of HPV in our series (HP, 14%; and NP, 50%) is much lower than the HPV prevalence in the OP (approximately 60%).1 The lower prevalence of HPV in the HP and NP results in lower positive predictive values for p16 in these sites (positive predictive value, 38% in HP SCC and 67% in NP SCC). This may be the reason why p16 does not appear to be closely correlated with HPV in this series, and it may explain why p16 status did not demonstrate prognostic significance in patients with HP and NP SCC. However, p16 had a negative predictive value of 100% in both the HP and NP, and thus a negative p16 value is sufficient to rule out the presence of HPV in these subsites.
Though HPV+/p16+ OP SCC has demonstrated increased radiation responsiveness, the exact mechanisms remain unknown. Data from our cohort and others suggest that patients with virally associated NP SCC may also have increased chemoradiation sensitivity, though p16 positivity alone did not confer increased chemoradiation sensitivity in NP or HP SCC.
The primary limitations of this study are its retrospective nature, small sample size, varied chemotherapy regimens, and the limited biomarker availability for some patients. Thus, it is possible that p16 could have a small effect on prognosis in HP and NP SCCs that our sample size was not large enough to detect. The small sample size is a particular limitation in making conclusions concerning the NP and HP subsites, though we believe the data to be hypothesis generating. In addition, our assay for HPV detection used ISH, which is less sensitive than polymerase chain reaction and reverse transcriptase–polymerase chain reaction. However, ISH is the method used at our center and at most other cancer centers, which makes our results generalizable. The strengths of this study are that all patients in the cohort had biopsy-proven SCC and were treated with intensity-modulated radiation therapy without primary surgical resection, allowing us to more accurately comment on the prognostic factors associated with radiation sensitivity of cancers in these subsites.
Clearly, there are limited data regarding the prognostic significance of p16 and HPV in head and neck subsites other than the OP. One would assume that the favorable prognostic significance of HPV and p16 would hold at other subsites because it has been demonstrated in multiple studies to confer favorable outcomes for patients with OP SCC. However, in our patients with HP and NP SCC, p16 and HPV correlate poorly and p16 is not a prognostic biomarker. Human papillomavirus status may still be predictive of outcomes, although fewer patients with HP SCC or NP SCC appear to have HPV-associated tumors compared with OP SCC.
Furthermore, p16 IHC is a rapid, easy, inexpensive, and reliable test that has a very high sensitivity and high negative predictive value for HPV in both the HP and NP. Thus, as a research approach p16 can be used as a screening test for HPV in these subsites with a lower prevalence of HPV to rule out the presence of HPV. Further confirmatory testing for HPV with ISH can be performed if the specimen is p16+ to establish if the tumor is virally associated. Because EBV+ NP SCC has been shown to consistently be p16− in our series and others,5,15 only p16− NP SCC may merit further testing for EBV.
In contrast to OP SCC, p16 expression in patients with HP SCC and NP SCC had a low positive predictive value for HPV and did not predict improved OS, LRC, or DFS. Thus, for HP SCC and NP SCC, p16 is not a prognostic biomarker, and the prognostic significance of HPV outside of the OP remains unclear. However, viral association may predict OS, LRC, and DFS in NP SCC. Caution must be taken when extrapolating the prognostic significance of p16 expression in patients with OP SCC to patients with head and neck SCC of other mucosal subsites with low HPV infection rates. On the basis of its high sensitivity and high negative predictive value for HPV, for future research, p16 IHC may be used as a screening test for HPV status in HP SCC and NP SCC, followed by confirmatory HPV ISH for p16+ tumors.
Submitted for Publication: November 4, 2013; final revision received March 31, 2014; accepted April 10, 2014.
Corresponding Author: Paul W. Read, MD, PhD, Department of Radiation Oncology, University of Virginia, PO Box 800383, Charlottesville, VA 22908 (firstname.lastname@example.org).
Published Online: May 29, 2014. doi:10.1001/jamaoto.2014.821.
Author Contributions: Dr Wilson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Wilson, Read.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Wilson, Read.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Wilson, Majithia.
Administrative, technical, or material support: Majithia, Levine, Read.
Study supervision: Shonka, Jameson, Levine, Read.
Conflict of Interest Disclosures: None reported.
Previous Presentation: This study was presented as a poster presentation at the American Society for Radiation Oncology 2012 national meeting; October 28-31, 2012; Boston, Massachusetts.