- Email: [email protected]
p16INK4a Expression, human papillomavirus, and survival in head and neck cancer
Oral Oncology (2008) 44, 133– 142
available at www.sciencedirect.com
journal homepage: http://intl.elsevierhealth.com/journals/oron/
p16INK4a Expression, human papillomavirus, and survival in head and neck cancer Elaine M. Smith a,*, Donghong Wang a, Yoonsang Kim b, Linda M. Rubenstein a, John H. Lee c, Thomas H. Haugen d, Lubomir P. Turek d a
Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA 52242, United States Department of Biostatistics, College of Public Health, University of Iowa, Iowa City, IA 52242, United States c Department of Otolaryngology, College of Medicine, University of Iowa, Iowa City, IA 52242, United States d Veterans Affairs Medical Center and Department of Pathology, College of Medicine, University of Iowa, Iowa City, IA 52242, United States b
Received 13 November 2006; accepted 15 January 2007 Available online 13 March 2007
KEYWORDS
Summary Development of head and neck cancer (HNC) is associated with human papillomavirus high-risk (HPV-HR) types. The HPV E7 oncoprotein inactivates the pRB protein increasing expression of p16INK4a. p16 Expression and HPV status have been associated with differences in clinical outcomes for HNC. This study examined whether HNC prognosis was different when these biomarkers were examined as individual or joint molecular effects. Tumor tissue from 301 HNC patients were analyzed and sequenced for HPV types. p16 was evaluated by immunohistochemical staining. p16 was expressed in 35% and HPV-HR was detected in 27% of HNC patients. After adjustment for age, tobacco, and alcohol, p16+ tumors were statistically significantly associated with HPV-HR (OR = 13.3, 7.1–24.9), histology, stage, grade, tumor site, and node involvement. Compared to p16+ HNC cases, those who did not express p16 had significantly worse disease-specific (DS) survival (Hazards Ratio, adj.HR = 2.0. 1.0–3.9) and recurrence (adj.HR = 3.6, 1.6–8.2); and HPV cases had worse DS survival (adj.HR = 2.8, 1.1–7.1) and recurrence (adj.HR = 2.0, 0.8–4.8) compared to HPV-HR patients. Each of the p16/HPV groups had different survival outcomes: p16+/HPV-HR cases (referent) had the best and p16 /HPV cases had the worst DS survival (adj.HR = 3.6; 53% versus 13%, p = 0.004) whereas p16+/HPV and p16 /HPV-HR had similar DS survival (adj.HR = 2.7/2.8). p16 /HPV-HR cases
Head and neck cancer survival; HPV; p16INK4a.
* Corresponding author. Tel.: +1 319 384 5014; fax: +1 319 384 5031. E-mail address: [email protected] (E.M. Smith).
1368-8375/$ - see front matter c 2007 Published by Elsevier Ltd. doi:10.1016/j.oraloncology.2007.01.010
134
E.M. Smith et al. had a worse recurrence rate (adj.HR = 7.0; 60% versus 0%, referent, p = 0.08) than p16 /HPV (adj.HR = 4.5) or p16+/HPV (adj.HR = 1.8) cases. The combined p16/HPV biomarker data are associated with different survival outcomes of HNC compared to each marker evaluated separately, indicating that the two molecular mechanisms evaluated together may provide a more accurate prediction of clinical outcomes. c 2007 Published by Elsevier Ltd.
Introduction Infection with HPV high-risk (HR) types is associated with the development of 25% of head and neck cancers (HNC) independent of tobacco and alcohol consumption.1–3 The HPV E7 oncoprotein inactivates the pRB tumor suppressor protein. The abrogation of pRB function is thought to be responsible for a prominent rise in the levels of p16INK4a, a cyclin-dependent kinase inhibitor, by the loss of negative feedback control.4 Expression of p16INK4a in association with HPV-HR infection has been demonstrated in a high percentage of high-grade cervical dysplasias and cancer.5,6 It has been suggested that p16INK4a expression may serve as either a surrogate biomarker of HPV oncogenic infection or play an independent role in predicting cervical cancer.6,7 However, the signals and pathways that determine p16INK4a expression in humans in general8 or p16INK4a upregulation in HPV-unrelated tumors are not well understood. The role of p16INK4a as a marker of HPV-associated head and neck tumors is less clear than it is for cervical cancer. Several small9–11 but no large-scale studies have examined the association between p16 expression as a marker for HPV-HR in HNC and because of their limited sample size they were unable to evaluate other covariates that might influence that association. p16INK4a might also serve as a marker to distinguish integrated and nonintegrated oncogenic HPV or nononcogenic HPV detected in HNC tumors. However, current findings regarding the prevalence of p16 expression observed in HNC vary widely, between 44% and 80%.12–14 Some12,13,15 but not other16 investigations also have found that p16 expression is a marker of better survival for HNC patients. In this study we evaluated a large number of HNC cases for differences in p16 expression and HPV-HR status in tumor cells and their interrelationship with HNC cancer controlling for other HNC risk factors and tumor characteristics. These two biomarkers also were evaluated to determine whether prognostic outcomes associated with HNC were different and more highly predictive when examined individually or in combination.
the time of the study analyses. After patients signed an approved Human Subjects form, they completed a selfadministered risk factor questionnaire regarding sociodemographics, tobacco habits, alcohol consumption, sexual practices, medical history including HPV-related diseases and oral lesions. Information regarding prior HNC history, tumor site, histology, stage, grade, node involvement, and treatment were identified from pathological reports and medical records.
Tumor specimens Hematoxylin and eosin-stained slides (H&E) from formalinfixed, paraffin-embedded tumor blocks were evaluated to verify the presence of malignant cells and tumor grade. Laser capture microdissection (LCM) was performed to verify that HPV DNA detected from the whole tissues was from the tumor cells. Procedures have been described previously1,17 and were performed when tissue specimens had less than 10% of tumor tissue or when they were initially found to contain an HPV nononcogenic, low-risk type. All histologic types of HNC were included in the study and were classified as squamous cell carcinoma (SCC) or nonSCC. Most tumors were SCC (92%) with a small percent of other histologic types: seven mucoepidermoid carcinomas, four adenoid cystic carcinomas, three acinar cell carcinomas, three adenocarcinoma, two verrucous carcinoma, one basal cell adenocarcinoma, one carcinoma, NOS, one polymorphous adenocarcinoma, one carcinoma in pleomorphic adenoma, and 1 mixed tumor, NOS. For each case, 1–2 blocks were selected with the highest percentage of tumor for HPV analysis and p16 immunohistochemical staining (IHC). After a block was selected, serial 4 lm thick sections were cut. Tumor staging was based on the 1997 American Joint Committee on Cancer (AJCC) criteria18 and tumor sites were grouped into oral cavity, oropharynx, and larynx/ hypopharynx (excluding nasopharynx) according to the AJCC Cancer Manual.18
p16 Immunohistochemical staining (IHC)
Methods Data collection The study included 301 histologically confirmed HNC cases recruited between 1994 and 2004 from the Department of Otolaryngology, University of Iowa Hospitals and Clinics, for whom p16 immunohistostaining had been performed at
Sections for p16 staining were deparaffined by passage through xylene and then dehydrated in graded alcohol. Two slides from the same cut series of the same block were used. One was stained by p16 mouse anti human monoclonal antibody E6H4 from CINtecä p16INK4a histology kits (Dako Corporation, CA); another slide was stained by monoclonal mouse IgG to Aspergillus niger gluxose oxidase from the same kit to verify specificity of staining observed with
p16INK4a and survival in HPV–Associated HNC monoclonal antibody. One known positive and negative p16 slides were used as controls in each assay. Before antibody staining, epitope retrieval was done by 0.1 mol/L citrate buffer in a steamer for 10 min. After cooling and rinsing through buffer, tissue sections were quenched of endogenous peroxidase by 3% hydrogen peroxide containing 15 mmol/L sodium azide. After rinsing, 1:25 dilute monoclonal antibody of p16 was applied to one slide and negative control reagent to another slide from the same block and then incubated 30 min at room temperature. The slide was then covered with 200 ll of visualization reagent consisting of secondary goat-anti-mouse immunoglobulin molecules and horseradish peroxidase molecules linked to a common dextran polymer backbone, eliminating the need for sequential application of link antibody and peroxidaseconjugated antibody. Localization of antigen was visualized using the chromogen DAB for 10 min. Finally, tissue sections were counter-stained in dilute hematoxylin, dehydrated in graded alcohol, cleared in xylene, and mounted. All p16 IHC slides were read by one observer (THH) and 10% were re-reviewed. Strong nuclear and cytoplasmic staining defined a positive reaction. P16 positive staining was limited to the tumor cells, occassionally including a few multinucleate giant cells which stained strongly positive, and less than 5% epithelium had focal positive stain. The p16 positive distribution was scored on a semiquantitative scale at first and then converted to positive and negative scores before statistical analyses. The scale was defined as: 1 = negative for staining; 2 = slightly positive for staining, 3 = 10–80% positive; 4 = > 80% positive; 5 = every tumor cell positive. Slides in categories 1 was defined as negative, and 2–5 were positive. Categories 2 (n = 2) and 3 (n = 9) were included as positive for IHC. Their inclusion did not alter the statistical findings in this large sample size (data not shown). Positive reaction of nontumor cells, such as multinucleated giant cells and epithelia, were not included as positive.
HPV detection and typing HPV detection and typing from paraffin-embedded tissue was performed as described elsewhere.19 Briefly, DNA isolation was performed by protocol from Qiagen QIAamp DNA mini kits. HPV detection was performed by PCR with MY09/11 primers and dot blot hybridization. Dot blot positive samples were further amplified by a second round of PCR with primers of GP5+/GP6+ or GP5+/MY09 and positive samples were typed by automated sequencing.
Statistical methods Risk factors and demographic characteristics included age, gender, education, tobacco, alcohol, number of sexual partners, oral–genital sex history, tumor site (oral cavity, oropharynx, larynx/hypopharynx), stage, grade, nodal involvement, and histology (SCC versus all others). Age, tobacco, and alcohol use were examined as both categorical and continuous variables in all analyses. Tobacco exposure included cigarettes, cigars, pipes, chewing tobacco, and snuff. Total cumulative exposure for an individual was expressed as pack-years (1 pack-year = 1 pack of cigarettes/
135 day for 1 year with 20 cigarettes/pack). Alcohol assessment included exposure to beer, wine, and liquor (one drink equivalent to 12-oz can/bottle of beer = 4-oz glass of wine = 1.5-oz shot of liquor) and dose-duration was calculated as the average number of drinks/week by number of years use. Both alcohol and tobacco use were also assessed as current, former, and never users. Former use was defined as not having used the substance for at least 1 year prior to diagnosis and never users defined as never having used the substance for 1 year or more during their life. The unadjusted relationships between continuous and categorical variables were assessed with the Wilcoxon rank sum test. Unadjusted associations between categorical variables were tested with either the Pearson chi-square test or Fisher’s exact test for small sample sizes. We also tested for possible interactions among risk factors. Tests for interactions were generated using both logistic regression and proportional hazards regression methods. Multivariate unconditional logistic regression was used to examine the association between dichotomous dependent variables and risk factors for HNC (e.g. age, gender, education, tobacco and alcohol exposure, and sexual behavior and history). Odds ratios (OR) and 95% confidence intervals (CI) were calculated to measure the magnitude of the associations. Survival and recurrence were measured in years from the date of diagnosis until death or until the patient was last known to be alive. In disease-specific survival analyses, patients who died of causes other than HNC were classified as censored observations. In the disease free patients, recurrence in patients was measured in years from the date of diagnosis until recurrence, while time for censored patients was measured in years from the date of diagnosis until the patient died or was last known to be alive. Dates of death or dates last known to be alive were available from the Iowa NCI SEER Cancer Registry,20 the university hospital tumor registry follow-up, and the National Death Index.21 Three sets of analyses were performed: overall survival, disease-specific survival, and time to recurrence. P16 and HPV were modeled independently and then together across all three sets of analyses of clinical outcomes. In disease-specific analyses, patients who died of causes other than HNC were classified as censored observations. Survival curves were estimated by the Kaplan–Meier method22 and the difference between curves was tested by the log-rank test.23 All survival curves were generated in S-PLUS 6.2.24 Cox proportional hazards models25 were used to assess the relationships between p16, HPV and time to death or recurrence, and to adjust for factors previously found to be prognostically significant in HNC.26 These analyses included age (continuous), tobacco, alcohol, stage (0/I/II: referent), and treatment (no treatment: referent). Tumor site (oral cavity, oropharynx, and larynx/ hypopharynx as the referent) also were examined in separate models. All variables were assessed for the proportional hazards assumption27 before inclusion in the models. P16 and HPV status were examined separately and then in four groups: p16+/HPV-HR (referent), p16+/HPV , p16 /HPVHR, p16 /HPV . The magnitude of the associations was assessed with hazard ratios (HRs) calculated from the Cox regression models. Statistical significance was based on two-tailed tests and p-values 60.05. Statistical analyses were performed using SAS 9.1.28
136
E.M. Smith et al.
Results P16 and HPV biomarkers and risk factors Among the 301 HNC cases, 57% were in the oral cavity, 30% in the oropharynx, and 13% in the larynx/hypopharynx. P16 was expressed in 35% and the risk of p16+ was higher in the oropharynx (65%) or oral cavity (25%) compared to the larynx (12%; Table 1). HPV-HR was detected in 27% of tumors: 95% HPV-16, 4% HPV-33, and 1% HPV-18. The most frequent site detected with HPV was the oropharynx (57%) followed by the oral cavity (15%), and larynx (10%). When HPV status was examined in association with p16, positive concordance (p16+/HPV-HR) was 21% and negative concordance (p16 /HPV ) was 59%, whereas 14% with p16+ expression were HPV and 6% of cases with HPV-HR did not express p16. Oropharyngeal cancer, clinicopathologic characteristics, and HPV-HR were significantly associated with p16 overexpression after adjustment for age, tobacco, and alcohol (Table 1). When HPV-HR and other pathologic characteristics were added to the adjusted model in Table 1, there was a drop in the odds ratio associated with p16+ status in tumors of the oropharynx but not the oral cavity compared to the larynx. The risk remained statistically significantly elevated in the p16+ patients for stage IV but not stage III, and other histologic types of HNCs showed a greater increased risk compared to SCC HNC histology. HPV-HR remained significantly associated with p16+ after controlling for risk factors and pathologic characteristics.
Table 1
Table 2 shows the risk of p16 status associated with potential HNC risk factors and confounders. After inclusion of HPV status in the adjusted ORs, male gender and younger age, although elevated, were no longer statistically significantly associated with p16 expression whereas greater number of lifetime sex partners (P4) remained significant. P16 expression was not associated with alcohol or tobacco use.
Survival Data for both p16 and HPV status were available in the survival models excluding three cases due to unknown cause of death and 35 who were never disease free. There were no differences in the odds ratios for histologic and risk factor characteristics when data were limited to patients with follow-up data on clinical outcomes compared to those without this information (data not shown). The median time to death or last follow-up was 1.8 years (range: 0.03–9 years). Adjusting for age, tobacco, alcohol, stage, treatment, histologic type, and tumor site, cases with p16+/HPV-HR (referent group) were significantly more likely to be detected in late stage disease (III or IV; p 6 0.001) and with tumors in the oropharynx (p < 0.0001) compared to each of the other three p16/HPV groups (p16 /HPV , p16+/HPV , p16 /HPV-HR; data not shown). The p16+/HPV-HR group also was more likely to be a former or current tobacco or alcohol user than the p16+/HPV (tobacco: p = 0.001, alcohol: p = 0.007), or p16 /HPV (p = 0.008, p = 0.07) groups, and to be younger (655 years) than each of the three other
Clinicopathologic characteristics of HNC by p16 status
Characteristics
p16 , N = 196 (%)
p16+, N = 105 (%)
Adjusted ORa (95% CI)
Adjusted ORb (95% CI)
Site Larynx/Hypopharynx Oral Cavity Oropharynx
36 (18.4) 128 (65.3) 32 (16.3)
5 (4.8) 42 (40.0) 58 (55.2)
1.0 2.5 (0.9–7.1) 12.5 (4.3–36.4)
1.0 2.4 (0.8–7.5) 5.7 (1.7–18.3)
77 (39.5) 31 (15.9) 87 (44.6)
18 (17.1) 18 (17.1) 69 (65.7)
1.0 2.5 (1.1–5.4) 3.3 (1.8–6.0)
1.0 1.5 (0.6–3.9) 2.4 (1.2–4.9)
Grade Well/moderately Poorly/Undifferentiated
154 (81.5) 35 (18.5)
65 (64.4) 36 (35.6)
1.0 2.4 (1.4–4.2)
1.0 1.9 (0.9–3.8)
Nodal involvement No Yes
122 (62.6) 73 (37.4)
37 (35.2) 68 (64.8)
1.0 2.9 (1.8–4.9)
1.0 1.1 (0.5–2.4)
Histology SCC Other
185 (94.4) 11 (5.6)
92 (87.6) 13 (12.4)
1.0 2.5 (1.03–6.0)
1.0 5.6 (2.1–15.1)
HPV status HPV-HR HPV
19 (9.8) 175 (90.2)
62 (59.0) 43 (41.0)
13.3 (7.1–24.9) 1.0
9.3 (4.7–18.5) 1.0
Stage 0/I/II III IV
a b
Adjusted for age, tobacco (pack-years), and alcohol (drinks/week). Adjusted for age, tobacco (pack-years), alcohol (drinks/week), tumor site, stage, grade, nodes, histology, and HPV status.
p16INK4a and survival in HPV–Associated HNC Table 2
137
Demographic and risk factor characteristics of HNC by p16 status
Characteristics
p16 , N = 196 (%)
p16+, N = 105 (%)
Adjusted ORa (95% CI)
Adjusted ORb (95% CI)
Gender Male Female
113 (57.7) 83 (42.3)
75 (71.4) 30 (28.6)
1.9 (1.1–3.2) 1.0
1.5 (0.8–2.8) 1.0
Age group 655 >55
68 (34.7) 129 (65.8)
50 (47.6) 55 (52.4)
1.8 (1.1–2.9) 1.0
1.5 (0.9–2.7) 1.0
Number of partners 63 P4
105 (57.4) 78 (42.6)
33 (33.7) 65 (66.3)
1.0 3.0 (1.7–5.4)
1.0 2.7 (1.4–5.4)
Oral-genital contact No Yes
61 (54.5) 51 (45.5)
28 (48.3) 30 (51.7)
1.0 1.2 (0.5–2.6)
1.0 0.5 (0.2–1.3)
Alcohola Never 1–21 >21
63 (32.5) 71 (36.6) 60 (30.9)
25 (24.0) 46 (44.2) 33 (31.7)
1.0 1.6 (0.9–3.0) 1.2 (0.6–2.5)
1.0 1.2 (0.6–2.4) 0.9 (0.4–2.1)
Tobaccoa Never 0–30 >30
50 (25.6) 48 (24.6) 97 (49.7)
21 (20.2) 30 (28.8) 53 (51.0)
1.0 1.5 (0.7–3.0) 1.5 (0.8–2.8)
1.0 0.7 (0.3–1.6) 1.3 (0.6–2.8)
a b
Adjusted for age, tobacco (pack-years) and alcohol (drinks/week). Adjusted for age, tobacco (pack-years), alcohol (drinks/week), and HPV status.
groups (p < 0.01 each). Despite having worse prognostic factors, the median disease-specific survival was better in the p16+/HPV-HR (2.5 years) compared to p16 /HPV (1.7 years, p = 0.0006), p16+/HPV (1.4 years, p = 0.02), and p16 /HPV-HR (1.5 years, p = 0.06) groups. The log-rank test showed a significant difference across the four p16/HPV groups (Fig. 1) for overall (p = 0.04) and time to recurrence (p = 0.005) but not for disease-specific survival (p = 0.17). Table 3 presents the adjusted hazards ratios of overall and disease-specific survival, and disease recurrence for p16, HPV, and the four p16/HPV groups. Model 1 examined all histologic types and model 2 was limited to SCC. Included risks and confounders are based on models from the four joint p16/HPV groups. Multivariate analyses indicated that age, stage, and treatment were associated with survival and should be included in the models whereas gender, tumor site, lymph node status, and grade were not associated with clinical outcomes after inclusion of stage and thus were not included in the final models. Alcohol and tobacco, important risk factors for HNC, also were included in these models. Number of sex partners (P4) was significantly associated with an increased risk of disease-specific death (HR = 2.0, 1.04–3.9), but not other clinical outcomes in the multivariate models. Tumor site was added to the adjusted models to compare oropharyngeal and oral cavity to laryngeal tumors but there were no significant differences in outcomes. Compared to those with p16 overexpression, patients who did not exhibit p16 had significantly worse overall and disease-specific survival, and recurrence (Table 3). HPV cases also had worse overall and disease-specific survival, and higher risk of recurrence compared to those with an
HPV-HR tumor. Disease-specific survival was worse for those who were HPV than it was for those who were p16 but lack of p16 expression was a better predictor of higher risk of recurrence than was HPV tumor status. Next we examined the combined p16/HPV groups for clinical outcomes in comparison to those based on the findings reported for the individual biomarkers. Overall survival was worse among those who were p16 /HPV or p16+/ HPV compared to the p16+/HPV-HR group (Table 3, and Fig. 1a). The p16 /HPV-HR group also showed worse survival although the results were not statistically significant. When analyses were limited to SCC-only (model 2), prognosis was worse in the p16+/HPV cases than for all HNC histologies combined (model 1) whereas the rates were unchanged for the p16 /HPV or p16 /HPV-HR groups by histologic group. Among the other risk factors, each additional year of age but not tobacco or alcohol, added to the risk of overall death. Hazards ratios increased with advanced stages of disease and were significant among those in stage IV in both models. Only the no surgery/radiation treatment group was associated with significantly greater overall mortality compared to surgery/radiation treatment (referent) whereas surgery or radiation had similar survival to the surgery/radiation group after controlling for stage, and other confounders associated with treatment type. The patterns for disease-specific survival were similar to those for overall survival by p16/HPV groups. Again only the p16+/HPV group had worse SCC disease-specific survival compared to all histologic types combined. Compared to the p16+/HPV-HR group, the other three p16/HPV groups showed similar and worse disease-specific survival regardless
138
E.M. Smith et al.
a 1.0
Overall Survival || | || |
| || || |
|| || ||
|
|| |
|| |||| |
0.8
| |||| ||| || ||||
||
||
0.6
||
||
|| ||
|| ||
||| | |||| || || || |||
|| ||
|| |||||| |
|
||
||||
|| || ||
||
||
||
|
| ||
||
||
|| || || ||
|
|| ||
|| |
0.4
|| ||
||
|
p16-negative / HPV-negative p16-negative / HPV-HR p16 + / HPV-negative p16 + / HPV-HR
0.0
0.2
Survival Probability
| ||
|| ||
||
4
2
0
6
8
Time inYears
b 1.0
Disease-Specific Survival |||| | ||
| | || | |
|| ||||
| || |
|| | || || || || || || || |||| || || |||| || |||| || || || | || || |||| | ||||| | |||| || || || || || | |||||| |||||| || || || || || || || ||
|| ||||| || |
0.6
||
||||
||
||
|| ||
||
|
|
|| ||
||
|
0.4
|| ||
p16-negative / HPV-negative p16-negative / HPV-HR p16+ / HPV-negative p16+ / HPV-HR
0.0
0.2
Survival Probability
0.8
||
4
2
0
6
8
Time in Years
c 1.0
Recurrence-Free Survival ||
|| |||
||
|||
|| || ||
||
||
||
|| ||
|| || ||
|
|| || ||
||
|| ||
||
||
||
||
||
||
||
||
|| ||| |||| || ||||
||||
|| |||| |||||| ||||
||
|| || | ||
0.6
|||| || ||
||
||
||
| |
|| ||
||||
|
|
0.4
||
0.2
p16-negative / HPV-negative p16-negative / HPV-HR p16+ / HPV-negative p16+ / HPV-HR
0.0
Recurrence-Free Probability
0.8
|| |
0
2
4
6
8
Time in Years
Figure 1 Survival and disease Recurrence by p16/HPV Status. Overall survival (a), disease-specific survival (b), and time-torecurrence (c) curves based on the Kaplan-Meier method were analyzed. The bold dashed line is for patients who were p16+/HPVHR, the solid bold line is for p16+/HPV , the dashed line is for patients who were p16 /HPV-HR, and the solid line is for p16+/ HPV . Vertical tick marks on curves indicate censored observations.
p16INK4a and survival in HPV–Associated HNC Table 3
Adjusted hazard ratiosa for p16/HPV models of survival and recurrence
Characteristics
Overall survival Model 1
P16 status p16+ p16
139
b
Disease-specific survival Model 2
c
Model 1
b
Model 2
Disease recurrence c
Model 1b
Model 2c
a
1.0 1.8 (1.0–3.2)
1.0 1.6 (0.9–2.8)
1.0 2.0 (1.0–3.9)
1.0 1.5 (0.8–3.2)
1.0 3.6 (1.6–8.2)
1.0 2.9 (1.2–7.3)
HPV statusa HPV-HR HPV
1.0 3.1 (1.5–6.5)
1.0 3.1 (1.5–6.7)
1.0 2.8 (1.1–7.1)
1.0 2.9 (1.1–7.7)
1.0 2.0 (0.8–4.8)
1.0 2.2 (0.8–5.6)
P16/HPV status P16+/HPV-HR p16 /HPVp16 /HPV-HR p16+/HPVAge Tobaccod Alcohole
1.0 3.6 (1.6–8.1) 2.2 (0.4–11.3) 3.3 (1.3–8.7) 1.03 (1.01–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
1.0 3.5 (1.5–8.1) 2.1 (0.4–10.5) 4.3 (1.5–11.8) 1.04 (1.01–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
1.0 3.6 (1.3–10.1) 2.7 (0.5–15.0) 2.8 (0.8–9.5) 1.03 (1.003–1.1) 1.0 (0.98–1.002) 1.0 (0.99–1.01)
1.0 3.6 (1.2–10.3) 2.5 (0.4–14.2) 4.3 (1.2– 15.6) 1.04 (1.01–1.1) 1.0 (0.98–1.001) 1.0 (0.99–1.01)
1.0 4.5 (1.5–13.9) 7.0 (1.6–31.1) 1.8 (0.4–7.7) 1.03 (1.001–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
1.0 5.2 (1.5–17.7) 7.9 (1.6–38.2) 3.7 (0.8–17.8) 1.03 (1.0–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
Stage 0/I/II III IV
1.0 2.0 (0.8–4.9) 3.2 (1.5–6.8)
1.0 2.2 (0.9–5.4) 3.6 (1.6–7.9)
1.0 3.7 (1.2–11.5) 4.4 (1.6–12.0)
1.0 4.4 (1.4–14.1) 5.1 (1.8–14.7)
1.0 1.1 (0.4–3.0) 1.2 (0.5–2.9)
1.0 0.9 (0.3–2.6) 0.9 (0.3–2.4)
4.7 (1.3–17.5)
4.7 (1.2–17.7)
5.4 (1.4–20.8)
5.6 (1.4–22.6)
4.1 (0.5–36.7)
3.3 (0.3–31.0)
1.0 0.7 (0.3–1.4) 1.4 (0.7–2.9)
1.0 0.8 (0.4–1.7) 1.4 (0.7–2.8)
1.0 0.5 (0.2–1.1) 1.4 (0.6–3.4)
1.0 0.6 (0.2–1.4) 1.4 (0.6–3.4)
1.0 0.6 (0.2–1.4) 0.9 (0.4–2.4)
1.0 0.4 (0.1–1.3) 0.8 (0.3–2.1)
1.0
1.0
1.0
1.0
1.0
1.0
1.1 (0.4–3.1) 1.0 (0.4–2.8)
1.2 (0.4–3.4) 1.1 (0.4–3.3)
1.1 (0.3–3.4) 0.8 (0.2–2.7)
1.1 (0.3–3.9) 0.9 (0.2–3.2)
1.1 (0.3–3.8) 1.1 (0.3–4.1)
1.1 (0.3–4.1) 1.3 (0.4–4.9)
Treatment No Surgf/No XRTg Surg & XRT Surg/No XRT XRT/No Surg Tumor site Larynx/ Hypopharynx Oral cavity Oropharynx a
HRs not adjusted for p16 and HPV, age, tobacco, alcohol, stage, treatment, tumor site; all other models are adjusted for p16, HPV, age, tobacco, alcohol, stage, treatment, tumor site. b Model based on all histologic types. c Model based on SCC patients only. d Pack-years. e Drinks/week. f Surg = surgery. g XRT = radiation treatment.
of time since diagnosis (Fig. 1b). Disease-specific HRs were similar to overall survival HRs for age, tobacco, alcohol, treatment type, and tumor site. Patients with stage III or IV disease had significantly worse disease-specific survival than overall survival independent of p16/HPV status and other confounders, and the risk was greater when analyses were limited to SCC (model 2). The p16+/HPV-HR group showed better 5-year disease-specific survival (53%) compared to each other group: 13% for p16 /HPV (p = 0.2), 0% for p16 /HPV-HR (p = 0.004), and 20% for p16+/HPV (p = 0.2). The median time to disease recurrence was longer in the p16+/HPV-HR (2.7 years) compared to each of the other groups: p16 /HPV– (1.5 years, p < 0.0001), p16+/HPV– (2.0 years, p = 0.06), and p16 /HPV-HR (1.1 years,
p = 0.03). Yet the probability of never being disease free was not different between any of the p16/HPV groups, a difference might otherwise bias the interpretation of those at risk of recurrence by p16/HPV group (data not shown). Figure 1c shows that patients who had p16+/HPV-HR tumors also had better recurrence-free survival than did those who did not express p16. The adjusted HRs for recurrence (Table 3) show that p16/HPV status, age, and no surgery/ no radiation (XRT) were significantly associated with recurrence whereas the previously significant factor, stage, was not. Those whose tumors expressed p16 that were HPV had the least elevated risk of recurrence compared to the referent group. The p16 /HPV-HR group showed the greatest likelihood of recurrence, worse than those who were p16 /HPV , which would not have been predicted based
140 on the individual biomarker results, suggesting that alternative pathways that are not well understood contribute to this outcome. The p16+/HPV-HR group again had better 5year recurrence-free survival compared to the other groups: 60% versus 33%, p16+/HPV (p = 0.6), 0%, p16 /HPV-HR (p = 0.08), and 8%, p16 /HPV (p = 0.001). Prognosis based on the individual biomarkers, p16 and HPV-HR, were compared to those based on the joint group effects (Table 3). Whereas overall survival, disease specific survival, and recurrence among those who did not express p16 or who had HPV status were worse than among those who did, differences in prognostic outcomes were more distinct when the two markers were examined together. For example, examination of the p16 status alone did not distinguish the variation in disease-specific survival associated with HPV status (model 1, p16 HR = 2.0 versus p16/HPV, HRs 3.6, 2.7, and 2.8); or the greater risk of recurrence by HPV status among the p16 groups (model 1, HR = 3.6 versus p16/HPV, HRs 4.5 and 7.0). Although it has been established that HPV-HR HNC cases have better survival, this information alone did not reveal the further distinctions in clinical outcomes when p16 expression was assessed with HPV status. When examined by HPV status alone, recurrence was greater in the HPV tumors (HR = 2.0), yet the highest recurrence rate was among those with an HPV-HR, not HPV , tumor that was p16 . Those who were p16 /HPVHR had better disease-specific survival than those who were p16 /HPV yet worse recurrence. The addition of HPV status interestingly showed that the risk of recurrence was actually greater among those who were p16 /HPV-HR than among either the p16+/ HPV or p16 /HPV group.
Discussion This is the first study to examine p16 expression and HPV-HR status in a large group of HNC cases to compare these biomarkers individually and jointly as predictors of clinical outcomes. Unlike most previous studies of HNC prognostic indicators, our investigation controlled for other risk factors and pathologic characteristics of the disease to examine survival associated with the individual and joint biomarkers. The results show different survival and recurrence outcomes in each of the four p16/HPV biomarker groups. Some studies5,29 have suggested that detection of p16 expression by immunohistochemistry could be used as a surrogate marker of an HPV-related tumor. If this were true, the p16INK4a IHC test, which is easier and quicker to perform than current HPV assays, could be used to predict tumors that are causally related to HPV and which have been shown to have better survival than HPV tumors.26,30 However, we found that p16 expression and HPV detection in the tumors diverged in 20% of cases, indicating that one cannot serve as a surrogate marker of the other in HNC. Furthermore, clinical outcomes varied between those patients whose cancers were p16+ and either HPV-HR or HPV . To verify these discordant joint marker results, a random sample of the p16 /HPV-HR and p16+/HPV cases were rerun and confirmed the original findings. Thus our data suggest that a test of p16 status alone is inadequate in predicting clinical outcomes. The significantly more favorable survival and less frequent disease recurrence among the p16+/HPV-HR HNC
E.M. Smith et al. cases compared to p16 /HPV cases was expected based on our and other findings.11,14,26,30 However, 20% of the cases were positive for one but not the other biomarker in this study. Although p16 upregulation in HPV HNC tumors has been reported in other small studies,9,10 we are the first to show that p16+/HPV patients have worse survival than the p16+/HPV-HR group but significantly lower tumor recurrence than either p16 /HPV (+ or ) group. The underlying molecular changes that result in HNC lesions that are p16 /HPV-HR, described here as well as in previous investigations,31,32 remain unclear. Because we took the precaution to laser microdissect tissues containing <20% of tumor cells, it is unlikely that the virus resided outside of the tumor itself. Genetic changes that would alter p16INK4a function, such as homozygous deletions, point or missense mutations have been reported in HNC but these were not examined in association with HPV status.9,10,31,33–35 In two small series, loss of p16INK4a expression in HPV-HR cases has been ascribed to alterations including loss of heterozygosity, point mutations or silencing by methylation.32,35 In the future, it will be important to examine the nature of p16INK4a alterations in HPV-HR HNC lesions to understand how such mutations affect pRB function in HPV+ tumors and whether the prognosis in cases with these mutations is independent of HPV/pRB effects. Changes that would lead to p16 overexpression in p16+/ HPV tumors have not been defined. p16 expression in HNC has varied between 44 and 80%9,13,14,36 with 35% in our study. The high frequency of p16+ tumors in other reports is likely due to the inclusion of more oropharyngeal and tonsillar HNCs which are more commonly HPV-HR-associated. Some of these differences may be due to the evaluation of p16 staining. In our study we included both categories 2 (slightly positive staining, n = 2) and 3 (10–80% positive, n = 9) but the inclusion or exclusion of the small number of cases in these two categories did not alter the statistical findings in our large sample size. Among studies that examined both p16 and HPV in the same patients, one found no association with prognosis16 whereas others found better survival among those who were positive for both markers.11,14 Differences in prognostic outcomes may be attributed to the fact that most investigations did not adjust for clinicopathologic factors potentially associated with prognosis or included different HNC sites. In contrast, we adjusted our results for stage, grade, tumor site and lymph node involvement and found more advanced pathological characteristics among those who were p16+ and HPV-HR. The reason for the significantly better survival shown by us and others in p16+/HPV-HR cases is as yet unclear. A further difference between this and previous studies is the duration of the follow-up period to evaluate outcomes. In a study of 57 SCC cases in the oral cavity and oropharynx, Paradiso et al.16 found no association between p16 status with HPV-HR status or with survival. In contrast, Bova et al.13 in 148 patients limited to cancer of the anterior tongue, found that while lack of p16 expression was not associated with histologic parameters, it was a significant independent predictor of reduced overall and disease-specific survival, similar to our findings. In the series of HNC from multiple tumor sites presented here, the adjusted HRs showed that lack of p16 expression was not significantly associated with worse clinicopathologic param-
p16INK4a and survival in HPV–Associated HNC eters but p16 patients did have worse survival. Geisler et al.33 examined 171 HNC cases and found no association between survival or recurrence and loss of p16 expression using IHC. Geiser and colleagues33 also examined the association between p16 status and radiation treatment for clinical outcomes and found reduced recurrence with this treatment (adj.HR = 0.5, 0.2–1.2), but did not assess HPV status which may have accounted for the improved survival. We also found better disease-specific survival and reduced recurrence among those who received XRT as well as those who had surgery only. Klussmann et al.14 found significantly better disease-specific survival in a group of 34 tonsillar cancer cases who expressed p16 compared to those who did not (70% versus 18%). They examined the combined effects of p16 expression and HPV-positivity and found that 89% of the HPV-HR tumors showed p16 expression. This is the largest study of p16 expression in HNC cases. We included all HNC sites and found that the three major tumor sites did not have significantly different prognostic outcomes after adjusting for HPV status, stage, treatment and 94% of the HPV cases were p16 . Although they found no difference in disease-specific survival by p16/HPV-HR status,14 recurrence occurred significantly more frequently in p16 /HPV cases as compared to those who were positive for both markers. In contrast, we found both better diseasespecific survival and less recurrence in the p16+/HPV-HR group compared to each of the other p16/HPV groups, which may be due to the larger sample or examination of multiple tumor sites, and other confounders. This does not preclude that there are differences in clinical outcomes by tumor subsites (e.g. buccal mucosa, tonsils) associated with p16 expression and HPV-HR status, but a larger sample size is needed to examine these effects. Another advantage of our study was the minimal loss to and high verification of follow-up status using the NCI SEER Cancer Registry and US National Death Index databases. In conclusion, the combined p16/HPV biomarker data demonstrate an association with different survival and recurrence for HNC compared to each marker evaluated separately. This finding suggests that the two markers should be evaluated together in other large studies to verify whether these joint markers better predict prognosis and to help target HNC treatment to specific molecular pathways in the future.
141
3.
4.
5.
6.
7.
8. 9.
10.
11.
12.
13.
14.
15.
Acknowledgements 16.
This study was supported by a grant from NIH NIDCR R01 DE11979 (EMS, YK, DW, LMR, THH, LPT), NIDCR R01 DE13110 (DW, LMR, EMS, LPT, THH), and Veterans Affairs Merit Review Funds (LPT, THH).
References 1. Smith EM, Ritchie JR, Summersgill KF, Hoffman HT, Wang DH, Haugen TH, et al. Human papillomavirus in oral exfoliated cells and risk of head and neck cancer. J Natl Cancer Inst 2004;96:449–55. 2. Mork J, Lie AK, Glattre E, Hallmans G, Jellum E, Koskela P, et al. Human papillomavirus infection as a risk factor for
17.
18.
19.
20.
squamous-cell carcinoma of the head and neck. N Engl J Med 2001;344:1125–31. Kreimer AR, Clifford GM, Boyle P, Franceschi S. Human papillomavirus types in head and neck squamous cell carcinomas worldwide: a systematic review. Cancer Epidemiol Biomark Prev 2005;14:467–75. Li Y, Nichols MA, Shay JW, Xiong Y. Transcriptional repression of the d-type cyclin-dependent kinase inhibitor p16 by the retinoblastoma susceptibility gene product pRb. Cancer Res 1994;54:6078–82. Wang SS, Trunk M, Schiffman M, Herrero R, Sherman ME, Burk RD, et al. Validation of p16INK4a as a marker of oncogenic human papillomavirus infection in cervical biopsies from a population-based cohort in Costa Rica. Cancer Epidemiol Biomark Prev 2004;13:1355–60. Klaes R, Benner A, Friedrich T, Ridder R, Herrington S, Jenkins D, et al. P16 immunohistochemistry improves interobserver agreement in the diagnosis of cervical intraepithelial neoplasia. Am J Surg Pathol 2002;26:1389–99. Sano T, Masuda N, Oyama T, Nakajima T. Overexpression of p16 and p14arf is associated with human papillomavirus infection in cervical squamous cell carcinoma and dysplasia. Pathol Int 2002;52:375–83. Lowe SW, Sherr CJ. Tumor suppression by Ink4a-Arf: progress and puzzles. Curr Opin Gene Dev 2003;13:77–83. Lang JC, Tobin EJ, Knobloch TJ, Schuller DE, Bartynski KJ, Mountain RE, et al. Frequent mutation of p16 in squamous cell carcinoma of the head and neck. Laryngoscope 1998;108: 923–8. El-Naggar AK, Lai S, Clayman G, Lee J-K, Luna MA, Goedfert H, et al. Methylation, a major mechanism of p16/CDKN2 gene inactivation in head and neck squamous carcinoma. Am J Pathol 1997;151:1767–74. Weinberger PM, Yu Z, Haffty BG, Kowalski D, Harigopal M, Brandsma J, et al. Molecular classification identifies a subset of human papillomavirus-associated oropharyngeal cancers with favorable prognosis. J Clin Oncol 2006;24:736–47. Andl T, Kahn T, Pfuhl A, Nicola T, Erber R, Conradt C, et al. Etiological involvement of oncogenic human papillomavirus in tonsillar squamous cell carcinomas lacking retinoblastoma cell cycle control. Cancer Res 1998;58:5–13. Bova RJ, Quinn DI, Nankervis JS, Cole IE, Sheridan BF, Jensen MJ, et al. Cyclin d1 and p16 expression predict reduced survival in carcinoma of the anterior tongue. Clin Cancer Res 1999;5:2810–9. Klussmann JP, Gultekin E, Weissenborn SJ, Wieland U, Dries V, Dienes HP, et al. Expression of p16 protein identifies a distinct entity of tonsillar carcinomas associated with human papillomavirus. Am J Pathol 2003;162:747–53. Li W, Thompson CH, O’Brien CJ, McNeil EB, Scolyer RA, Cossart YE, et al. Human papillomavirus positivity predicts favourable outcome for squamous carcinoma of the tonsil. Int J Cancer 2003;106:553–8. Paradiso A, Ranieri G, Stea B, Zito A, Zehbe I, Tommasino M, et al. Altered p16INK4a and Fhit expression in carcinogenesis and progression of human oral cancer. Int J Oncol 2004;24:249–55. Emmert-Buck MR, Bonner RF, Smith PD, Chuaqui RF, Zhuang Z, Goldstein SR, et al. Laser capture microdissection. Science 1996;274:998–1001. Fleming ID, Cooper JS, Henson DE, Hutter RVP, Kennedy BJ, Murphy GP, et al., editors. AJCC cancer staging manual. 5th ed. Philadelphia (PA): Lippincott-Raven; 1997. Summersgill KF, Smith EM, Levy BT, Allen JM, Haugen TH, Turek LP. Human papillomavirus in the oral cavities of children and adolescents. Oral Surg Oral Med Oral Pathol Oral Radiol Endo 2001;91:62–9. NCI SEER. http://seer.cancer.gov/csr/1973–1999; 2002, 2004.
142 21. National Death Index. http://www.cdc.gov/nchs/r&d/ndi/ ndi.htm. 22. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Amer Stat Assoc 1958;53:457–81. 23. Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966;50:163–70. 24. S-PLUS 2000 Academic Site, editor. Version 6.2. Seattle WA: Math Soft, Inc.; 2003. 25. Cox DR. Regression models and life tables. JR Stat Soc (B) 1972;34:187–220. 26. Ritchie JM, Smith EM, Summersgill KF, Hoffman HT, Wang D, Klussmann JP, et al. Human papillomavirus infection as a prognostic factor in carcinomas of the oral cavity and oropharynx. Int J Cancer 2003;104:336–44. 27. Allison PD. Survival analysis using the SAS System, a practical guide. Cary, NC: SAS Institute, Inc.; 1995. 28. SAS System for Windows, version 9.1. Cary, NC: SAS Institute; 2002–2003. 29. Dray M, Russell P, Dalrymple C, Wallman N, Anous G, Lelong A, et al. p16INK4a as a complementary marker of high-grade intraepithelial lesions of the uterine cervix. I: Experience with squamous lesions in 189 consecutive cervical biopsies. Pathol 2005;37:112–24. 30. Gillison ML, Koch WM, Capone RB, Spafford M, Westra WH, Wu L, et al. Evidence for a causal association between human
E.M. Smith et al.
31.
32.
33.
34.
35.
36.
papillomavirus and a subset of head and neck cancers. J Natl Cancer Inst 2000;92:709–20. Li W, Thompson CH, Cossart YE, O’Brien CJ, McNeil EB, Scolyer RA, et al. The expression of key cell cycle markers and presence of human papillomavirus in squamous cell carcinoma of the tonsil. Head Neck 2004;26:1–9. Fregonesis PA, Teresa DB, Duarte RA, Neto CB, Brancini de Oliveira MR, Soares CP. p16INK4a immunohistochemical overexpression in premalignant and malignant oral lesions infected with human papillomavirus. J Histochem Cytochem 2003;51:1291–7. Geisler SA, Olshan AF, Weissler MC, Cai J, Funkhouser WK, Smith J, et al. P16 and p53 protein expression as prognostic indicators of survival and disease recurrence from head and neck cancer. Clin Cancer Res 2002;8:3445–53. Lang JC, Borchers J, Danahey D, Smith S, Stover DG, Agrawal A, et al. Mutational status of overexpressed p16 in head and neck cancer: evidence for germline mutation of p16/p14arf. Int J Oncol 2002;21:401–8. Reed AL, Califano J, Cairns P, Westra WH, Jones RM, Koch W, et al. High frequency of p16 (cdkn2/mts-1/) inactivation in head and neck squamous cell carcinoma. Cancer Res 1996;56:3630–3. Ambrosch P, Schlott T, Hilmes D, Ruschenburg I. P16 alterations and retinoblastoma protein expression in squamous cell carcinoma and neighboring dysplasia from the upper aerodigestive tract. Virchows Arch 2001;438:343–9.
available at www.sciencedirect.com
journal homepage: http://intl.elsevierhealth.com/journals/oron/
p16INK4a Expression, human papillomavirus, and survival in head and neck cancer Elaine M. Smith a,*, Donghong Wang a, Yoonsang Kim b, Linda M. Rubenstein a, John H. Lee c, Thomas H. Haugen d, Lubomir P. Turek d a
Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA 52242, United States Department of Biostatistics, College of Public Health, University of Iowa, Iowa City, IA 52242, United States c Department of Otolaryngology, College of Medicine, University of Iowa, Iowa City, IA 52242, United States d Veterans Affairs Medical Center and Department of Pathology, College of Medicine, University of Iowa, Iowa City, IA 52242, United States b
Received 13 November 2006; accepted 15 January 2007 Available online 13 March 2007
KEYWORDS
Summary Development of head and neck cancer (HNC) is associated with human papillomavirus high-risk (HPV-HR) types. The HPV E7 oncoprotein inactivates the pRB protein increasing expression of p16INK4a. p16 Expression and HPV status have been associated with differences in clinical outcomes for HNC. This study examined whether HNC prognosis was different when these biomarkers were examined as individual or joint molecular effects. Tumor tissue from 301 HNC patients were analyzed and sequenced for HPV types. p16 was evaluated by immunohistochemical staining. p16 was expressed in 35% and HPV-HR was detected in 27% of HNC patients. After adjustment for age, tobacco, and alcohol, p16+ tumors were statistically significantly associated with HPV-HR (OR = 13.3, 7.1–24.9), histology, stage, grade, tumor site, and node involvement. Compared to p16+ HNC cases, those who did not express p16 had significantly worse disease-specific (DS) survival (Hazards Ratio, adj.HR = 2.0. 1.0–3.9) and recurrence (adj.HR = 3.6, 1.6–8.2); and HPV cases had worse DS survival (adj.HR = 2.8, 1.1–7.1) and recurrence (adj.HR = 2.0, 0.8–4.8) compared to HPV-HR patients. Each of the p16/HPV groups had different survival outcomes: p16+/HPV-HR cases (referent) had the best and p16 /HPV cases had the worst DS survival (adj.HR = 3.6; 53% versus 13%, p = 0.004) whereas p16+/HPV and p16 /HPV-HR had similar DS survival (adj.HR = 2.7/2.8). p16 /HPV-HR cases
Head and neck cancer survival; HPV; p16INK4a.
* Corresponding author. Tel.: +1 319 384 5014; fax: +1 319 384 5031. E-mail address: [email protected] (E.M. Smith).
1368-8375/$ - see front matter c 2007 Published by Elsevier Ltd. doi:10.1016/j.oraloncology.2007.01.010
134
E.M. Smith et al. had a worse recurrence rate (adj.HR = 7.0; 60% versus 0%, referent, p = 0.08) than p16 /HPV (adj.HR = 4.5) or p16+/HPV (adj.HR = 1.8) cases. The combined p16/HPV biomarker data are associated with different survival outcomes of HNC compared to each marker evaluated separately, indicating that the two molecular mechanisms evaluated together may provide a more accurate prediction of clinical outcomes. c 2007 Published by Elsevier Ltd.
Introduction Infection with HPV high-risk (HR) types is associated with the development of 25% of head and neck cancers (HNC) independent of tobacco and alcohol consumption.1–3 The HPV E7 oncoprotein inactivates the pRB tumor suppressor protein. The abrogation of pRB function is thought to be responsible for a prominent rise in the levels of p16INK4a, a cyclin-dependent kinase inhibitor, by the loss of negative feedback control.4 Expression of p16INK4a in association with HPV-HR infection has been demonstrated in a high percentage of high-grade cervical dysplasias and cancer.5,6 It has been suggested that p16INK4a expression may serve as either a surrogate biomarker of HPV oncogenic infection or play an independent role in predicting cervical cancer.6,7 However, the signals and pathways that determine p16INK4a expression in humans in general8 or p16INK4a upregulation in HPV-unrelated tumors are not well understood. The role of p16INK4a as a marker of HPV-associated head and neck tumors is less clear than it is for cervical cancer. Several small9–11 but no large-scale studies have examined the association between p16 expression as a marker for HPV-HR in HNC and because of their limited sample size they were unable to evaluate other covariates that might influence that association. p16INK4a might also serve as a marker to distinguish integrated and nonintegrated oncogenic HPV or nononcogenic HPV detected in HNC tumors. However, current findings regarding the prevalence of p16 expression observed in HNC vary widely, between 44% and 80%.12–14 Some12,13,15 but not other16 investigations also have found that p16 expression is a marker of better survival for HNC patients. In this study we evaluated a large number of HNC cases for differences in p16 expression and HPV-HR status in tumor cells and their interrelationship with HNC cancer controlling for other HNC risk factors and tumor characteristics. These two biomarkers also were evaluated to determine whether prognostic outcomes associated with HNC were different and more highly predictive when examined individually or in combination.
the time of the study analyses. After patients signed an approved Human Subjects form, they completed a selfadministered risk factor questionnaire regarding sociodemographics, tobacco habits, alcohol consumption, sexual practices, medical history including HPV-related diseases and oral lesions. Information regarding prior HNC history, tumor site, histology, stage, grade, node involvement, and treatment were identified from pathological reports and medical records.
Tumor specimens Hematoxylin and eosin-stained slides (H&E) from formalinfixed, paraffin-embedded tumor blocks were evaluated to verify the presence of malignant cells and tumor grade. Laser capture microdissection (LCM) was performed to verify that HPV DNA detected from the whole tissues was from the tumor cells. Procedures have been described previously1,17 and were performed when tissue specimens had less than 10% of tumor tissue or when they were initially found to contain an HPV nononcogenic, low-risk type. All histologic types of HNC were included in the study and were classified as squamous cell carcinoma (SCC) or nonSCC. Most tumors were SCC (92%) with a small percent of other histologic types: seven mucoepidermoid carcinomas, four adenoid cystic carcinomas, three acinar cell carcinomas, three adenocarcinoma, two verrucous carcinoma, one basal cell adenocarcinoma, one carcinoma, NOS, one polymorphous adenocarcinoma, one carcinoma in pleomorphic adenoma, and 1 mixed tumor, NOS. For each case, 1–2 blocks were selected with the highest percentage of tumor for HPV analysis and p16 immunohistochemical staining (IHC). After a block was selected, serial 4 lm thick sections were cut. Tumor staging was based on the 1997 American Joint Committee on Cancer (AJCC) criteria18 and tumor sites were grouped into oral cavity, oropharynx, and larynx/ hypopharynx (excluding nasopharynx) according to the AJCC Cancer Manual.18
p16 Immunohistochemical staining (IHC)
Methods Data collection The study included 301 histologically confirmed HNC cases recruited between 1994 and 2004 from the Department of Otolaryngology, University of Iowa Hospitals and Clinics, for whom p16 immunohistostaining had been performed at
Sections for p16 staining were deparaffined by passage through xylene and then dehydrated in graded alcohol. Two slides from the same cut series of the same block were used. One was stained by p16 mouse anti human monoclonal antibody E6H4 from CINtecä p16INK4a histology kits (Dako Corporation, CA); another slide was stained by monoclonal mouse IgG to Aspergillus niger gluxose oxidase from the same kit to verify specificity of staining observed with
p16INK4a and survival in HPV–Associated HNC monoclonal antibody. One known positive and negative p16 slides were used as controls in each assay. Before antibody staining, epitope retrieval was done by 0.1 mol/L citrate buffer in a steamer for 10 min. After cooling and rinsing through buffer, tissue sections were quenched of endogenous peroxidase by 3% hydrogen peroxide containing 15 mmol/L sodium azide. After rinsing, 1:25 dilute monoclonal antibody of p16 was applied to one slide and negative control reagent to another slide from the same block and then incubated 30 min at room temperature. The slide was then covered with 200 ll of visualization reagent consisting of secondary goat-anti-mouse immunoglobulin molecules and horseradish peroxidase molecules linked to a common dextran polymer backbone, eliminating the need for sequential application of link antibody and peroxidaseconjugated antibody. Localization of antigen was visualized using the chromogen DAB for 10 min. Finally, tissue sections were counter-stained in dilute hematoxylin, dehydrated in graded alcohol, cleared in xylene, and mounted. All p16 IHC slides were read by one observer (THH) and 10% were re-reviewed. Strong nuclear and cytoplasmic staining defined a positive reaction. P16 positive staining was limited to the tumor cells, occassionally including a few multinucleate giant cells which stained strongly positive, and less than 5% epithelium had focal positive stain. The p16 positive distribution was scored on a semiquantitative scale at first and then converted to positive and negative scores before statistical analyses. The scale was defined as: 1 = negative for staining; 2 = slightly positive for staining, 3 = 10–80% positive; 4 = > 80% positive; 5 = every tumor cell positive. Slides in categories 1 was defined as negative, and 2–5 were positive. Categories 2 (n = 2) and 3 (n = 9) were included as positive for IHC. Their inclusion did not alter the statistical findings in this large sample size (data not shown). Positive reaction of nontumor cells, such as multinucleated giant cells and epithelia, were not included as positive.
HPV detection and typing HPV detection and typing from paraffin-embedded tissue was performed as described elsewhere.19 Briefly, DNA isolation was performed by protocol from Qiagen QIAamp DNA mini kits. HPV detection was performed by PCR with MY09/11 primers and dot blot hybridization. Dot blot positive samples were further amplified by a second round of PCR with primers of GP5+/GP6+ or GP5+/MY09 and positive samples were typed by automated sequencing.
Statistical methods Risk factors and demographic characteristics included age, gender, education, tobacco, alcohol, number of sexual partners, oral–genital sex history, tumor site (oral cavity, oropharynx, larynx/hypopharynx), stage, grade, nodal involvement, and histology (SCC versus all others). Age, tobacco, and alcohol use were examined as both categorical and continuous variables in all analyses. Tobacco exposure included cigarettes, cigars, pipes, chewing tobacco, and snuff. Total cumulative exposure for an individual was expressed as pack-years (1 pack-year = 1 pack of cigarettes/
135 day for 1 year with 20 cigarettes/pack). Alcohol assessment included exposure to beer, wine, and liquor (one drink equivalent to 12-oz can/bottle of beer = 4-oz glass of wine = 1.5-oz shot of liquor) and dose-duration was calculated as the average number of drinks/week by number of years use. Both alcohol and tobacco use were also assessed as current, former, and never users. Former use was defined as not having used the substance for at least 1 year prior to diagnosis and never users defined as never having used the substance for 1 year or more during their life. The unadjusted relationships between continuous and categorical variables were assessed with the Wilcoxon rank sum test. Unadjusted associations between categorical variables were tested with either the Pearson chi-square test or Fisher’s exact test for small sample sizes. We also tested for possible interactions among risk factors. Tests for interactions were generated using both logistic regression and proportional hazards regression methods. Multivariate unconditional logistic regression was used to examine the association between dichotomous dependent variables and risk factors for HNC (e.g. age, gender, education, tobacco and alcohol exposure, and sexual behavior and history). Odds ratios (OR) and 95% confidence intervals (CI) were calculated to measure the magnitude of the associations. Survival and recurrence were measured in years from the date of diagnosis until death or until the patient was last known to be alive. In disease-specific survival analyses, patients who died of causes other than HNC were classified as censored observations. In the disease free patients, recurrence in patients was measured in years from the date of diagnosis until recurrence, while time for censored patients was measured in years from the date of diagnosis until the patient died or was last known to be alive. Dates of death or dates last known to be alive were available from the Iowa NCI SEER Cancer Registry,20 the university hospital tumor registry follow-up, and the National Death Index.21 Three sets of analyses were performed: overall survival, disease-specific survival, and time to recurrence. P16 and HPV were modeled independently and then together across all three sets of analyses of clinical outcomes. In disease-specific analyses, patients who died of causes other than HNC were classified as censored observations. Survival curves were estimated by the Kaplan–Meier method22 and the difference between curves was tested by the log-rank test.23 All survival curves were generated in S-PLUS 6.2.24 Cox proportional hazards models25 were used to assess the relationships between p16, HPV and time to death or recurrence, and to adjust for factors previously found to be prognostically significant in HNC.26 These analyses included age (continuous), tobacco, alcohol, stage (0/I/II: referent), and treatment (no treatment: referent). Tumor site (oral cavity, oropharynx, and larynx/ hypopharynx as the referent) also were examined in separate models. All variables were assessed for the proportional hazards assumption27 before inclusion in the models. P16 and HPV status were examined separately and then in four groups: p16+/HPV-HR (referent), p16+/HPV , p16 /HPVHR, p16 /HPV . The magnitude of the associations was assessed with hazard ratios (HRs) calculated from the Cox regression models. Statistical significance was based on two-tailed tests and p-values 60.05. Statistical analyses were performed using SAS 9.1.28
136
E.M. Smith et al.
Results P16 and HPV biomarkers and risk factors Among the 301 HNC cases, 57% were in the oral cavity, 30% in the oropharynx, and 13% in the larynx/hypopharynx. P16 was expressed in 35% and the risk of p16+ was higher in the oropharynx (65%) or oral cavity (25%) compared to the larynx (12%; Table 1). HPV-HR was detected in 27% of tumors: 95% HPV-16, 4% HPV-33, and 1% HPV-18. The most frequent site detected with HPV was the oropharynx (57%) followed by the oral cavity (15%), and larynx (10%). When HPV status was examined in association with p16, positive concordance (p16+/HPV-HR) was 21% and negative concordance (p16 /HPV ) was 59%, whereas 14% with p16+ expression were HPV and 6% of cases with HPV-HR did not express p16. Oropharyngeal cancer, clinicopathologic characteristics, and HPV-HR were significantly associated with p16 overexpression after adjustment for age, tobacco, and alcohol (Table 1). When HPV-HR and other pathologic characteristics were added to the adjusted model in Table 1, there was a drop in the odds ratio associated with p16+ status in tumors of the oropharynx but not the oral cavity compared to the larynx. The risk remained statistically significantly elevated in the p16+ patients for stage IV but not stage III, and other histologic types of HNCs showed a greater increased risk compared to SCC HNC histology. HPV-HR remained significantly associated with p16+ after controlling for risk factors and pathologic characteristics.
Table 1
Table 2 shows the risk of p16 status associated with potential HNC risk factors and confounders. After inclusion of HPV status in the adjusted ORs, male gender and younger age, although elevated, were no longer statistically significantly associated with p16 expression whereas greater number of lifetime sex partners (P4) remained significant. P16 expression was not associated with alcohol or tobacco use.
Survival Data for both p16 and HPV status were available in the survival models excluding three cases due to unknown cause of death and 35 who were never disease free. There were no differences in the odds ratios for histologic and risk factor characteristics when data were limited to patients with follow-up data on clinical outcomes compared to those without this information (data not shown). The median time to death or last follow-up was 1.8 years (range: 0.03–9 years). Adjusting for age, tobacco, alcohol, stage, treatment, histologic type, and tumor site, cases with p16+/HPV-HR (referent group) were significantly more likely to be detected in late stage disease (III or IV; p 6 0.001) and with tumors in the oropharynx (p < 0.0001) compared to each of the other three p16/HPV groups (p16 /HPV , p16+/HPV , p16 /HPV-HR; data not shown). The p16+/HPV-HR group also was more likely to be a former or current tobacco or alcohol user than the p16+/HPV (tobacco: p = 0.001, alcohol: p = 0.007), or p16 /HPV (p = 0.008, p = 0.07) groups, and to be younger (655 years) than each of the three other
Clinicopathologic characteristics of HNC by p16 status
Characteristics
p16 , N = 196 (%)
p16+, N = 105 (%)
Adjusted ORa (95% CI)
Adjusted ORb (95% CI)
Site Larynx/Hypopharynx Oral Cavity Oropharynx
36 (18.4) 128 (65.3) 32 (16.3)
5 (4.8) 42 (40.0) 58 (55.2)
1.0 2.5 (0.9–7.1) 12.5 (4.3–36.4)
1.0 2.4 (0.8–7.5) 5.7 (1.7–18.3)
77 (39.5) 31 (15.9) 87 (44.6)
18 (17.1) 18 (17.1) 69 (65.7)
1.0 2.5 (1.1–5.4) 3.3 (1.8–6.0)
1.0 1.5 (0.6–3.9) 2.4 (1.2–4.9)
Grade Well/moderately Poorly/Undifferentiated
154 (81.5) 35 (18.5)
65 (64.4) 36 (35.6)
1.0 2.4 (1.4–4.2)
1.0 1.9 (0.9–3.8)
Nodal involvement No Yes
122 (62.6) 73 (37.4)
37 (35.2) 68 (64.8)
1.0 2.9 (1.8–4.9)
1.0 1.1 (0.5–2.4)
Histology SCC Other
185 (94.4) 11 (5.6)
92 (87.6) 13 (12.4)
1.0 2.5 (1.03–6.0)
1.0 5.6 (2.1–15.1)
HPV status HPV-HR HPV
19 (9.8) 175 (90.2)
62 (59.0) 43 (41.0)
13.3 (7.1–24.9) 1.0
9.3 (4.7–18.5) 1.0
Stage 0/I/II III IV
a b
Adjusted for age, tobacco (pack-years), and alcohol (drinks/week). Adjusted for age, tobacco (pack-years), alcohol (drinks/week), tumor site, stage, grade, nodes, histology, and HPV status.
p16INK4a and survival in HPV–Associated HNC Table 2
137
Demographic and risk factor characteristics of HNC by p16 status
Characteristics
p16 , N = 196 (%)
p16+, N = 105 (%)
Adjusted ORa (95% CI)
Adjusted ORb (95% CI)
Gender Male Female
113 (57.7) 83 (42.3)
75 (71.4) 30 (28.6)
1.9 (1.1–3.2) 1.0
1.5 (0.8–2.8) 1.0
Age group 655 >55
68 (34.7) 129 (65.8)
50 (47.6) 55 (52.4)
1.8 (1.1–2.9) 1.0
1.5 (0.9–2.7) 1.0
Number of partners 63 P4
105 (57.4) 78 (42.6)
33 (33.7) 65 (66.3)
1.0 3.0 (1.7–5.4)
1.0 2.7 (1.4–5.4)
Oral-genital contact No Yes
61 (54.5) 51 (45.5)
28 (48.3) 30 (51.7)
1.0 1.2 (0.5–2.6)
1.0 0.5 (0.2–1.3)
Alcohola Never 1–21 >21
63 (32.5) 71 (36.6) 60 (30.9)
25 (24.0) 46 (44.2) 33 (31.7)
1.0 1.6 (0.9–3.0) 1.2 (0.6–2.5)
1.0 1.2 (0.6–2.4) 0.9 (0.4–2.1)
Tobaccoa Never 0–30 >30
50 (25.6) 48 (24.6) 97 (49.7)
21 (20.2) 30 (28.8) 53 (51.0)
1.0 1.5 (0.7–3.0) 1.5 (0.8–2.8)
1.0 0.7 (0.3–1.6) 1.3 (0.6–2.8)
a b
Adjusted for age, tobacco (pack-years) and alcohol (drinks/week). Adjusted for age, tobacco (pack-years), alcohol (drinks/week), and HPV status.
groups (p < 0.01 each). Despite having worse prognostic factors, the median disease-specific survival was better in the p16+/HPV-HR (2.5 years) compared to p16 /HPV (1.7 years, p = 0.0006), p16+/HPV (1.4 years, p = 0.02), and p16 /HPV-HR (1.5 years, p = 0.06) groups. The log-rank test showed a significant difference across the four p16/HPV groups (Fig. 1) for overall (p = 0.04) and time to recurrence (p = 0.005) but not for disease-specific survival (p = 0.17). Table 3 presents the adjusted hazards ratios of overall and disease-specific survival, and disease recurrence for p16, HPV, and the four p16/HPV groups. Model 1 examined all histologic types and model 2 was limited to SCC. Included risks and confounders are based on models from the four joint p16/HPV groups. Multivariate analyses indicated that age, stage, and treatment were associated with survival and should be included in the models whereas gender, tumor site, lymph node status, and grade were not associated with clinical outcomes after inclusion of stage and thus were not included in the final models. Alcohol and tobacco, important risk factors for HNC, also were included in these models. Number of sex partners (P4) was significantly associated with an increased risk of disease-specific death (HR = 2.0, 1.04–3.9), but not other clinical outcomes in the multivariate models. Tumor site was added to the adjusted models to compare oropharyngeal and oral cavity to laryngeal tumors but there were no significant differences in outcomes. Compared to those with p16 overexpression, patients who did not exhibit p16 had significantly worse overall and disease-specific survival, and recurrence (Table 3). HPV cases also had worse overall and disease-specific survival, and higher risk of recurrence compared to those with an
HPV-HR tumor. Disease-specific survival was worse for those who were HPV than it was for those who were p16 but lack of p16 expression was a better predictor of higher risk of recurrence than was HPV tumor status. Next we examined the combined p16/HPV groups for clinical outcomes in comparison to those based on the findings reported for the individual biomarkers. Overall survival was worse among those who were p16 /HPV or p16+/ HPV compared to the p16+/HPV-HR group (Table 3, and Fig. 1a). The p16 /HPV-HR group also showed worse survival although the results were not statistically significant. When analyses were limited to SCC-only (model 2), prognosis was worse in the p16+/HPV cases than for all HNC histologies combined (model 1) whereas the rates were unchanged for the p16 /HPV or p16 /HPV-HR groups by histologic group. Among the other risk factors, each additional year of age but not tobacco or alcohol, added to the risk of overall death. Hazards ratios increased with advanced stages of disease and were significant among those in stage IV in both models. Only the no surgery/radiation treatment group was associated with significantly greater overall mortality compared to surgery/radiation treatment (referent) whereas surgery or radiation had similar survival to the surgery/radiation group after controlling for stage, and other confounders associated with treatment type. The patterns for disease-specific survival were similar to those for overall survival by p16/HPV groups. Again only the p16+/HPV group had worse SCC disease-specific survival compared to all histologic types combined. Compared to the p16+/HPV-HR group, the other three p16/HPV groups showed similar and worse disease-specific survival regardless
138
E.M. Smith et al.
a 1.0
Overall Survival || | || |
| || || |
|| || ||
|
|| |
|| |||| |
0.8
| |||| ||| || ||||
||
||
0.6
||
||
|| ||
|| ||
||| | |||| || || || |||
|| ||
|| |||||| |
|
||
||||
|| || ||
||
||
||
|
| ||
||
||
|| || || ||
|
|| ||
|| |
0.4
|| ||
||
|
p16-negative / HPV-negative p16-negative / HPV-HR p16 + / HPV-negative p16 + / HPV-HR
0.0
0.2
Survival Probability
| ||
|| ||
||
4
2
0
6
8
Time inYears
b 1.0
Disease-Specific Survival |||| | ||
| | || | |
|| ||||
| || |
|| | || || || || || || || |||| || || |||| || |||| || || || | || || |||| | ||||| | |||| || || || || || | |||||| |||||| || || || || || || || ||
|| ||||| || |
0.6
||
||||
||
||
|| ||
||
|
|
|| ||
||
|
0.4
|| ||
p16-negative / HPV-negative p16-negative / HPV-HR p16+ / HPV-negative p16+ / HPV-HR
0.0
0.2
Survival Probability
0.8
||
4
2
0
6
8
Time in Years
c 1.0
Recurrence-Free Survival ||
|| |||
||
|||
|| || ||
||
||
||
|| ||
|| || ||
|
|| || ||
||
|| ||
||
||
||
||
||
||
||
||
|| ||| |||| || ||||
||||
|| |||| |||||| ||||
||
|| || | ||
0.6
|||| || ||
||
||
||
| |
|| ||
||||
|
|
0.4
||
0.2
p16-negative / HPV-negative p16-negative / HPV-HR p16+ / HPV-negative p16+ / HPV-HR
0.0
Recurrence-Free Probability
0.8
|| |
0
2
4
6
8
Time in Years
Figure 1 Survival and disease Recurrence by p16/HPV Status. Overall survival (a), disease-specific survival (b), and time-torecurrence (c) curves based on the Kaplan-Meier method were analyzed. The bold dashed line is for patients who were p16+/HPVHR, the solid bold line is for p16+/HPV , the dashed line is for patients who were p16 /HPV-HR, and the solid line is for p16+/ HPV . Vertical tick marks on curves indicate censored observations.
p16INK4a and survival in HPV–Associated HNC Table 3
Adjusted hazard ratiosa for p16/HPV models of survival and recurrence
Characteristics
Overall survival Model 1
P16 status p16+ p16
139
b
Disease-specific survival Model 2
c
Model 1
b
Model 2
Disease recurrence c
Model 1b
Model 2c
a
1.0 1.8 (1.0–3.2)
1.0 1.6 (0.9–2.8)
1.0 2.0 (1.0–3.9)
1.0 1.5 (0.8–3.2)
1.0 3.6 (1.6–8.2)
1.0 2.9 (1.2–7.3)
HPV statusa HPV-HR HPV
1.0 3.1 (1.5–6.5)
1.0 3.1 (1.5–6.7)
1.0 2.8 (1.1–7.1)
1.0 2.9 (1.1–7.7)
1.0 2.0 (0.8–4.8)
1.0 2.2 (0.8–5.6)
P16/HPV status P16+/HPV-HR p16 /HPVp16 /HPV-HR p16+/HPVAge Tobaccod Alcohole
1.0 3.6 (1.6–8.1) 2.2 (0.4–11.3) 3.3 (1.3–8.7) 1.03 (1.01–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
1.0 3.5 (1.5–8.1) 2.1 (0.4–10.5) 4.3 (1.5–11.8) 1.04 (1.01–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
1.0 3.6 (1.3–10.1) 2.7 (0.5–15.0) 2.8 (0.8–9.5) 1.03 (1.003–1.1) 1.0 (0.98–1.002) 1.0 (0.99–1.01)
1.0 3.6 (1.2–10.3) 2.5 (0.4–14.2) 4.3 (1.2– 15.6) 1.04 (1.01–1.1) 1.0 (0.98–1.001) 1.0 (0.99–1.01)
1.0 4.5 (1.5–13.9) 7.0 (1.6–31.1) 1.8 (0.4–7.7) 1.03 (1.001–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
1.0 5.2 (1.5–17.7) 7.9 (1.6–38.2) 3.7 (0.8–17.8) 1.03 (1.0–1.1) 1.0 (0.99–1.01) 1.0 (0.99–1.01)
Stage 0/I/II III IV
1.0 2.0 (0.8–4.9) 3.2 (1.5–6.8)
1.0 2.2 (0.9–5.4) 3.6 (1.6–7.9)
1.0 3.7 (1.2–11.5) 4.4 (1.6–12.0)
1.0 4.4 (1.4–14.1) 5.1 (1.8–14.7)
1.0 1.1 (0.4–3.0) 1.2 (0.5–2.9)
1.0 0.9 (0.3–2.6) 0.9 (0.3–2.4)
4.7 (1.3–17.5)
4.7 (1.2–17.7)
5.4 (1.4–20.8)
5.6 (1.4–22.6)
4.1 (0.5–36.7)
3.3 (0.3–31.0)
1.0 0.7 (0.3–1.4) 1.4 (0.7–2.9)
1.0 0.8 (0.4–1.7) 1.4 (0.7–2.8)
1.0 0.5 (0.2–1.1) 1.4 (0.6–3.4)
1.0 0.6 (0.2–1.4) 1.4 (0.6–3.4)
1.0 0.6 (0.2–1.4) 0.9 (0.4–2.4)
1.0 0.4 (0.1–1.3) 0.8 (0.3–2.1)
1.0
1.0
1.0
1.0
1.0
1.0
1.1 (0.4–3.1) 1.0 (0.4–2.8)
1.2 (0.4–3.4) 1.1 (0.4–3.3)
1.1 (0.3–3.4) 0.8 (0.2–2.7)
1.1 (0.3–3.9) 0.9 (0.2–3.2)
1.1 (0.3–3.8) 1.1 (0.3–4.1)
1.1 (0.3–4.1) 1.3 (0.4–4.9)
Treatment No Surgf/No XRTg Surg & XRT Surg/No XRT XRT/No Surg Tumor site Larynx/ Hypopharynx Oral cavity Oropharynx a
HRs not adjusted for p16 and HPV, age, tobacco, alcohol, stage, treatment, tumor site; all other models are adjusted for p16, HPV, age, tobacco, alcohol, stage, treatment, tumor site. b Model based on all histologic types. c Model based on SCC patients only. d Pack-years. e Drinks/week. f Surg = surgery. g XRT = radiation treatment.
of time since diagnosis (Fig. 1b). Disease-specific HRs were similar to overall survival HRs for age, tobacco, alcohol, treatment type, and tumor site. Patients with stage III or IV disease had significantly worse disease-specific survival than overall survival independent of p16/HPV status and other confounders, and the risk was greater when analyses were limited to SCC (model 2). The p16+/HPV-HR group showed better 5-year disease-specific survival (53%) compared to each other group: 13% for p16 /HPV (p = 0.2), 0% for p16 /HPV-HR (p = 0.004), and 20% for p16+/HPV (p = 0.2). The median time to disease recurrence was longer in the p16+/HPV-HR (2.7 years) compared to each of the other groups: p16 /HPV– (1.5 years, p < 0.0001), p16+/HPV– (2.0 years, p = 0.06), and p16 /HPV-HR (1.1 years,
p = 0.03). Yet the probability of never being disease free was not different between any of the p16/HPV groups, a difference might otherwise bias the interpretation of those at risk of recurrence by p16/HPV group (data not shown). Figure 1c shows that patients who had p16+/HPV-HR tumors also had better recurrence-free survival than did those who did not express p16. The adjusted HRs for recurrence (Table 3) show that p16/HPV status, age, and no surgery/ no radiation (XRT) were significantly associated with recurrence whereas the previously significant factor, stage, was not. Those whose tumors expressed p16 that were HPV had the least elevated risk of recurrence compared to the referent group. The p16 /HPV-HR group showed the greatest likelihood of recurrence, worse than those who were p16 /HPV , which would not have been predicted based
140 on the individual biomarker results, suggesting that alternative pathways that are not well understood contribute to this outcome. The p16+/HPV-HR group again had better 5year recurrence-free survival compared to the other groups: 60% versus 33%, p16+/HPV (p = 0.6), 0%, p16 /HPV-HR (p = 0.08), and 8%, p16 /HPV (p = 0.001). Prognosis based on the individual biomarkers, p16 and HPV-HR, were compared to those based on the joint group effects (Table 3). Whereas overall survival, disease specific survival, and recurrence among those who did not express p16 or who had HPV status were worse than among those who did, differences in prognostic outcomes were more distinct when the two markers were examined together. For example, examination of the p16 status alone did not distinguish the variation in disease-specific survival associated with HPV status (model 1, p16 HR = 2.0 versus p16/HPV, HRs 3.6, 2.7, and 2.8); or the greater risk of recurrence by HPV status among the p16 groups (model 1, HR = 3.6 versus p16/HPV, HRs 4.5 and 7.0). Although it has been established that HPV-HR HNC cases have better survival, this information alone did not reveal the further distinctions in clinical outcomes when p16 expression was assessed with HPV status. When examined by HPV status alone, recurrence was greater in the HPV tumors (HR = 2.0), yet the highest recurrence rate was among those with an HPV-HR, not HPV , tumor that was p16 . Those who were p16 /HPVHR had better disease-specific survival than those who were p16 /HPV yet worse recurrence. The addition of HPV status interestingly showed that the risk of recurrence was actually greater among those who were p16 /HPV-HR than among either the p16+/ HPV or p16 /HPV group.
Discussion This is the first study to examine p16 expression and HPV-HR status in a large group of HNC cases to compare these biomarkers individually and jointly as predictors of clinical outcomes. Unlike most previous studies of HNC prognostic indicators, our investigation controlled for other risk factors and pathologic characteristics of the disease to examine survival associated with the individual and joint biomarkers. The results show different survival and recurrence outcomes in each of the four p16/HPV biomarker groups. Some studies5,29 have suggested that detection of p16 expression by immunohistochemistry could be used as a surrogate marker of an HPV-related tumor. If this were true, the p16INK4a IHC test, which is easier and quicker to perform than current HPV assays, could be used to predict tumors that are causally related to HPV and which have been shown to have better survival than HPV tumors.26,30 However, we found that p16 expression and HPV detection in the tumors diverged in 20% of cases, indicating that one cannot serve as a surrogate marker of the other in HNC. Furthermore, clinical outcomes varied between those patients whose cancers were p16+ and either HPV-HR or HPV . To verify these discordant joint marker results, a random sample of the p16 /HPV-HR and p16+/HPV cases were rerun and confirmed the original findings. Thus our data suggest that a test of p16 status alone is inadequate in predicting clinical outcomes. The significantly more favorable survival and less frequent disease recurrence among the p16+/HPV-HR HNC
E.M. Smith et al. cases compared to p16 /HPV cases was expected based on our and other findings.11,14,26,30 However, 20% of the cases were positive for one but not the other biomarker in this study. Although p16 upregulation in HPV HNC tumors has been reported in other small studies,9,10 we are the first to show that p16+/HPV patients have worse survival than the p16+/HPV-HR group but significantly lower tumor recurrence than either p16 /HPV (+ or ) group. The underlying molecular changes that result in HNC lesions that are p16 /HPV-HR, described here as well as in previous investigations,31,32 remain unclear. Because we took the precaution to laser microdissect tissues containing <20% of tumor cells, it is unlikely that the virus resided outside of the tumor itself. Genetic changes that would alter p16INK4a function, such as homozygous deletions, point or missense mutations have been reported in HNC but these were not examined in association with HPV status.9,10,31,33–35 In two small series, loss of p16INK4a expression in HPV-HR cases has been ascribed to alterations including loss of heterozygosity, point mutations or silencing by methylation.32,35 In the future, it will be important to examine the nature of p16INK4a alterations in HPV-HR HNC lesions to understand how such mutations affect pRB function in HPV+ tumors and whether the prognosis in cases with these mutations is independent of HPV/pRB effects. Changes that would lead to p16 overexpression in p16+/ HPV tumors have not been defined. p16 expression in HNC has varied between 44 and 80%9,13,14,36 with 35% in our study. The high frequency of p16+ tumors in other reports is likely due to the inclusion of more oropharyngeal and tonsillar HNCs which are more commonly HPV-HR-associated. Some of these differences may be due to the evaluation of p16 staining. In our study we included both categories 2 (slightly positive staining, n = 2) and 3 (10–80% positive, n = 9) but the inclusion or exclusion of the small number of cases in these two categories did not alter the statistical findings in our large sample size. Among studies that examined both p16 and HPV in the same patients, one found no association with prognosis16 whereas others found better survival among those who were positive for both markers.11,14 Differences in prognostic outcomes may be attributed to the fact that most investigations did not adjust for clinicopathologic factors potentially associated with prognosis or included different HNC sites. In contrast, we adjusted our results for stage, grade, tumor site and lymph node involvement and found more advanced pathological characteristics among those who were p16+ and HPV-HR. The reason for the significantly better survival shown by us and others in p16+/HPV-HR cases is as yet unclear. A further difference between this and previous studies is the duration of the follow-up period to evaluate outcomes. In a study of 57 SCC cases in the oral cavity and oropharynx, Paradiso et al.16 found no association between p16 status with HPV-HR status or with survival. In contrast, Bova et al.13 in 148 patients limited to cancer of the anterior tongue, found that while lack of p16 expression was not associated with histologic parameters, it was a significant independent predictor of reduced overall and disease-specific survival, similar to our findings. In the series of HNC from multiple tumor sites presented here, the adjusted HRs showed that lack of p16 expression was not significantly associated with worse clinicopathologic param-
p16INK4a and survival in HPV–Associated HNC eters but p16 patients did have worse survival. Geisler et al.33 examined 171 HNC cases and found no association between survival or recurrence and loss of p16 expression using IHC. Geiser and colleagues33 also examined the association between p16 status and radiation treatment for clinical outcomes and found reduced recurrence with this treatment (adj.HR = 0.5, 0.2–1.2), but did not assess HPV status which may have accounted for the improved survival. We also found better disease-specific survival and reduced recurrence among those who received XRT as well as those who had surgery only. Klussmann et al.14 found significantly better disease-specific survival in a group of 34 tonsillar cancer cases who expressed p16 compared to those who did not (70% versus 18%). They examined the combined effects of p16 expression and HPV-positivity and found that 89% of the HPV-HR tumors showed p16 expression. This is the largest study of p16 expression in HNC cases. We included all HNC sites and found that the three major tumor sites did not have significantly different prognostic outcomes after adjusting for HPV status, stage, treatment and 94% of the HPV cases were p16 . Although they found no difference in disease-specific survival by p16/HPV-HR status,14 recurrence occurred significantly more frequently in p16 /HPV cases as compared to those who were positive for both markers. In contrast, we found both better diseasespecific survival and less recurrence in the p16+/HPV-HR group compared to each of the other p16/HPV groups, which may be due to the larger sample or examination of multiple tumor sites, and other confounders. This does not preclude that there are differences in clinical outcomes by tumor subsites (e.g. buccal mucosa, tonsils) associated with p16 expression and HPV-HR status, but a larger sample size is needed to examine these effects. Another advantage of our study was the minimal loss to and high verification of follow-up status using the NCI SEER Cancer Registry and US National Death Index databases. In conclusion, the combined p16/HPV biomarker data demonstrate an association with different survival and recurrence for HNC compared to each marker evaluated separately. This finding suggests that the two markers should be evaluated together in other large studies to verify whether these joint markers better predict prognosis and to help target HNC treatment to specific molecular pathways in the future.
141
3.
4.
5.
6.
7.
8. 9.
10.
11.
12.
13.
14.
15.
Acknowledgements 16.
This study was supported by a grant from NIH NIDCR R01 DE11979 (EMS, YK, DW, LMR, THH, LPT), NIDCR R01 DE13110 (DW, LMR, EMS, LPT, THH), and Veterans Affairs Merit Review Funds (LPT, THH).
References 1. Smith EM, Ritchie JR, Summersgill KF, Hoffman HT, Wang DH, Haugen TH, et al. Human papillomavirus in oral exfoliated cells and risk of head and neck cancer. J Natl Cancer Inst 2004;96:449–55. 2. Mork J, Lie AK, Glattre E, Hallmans G, Jellum E, Koskela P, et al. Human papillomavirus infection as a risk factor for
17.
18.
19.
20.
squamous-cell carcinoma of the head and neck. N Engl J Med 2001;344:1125–31. Kreimer AR, Clifford GM, Boyle P, Franceschi S. Human papillomavirus types in head and neck squamous cell carcinomas worldwide: a systematic review. Cancer Epidemiol Biomark Prev 2005;14:467–75. Li Y, Nichols MA, Shay JW, Xiong Y. Transcriptional repression of the d-type cyclin-dependent kinase inhibitor p16 by the retinoblastoma susceptibility gene product pRb. Cancer Res 1994;54:6078–82. Wang SS, Trunk M, Schiffman M, Herrero R, Sherman ME, Burk RD, et al. Validation of p16INK4a as a marker of oncogenic human papillomavirus infection in cervical biopsies from a population-based cohort in Costa Rica. Cancer Epidemiol Biomark Prev 2004;13:1355–60. Klaes R, Benner A, Friedrich T, Ridder R, Herrington S, Jenkins D, et al. P16 immunohistochemistry improves interobserver agreement in the diagnosis of cervical intraepithelial neoplasia. Am J Surg Pathol 2002;26:1389–99. Sano T, Masuda N, Oyama T, Nakajima T. Overexpression of p16 and p14arf is associated with human papillomavirus infection in cervical squamous cell carcinoma and dysplasia. Pathol Int 2002;52:375–83. Lowe SW, Sherr CJ. Tumor suppression by Ink4a-Arf: progress and puzzles. Curr Opin Gene Dev 2003;13:77–83. Lang JC, Tobin EJ, Knobloch TJ, Schuller DE, Bartynski KJ, Mountain RE, et al. Frequent mutation of p16 in squamous cell carcinoma of the head and neck. Laryngoscope 1998;108: 923–8. El-Naggar AK, Lai S, Clayman G, Lee J-K, Luna MA, Goedfert H, et al. Methylation, a major mechanism of p16/CDKN2 gene inactivation in head and neck squamous carcinoma. Am J Pathol 1997;151:1767–74. Weinberger PM, Yu Z, Haffty BG, Kowalski D, Harigopal M, Brandsma J, et al. Molecular classification identifies a subset of human papillomavirus-associated oropharyngeal cancers with favorable prognosis. J Clin Oncol 2006;24:736–47. Andl T, Kahn T, Pfuhl A, Nicola T, Erber R, Conradt C, et al. Etiological involvement of oncogenic human papillomavirus in tonsillar squamous cell carcinomas lacking retinoblastoma cell cycle control. Cancer Res 1998;58:5–13. Bova RJ, Quinn DI, Nankervis JS, Cole IE, Sheridan BF, Jensen MJ, et al. Cyclin d1 and p16 expression predict reduced survival in carcinoma of the anterior tongue. Clin Cancer Res 1999;5:2810–9. Klussmann JP, Gultekin E, Weissenborn SJ, Wieland U, Dries V, Dienes HP, et al. Expression of p16 protein identifies a distinct entity of tonsillar carcinomas associated with human papillomavirus. Am J Pathol 2003;162:747–53. Li W, Thompson CH, O’Brien CJ, McNeil EB, Scolyer RA, Cossart YE, et al. Human papillomavirus positivity predicts favourable outcome for squamous carcinoma of the tonsil. Int J Cancer 2003;106:553–8. Paradiso A, Ranieri G, Stea B, Zito A, Zehbe I, Tommasino M, et al. Altered p16INK4a and Fhit expression in carcinogenesis and progression of human oral cancer. Int J Oncol 2004;24:249–55. Emmert-Buck MR, Bonner RF, Smith PD, Chuaqui RF, Zhuang Z, Goldstein SR, et al. Laser capture microdissection. Science 1996;274:998–1001. Fleming ID, Cooper JS, Henson DE, Hutter RVP, Kennedy BJ, Murphy GP, et al., editors. AJCC cancer staging manual. 5th ed. Philadelphia (PA): Lippincott-Raven; 1997. Summersgill KF, Smith EM, Levy BT, Allen JM, Haugen TH, Turek LP. Human papillomavirus in the oral cavities of children and adolescents. Oral Surg Oral Med Oral Pathol Oral Radiol Endo 2001;91:62–9. NCI SEER. http://seer.cancer.gov/csr/1973–1999; 2002, 2004.
142 21. National Death Index. http://www.cdc.gov/nchs/r&d/ndi/ ndi.htm. 22. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Amer Stat Assoc 1958;53:457–81. 23. Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966;50:163–70. 24. S-PLUS 2000 Academic Site, editor. Version 6.2. Seattle WA: Math Soft, Inc.; 2003. 25. Cox DR. Regression models and life tables. JR Stat Soc (B) 1972;34:187–220. 26. Ritchie JM, Smith EM, Summersgill KF, Hoffman HT, Wang D, Klussmann JP, et al. Human papillomavirus infection as a prognostic factor in carcinomas of the oral cavity and oropharynx. Int J Cancer 2003;104:336–44. 27. Allison PD. Survival analysis using the SAS System, a practical guide. Cary, NC: SAS Institute, Inc.; 1995. 28. SAS System for Windows, version 9.1. Cary, NC: SAS Institute; 2002–2003. 29. Dray M, Russell P, Dalrymple C, Wallman N, Anous G, Lelong A, et al. p16INK4a as a complementary marker of high-grade intraepithelial lesions of the uterine cervix. I: Experience with squamous lesions in 189 consecutive cervical biopsies. Pathol 2005;37:112–24. 30. Gillison ML, Koch WM, Capone RB, Spafford M, Westra WH, Wu L, et al. Evidence for a causal association between human
E.M. Smith et al.
31.
32.
33.
34.
35.
36.
papillomavirus and a subset of head and neck cancers. J Natl Cancer Inst 2000;92:709–20. Li W, Thompson CH, Cossart YE, O’Brien CJ, McNeil EB, Scolyer RA, et al. The expression of key cell cycle markers and presence of human papillomavirus in squamous cell carcinoma of the tonsil. Head Neck 2004;26:1–9. Fregonesis PA, Teresa DB, Duarte RA, Neto CB, Brancini de Oliveira MR, Soares CP. p16INK4a immunohistochemical overexpression in premalignant and malignant oral lesions infected with human papillomavirus. J Histochem Cytochem 2003;51:1291–7. Geisler SA, Olshan AF, Weissler MC, Cai J, Funkhouser WK, Smith J, et al. P16 and p53 protein expression as prognostic indicators of survival and disease recurrence from head and neck cancer. Clin Cancer Res 2002;8:3445–53. Lang JC, Borchers J, Danahey D, Smith S, Stover DG, Agrawal A, et al. Mutational status of overexpressed p16 in head and neck cancer: evidence for germline mutation of p16/p14arf. Int J Oncol 2002;21:401–8. Reed AL, Califano J, Cairns P, Westra WH, Jones RM, Koch W, et al. High frequency of p16 (cdkn2/mts-1/) inactivation in head and neck squamous cell carcinoma. Cancer Res 1996;56:3630–3. Ambrosch P, Schlott T, Hilmes D, Ruschenburg I. P16 alterations and retinoblastoma protein expression in squamous cell carcinoma and neighboring dysplasia from the upper aerodigestive tract. Virchows Arch 2001;438:343–9.