Immunohistochemical staining of formalin-fixed, paraffin-embedded tumor tissue with monoclonal antibodies directed against vascular endothelial growth factor (A), p53 (B), cyclin D1 (C), and epidermal growth factor receptor (D) (original magnification ×100, scale bar = 200 µm).
Smith BD, Smith GL, Carter D, DiGiovanna MP, Kasowitz KM, Sasaki CT, Haffty BG. Molecular Marker Expression in Oral and Oropharyngeal Squamous Cell Carcinoma. Arch Otolaryngol Head Neck Surg. 2001;127(7):780-785. doi:10-1001/pubs.Arch Otolaryngol. Head Neck Surg.-ISSN-0886-4470-127-7-ooa00175
To determine the relative prognostic value of p53, cyclin D1, epidermal growth factor receptor (EGFR), and vascular endothelial growth factor (VEGF) expression in patients with oral and oropharyngeal squamous cell carcinoma.
Retrospective cohort study.
Fifty-six patients with oral and oropharyngeal squamous cell carcinoma referred to the Department of Therapeutic Radiology at Yale–New Haven Hospital (Conn) between 1981 and 1992 who were treated with gross total surgical resection and postoperative external beam radiotherapy.
Archival tumor tissue was stained with monoclonal antibodies directed against these 4 oncoproteins and scored for staining intensity and percent distribution by a pathologist blinded to the patients' clinical outcomes. Frequency of marker expression was 48% for p53, 20% for cyclin D1, 64% for EGFR, and 41% for VEGF. In multivariate analysis, EGFR positivity was protective against locoregional relapse (relative risk [RR], 0.27; 95% confidence interval [CI], 0.11-0.66; P = .002). In contrast, advanced N stage predicted poor locoregional relapse-free survival (RR, 1.94; 95% CI, 1.03-3.66; P = .04). Predictors of poor overall survival in multivariate analysis included VEGF positivity (RR, 3.53; 95% CI, 1.75-7.13; P<.001) and black race (RR, 2.48; 95% CI, 1.05-5.85; P = .04). Cyclin D1 and p53 expression were not significantly associated with prognosis in this cohort.
In oral and oropharyngeal squamous cell carcinoma treated with surgery and postoperative radiotherapy, VEGF and EGFR expression may influence clinical outcome. If confirmed, these results have potential implications for the determination of patient prognosis and the development of biologically based pharmacotherapies.
IN RECENT YEARS, a number of studies have attempted to determine the prognostic relevance of certain molecular defects commonly found in head and neck squamous cell carcinoma (HNSCC). Information gained from these studies may ultimately supplement clinical staging, thereby enabling more accurate determination of patient prognosis and more effective selection of therapeutic modalities. In addition, identification of specific molecular defects that influence tumor behavior may provide new avenues for the development of molecularly based cancer pharmacotherapies.
Of the molecular markers studied in HNSCC, p53, cyclin D1, and epidermal growth factor receptor (EGFR) have received particular attention in multiple studies as possible factors influencing outcome.1p53 is a tumor suppressor gene that protects the integrity of the genome by initiating either apoptosis or cell cycle arrest following DNA damage.2Cyclin D1 is a proto-oncogene that responds to extracellular mitogens by promoting G1-phase progression through the cell cycle.3 Similarly, EGFR is a proto-oncogene that, when activated at the cell surface by transforming growth factor-α, serves to promote cellular proliferation.4 Inactivation of tumor suppressor genes such as p53 and activation of proto-oncogenes such as cyclin D1 and EGFR results in deregulation of the cell cycle and thus promotes neoplasia. These alterations are amenable to assessment via immunohistochemistry, as p53 gene mutations often confer stability to the p53 protein resulting in elevated intracellular levels,5 and activation of cyclin D1 and EGFR is often associated with increased protein expression.
Another possible determinant of tumor aggressiveness is vascular endothelial growth factor (VEGF), a protein that is induced under hypoxic conditions and promotes vascular permeability and endothelial cell proliferation.6 We have recently published the first report to demonstrate that VEGF protein levels are an independent risk factor for poor survival in HNSCC.7 This study showed high levels of VEGF expression to correlate with local recurrence, distant metastatic disease, poor disease-free survival, and poor overall survival.
Given these intriguing results, the present study examined the same cohort of patients to compare the prognostic utility of VEGF protein levels with the prognostic value of p53, cyclin D1, and EGFR protein levels. Specifically, this study used multivariate modeling to assess the prognostic significance of VEGF, p53, cyclin D1, and EGFR protein levels with respect to the specific end points of locoregional recurrence-free survival and overall survival. In an effort to ensure homogeneity and enhance the validity of the study, the patient population selected was limited to patients with squamous cell carcinoma of the oral cavity and oropharynx who were treated with gross total surgical resection and postoperative external beam radiation therapy. In addition, we sought to compare the prognostic utility of these molecular markers with standard clinicopathologic variables, including TNM stage, tumor site, tumor grade, margin status, race, sex, and age at diagnosis.
Criteria for patient inclusion in this study were as follows: (1) presentation to the Department of Therapeutic Radiology at Yale–New Haven Hospital (Conn) between 1981 and 1992 with a histologically confirmed diagnosis of squamous cell carcinoma in the oral cavity or oropharynx and (2) treatment with primary surgical excision and postoperative external beam radiotherapy with curative intent. Patients were excluded from this study if they had a prior history of HNSCC, presented with metastatic disease, or failed to receive a full course of radiation therapy. A review of radiation records identified 77 patients who met the entry criteria. Of these, 14 had incomplete follow-up and 10 tissue samples were unavailable, leaving 56 patients for inclusion in the study.
With the approval of the appropriate institutional review boards, paraffin-embedded tissue samples were obtained from the hospital archives. Covariables including demographic, staging, clinical, pathologic, and treatment parameters were extracted from patient charts. Local recurrence was defined as recurrence of disease at a site within the upper aerodigestive tract anatomically contiguous with the primary tumor. Regional recurrence was considered recurrence of disease within the cervical lymphatic system. Distant recurrence was considered recurrence of disease that did not meet the definitions of local or regional recurrence and was not considered to represent a second primary tumor based on its histological and/or clinical manifestations. Forty-one patients (73%) were followed up until death, and the median follow-up time of the surviving patients was 6.1 years with a minimum of 2.8 years.
All patients were treated with gross total surgical resection and postoperative external beam radiotherapy to a median dose of 60 Gy. Final surgical margins were negative in 38 patients (68%). Patients with evidence of tumor approaching within 1 high-power field of final surgical margins were categorized as having positive margins in accordance with the method of Beitler et al.8 Forty-nine patients (88%) received a radical or modified neck dissection. Eleven patients (20%) received adjuvant chemotherapy with the hypoxic cytotoxins mitomycin or porfiromycin as part of an institutional protocol.9,10 Five patients (9%) received intraoperative brachytherapy. All patients were staged clinically according to the American Joint Committee on Cancer TNM classification system.11 Clinically N0 patients were restaged for the purposes of this analysis if pathologic examination results of the neck dissection specimen were positive for nodal metastases.
Immunohistochemical analysis was performed on 5-µm-thick tissue sections prepared from formalin-fixed, paraffin-embedded archival tissue from the resected primary tumor. Tissue sections were deparaffinized and then quenched in 2% hydrogen peroxide–methanol solution. Samples stained for VEGF were pretreated by microwaving at 50% power 3 times for 5 minutes in 10-mmol/L sodium citrate (pH 6.0). Samples stained for EGFR were pretreated with 0.4% pepsin in 0.1% hydrochloric acid for 15 minutes at 37°C. Samples stained for p53 and cyclin D1 did not require an antigen retrieval step. Slides were incubated overnight at 4°C with the following antibodies: (1) a mouse monoclonal IgG1 reactive with the 34- to 50-kd isoforms of VEGF (1:200 dilution, clone JH121; Oncogene Research Products, Cambridge, Mass); (2) a mouse monoclonal IgG2b reactive with wild-type and mutant forms of p53 (1:50 dilution, clone DO-7; DAKO, Carpinteria, Calif); (3) a mouse monoclonal IgG2b reactive with human cyclin D1 (1:2500 dilution, clone A-12; Santa Cruz Biotechnology, Santa Cruz, Calif); and (4) a mouse monoclonal IgG1 antibody specific for the protein portion of the extracellular domain of EGFR (prediluted, clone E 30; BioGenex, San Ramon, Calif). The next day, slides were incubated with horse antimouse secondary antibody, labeled with avidin-biotin complex streptavidin-peroxidase (Elite; Vector Laboratories, Inc, Burlingame, Calif), incubated with the chromogen diaminobenzidine tetrahydrochloride, counterstained with hematoxylin, and mounted.
An experienced pathologist (D.C.), blinded to the clinical outcomes, examined multiple microscopic fields to score the tissue sections for tumor staining intensity (0, none; 1, faint and focal; 2, moderate; 3, strong; 4, intense) and percent distribution. Positivity for cyclin D1 staining was defined as 5% or more staining distribution in accordance with the method of Pignataro et al.12 Positivity for p53 staining was defined as 10% or greater staining, a cut point used in previous studies.13,14 To maximize the power of statistical comparisons using the VEGF status variable, VEGF status was determined by dichotomizing the VEGF staining intensity variable with the cut point at the median between moderate and strong staining. The EGFR status was also determined by dichotomizing the EGFR staining intensity variable with the cut point at the median between faint and moderate staining. The EGFR status was not divided into tertiles as described by Grandis et al15 due to the smaller sample size in this study.
Molecular marker status and relevant covariables were assembled in a database and analyzed using statistical software.16 All tests of statistical significance were 2-sided. Follow-up time and time to recurrence were calculated from the date of surgery to the date of the relevant outcome. To enhance the power of statistical comparisons, the following categories were collapsed: T stages 1 and 2, T stages 3 and 4, N stages 2 and 3, and tumor grades poorly and poorly to moderately differentiated.
Bivariate analyses for the associations between predictor variables and the main outcome variables (locoregional relapse-free survival and overall survival) were conducted using the Kaplan-Meier log-rank test. Variables violating the linearity assumption entered into the analysis as categorical variables, with cut points determined by quartiles. Unadjusted relative risks were calculated using a Cox proportional hazards model.
In the multivariate analysis, Cox proportional hazards regression determined significant predictors of locoregional recurrence-free survival and overall survival at an α = .05 level in the final model. Molecular marker status and all clinicopathologic covariables were initially entered into a forward parsimonious selection. Variables with P≤.10 were retained for the final model.
A total of 77 patients were eligible for the analysis, but 21 were excluded due to incomplete follow-up data and/or insufficient archival tissue, leaving 56 patients (73%) in the analysis. Frequency statistics for clinicopathologic characteristics of the patient cohort are presented in Table 1. Fifty-one patients (91%) presented with stage III or IV disease, with the remaining 5 patients presenting with stage II. A total of 14 patients (25%) experienced local recurrence, 10 had regional recurrence (18%), and 14 had distant recurrence (25%). Forty-one patients (73%) died during the follow-up period. Five-year survival for this cohort was 40% (22 patients).
Positivity rates for molecular marker expression were as follows: 41% for VEGF, 48% for p53, 20% for cyclin D1, and 64% for EGFR. Examples of tumors positive for each of these molecular markers are presented in Figure 1. The VEGF-positive tumors demonstrated cytoplasmic granular staining, p53 and cyclin D1–positive tumors displayed nuclear staining, and EGFR-positive tumors showed membranous staining.
The relationships between clinicopathologic covariables and the main outcome measures in this study (locoregional relapse-free survival and overall survival) are presented in Table 1. None of the clinicopathologic variables were significantly predictive of locoregional relapse-free survival in log-rank statistics at α = .05. N stage, however, was significantly predictive of locoregional relapse-free survival in an unadjusted Cox proportional hazards model (relative risk [RR], 1.97; 95% confidence interval [CI], 1.03-3.80). Several clinicopathologic covariables were significant risk factors for poor overall survival, including advanced T stage (RR, 2.09; 95% CI, 1.01-4.32; P = .04), advanced N stage (RR, 1.91; 95% CI, 1.20-3.05; P = .02), and black race (RR, 2.57; 95% CI, 1.19-5.57; P = .01).
The relationships between molecular marker positivity and the main outcome variables are presented in Table 2. The EGFR positivity was strongly protective against locoregional recurrence (RR, 0.27; 95% CI, 0.11-0.66, P = .002). Borderline predictors of locoregional recurrence included VEGF status (RR, 2.09; 95% CI, 0.87-5.02; P = .09) and p53 status (RR, 0.45; 95% CI, 0.18-1.17; P = .09). With respect to overall survival, the only significant predictor was VEGF status (RR, 3.21; 95% CI, 1.63-6.32; P<.001).
Finally, the relationships between molecular marker expression and distant recurrence were examined (Table 2). Significant predictors included VEGF status (RR, 4.62; 95% CI, 1.41-15.10; P = .006), p53 status (RR, 3.02; 95% CI, 0.95-9.64; P = .05), and cyclin D1 status (RR, 2.88; 95% CI, 0.96-8.65; P = .05).
Both Cox proportional hazards models included all 56 patients. The VEGF, p53, cyclin D1, and EGFR status, along with the aforementioned covariables, were entered into the model. For locoregional relapse-free survival, the variables N stage and EGFR status were retained at a level of P = .10 (Table 3). With a final α = .05 criterion, both variables were predictive of locoregional relapse-free survival. The most significant predictor, EGFR positivity, was protective against locoregional recurrence (RR, 0.27; 95% CI, 0.11-0.66; P = .002). Thus, EGFR-negative tumors recurred locally or regionally more than 3 times as often as EGFR-positive tumors. Expression of VEGF, cyclin D1, and p53 was not significantly associated with locoregional relapse-free survival in this multivariate model. N stage, the only clinicopathologic variable included in the final model, was positively correlated with locoregional recurrence (RR, 1.94; 95% CI, 1.03-3.66; P = .04).
For overall survival, age, race, and VEGF status were retained at a level of P = .10 (Table 3). Because age was entered as a 3-level dummy variable, all age categories were forced into the final model. The most significant predictor of poor survival in this cohort was VEGF status, with death more than 3 times as likely in VEGF-positive patients (RR, 3.53; 95% CI, 1.75-7.13; P<.001). Despite entry of cyclin D1, EGFR, and p53 status into the initial model, none of these markers significantly predicted overall survival when controlling for other relevant variables. Black race, the only demographic variable included in the final model, was associated with increased mortality (RR, 2.48; 95% CI, 1.05-5.85; P = .04). T stage and N stage were not predictive of survival in multivariate modeling.
This study, conducted on patients with oral cavity and oropharyngeal primary tumors treated with surgical excision and postoperative radiotherapy, demonstrates the potential prognostic utility of molecular marker analysis. Specifically, low levels of EGFR expression were predictive of poor locoregional recurrence-free survival in multivariate analysis. In addition, high levels of VEGF expression were highly predictive of poor overall survival in multivariate analysis. Thus, in this cohort of patients with relatively advanced disease and poor prognosis, EGFR and VEGF status proved to be better indicators of outcome than traditional staging parameters, such as T stage and N stage. In addition, EGFR and VEGF status were more important predictors of outcome than p53 and cyclin D1 expression.
This study was limited to a relatively homogeneous cohort to control for variations in tumor behavior imposed by different primary tumor sites and therapeutic modalities. These criteria enhanced the internal validity of the study, while limiting its external validity and power. Larger studies will be needed to verify the results of this study and determine their applicability to other sites within the head and neck. Furthermore, the scoring systems used in this study to define VEGF and EGFR positivity have not been used in previous studies and will need to be prospectively verified in future work.
This study was also limited to the qualitative immunohistochemical assessment of oncoprotein expression. This technique offers important advantages because it is relatively inexpensive and readily available in most pathology departments. Hence, molecular markers that can be assessed via immunohistochemical methods may be more easily integrated into clinical practice than markers dependent on various techniques of molecular biology. However, qualitative immunohistochemistry is subject to interobserver variability and, at least in the case of p53, does not always correlate with the specific genotypic changes shown to be associated with poor prognosis.5 As a result, a study that uses immunohistochemistry alone can never rule out the possibility that a specific genetic marker affects prognosis.
Despite these limitations, the results of this study provide novel information that may prove clinically relevant. Because a high level of EGFR was the most important predictor of locoregional control, EGFR may serve as a marker of radiosensitivity in this cohort. Indeed, at least 3 studies conducted on cell lines derived from human squamous cell carcinomas have shown that in vitro activation of EGFR enhances radiosensitivity.17- 19 However, in a different cell line derived from a human HNSCC, Huang et al20 found that inhibition of EGFR with a specific monoclonal antibody enhanced radiosensitivity. Hence, the role of EGFR in mediating radiosensitization may depend on the downstream signaling components present in a specific clonal population of neoplastic cells. This study suggests that, in neoplasms derived from the oral cavity and oropharynx, high levels of EGFR expression may correlate with increased radiosensitivity.
In contradistinction to the results of this study, at least 3 excellent clinical studies have shown a relationship between elevated levels of EGFR and poor prognosis.15,21,22 These conflicting results could be attributed to differences in the anatomical subsites of HNSCC included in the specific studies and differences in therapeutic modalities used. Indeed, none of these studies were limited to oral cavity and oropharyngeal primary tumors, and only 1 study included a significant number of patients (62%) treated with radiotherapy.15 In contrast to these 3 studies, Wen et al23 found that EGFR expression was not predictive of recurrent disease or overall survival in a cohort of 68 patients with early stage laryngeal cancer treated exclusively with radiotherapy. Hence, although elevated EGFR expression may generally portend aggressive disease, the specific contribution of EGFR expression to radiosensitivity has yet to be consistently determined in clinical series. Additional clinical information regarding the role of EGFR in mediating response to radiotherapy will be forthcoming, as a phase 3 clinical trial using C225, a monoclonal antibody targeted against EGFR, recently began accrual for patients with advanced head and neck tumors.24
In this cohort, VEGF was the only molecular marker that significantly predicted overall survival in multivariate analysis. Hence, VEGF may prove to be a more important determinant of prognosis in patients with head and neck cancer than EGFR, p53, and cyclin D1. In addition, VEGF expression was a more significant predictor of overall survival than clinicopathologic variables such as T stage and N stage. If this relationship is verified in further studies, quantitation of tumor VEGF expression may provide useful information in staging for patients with head and neck cancer.
Vascular endothelial growth factor may mediate poor prognosis through several potential mechanisms. For example, because VEGF is induced under hypoxic conditions, and hypoxic tumors tend to display increased radioresistance, VEGF protein expression may serve as a surrogate marker of hypoxic radioresistance. This hypothesis is consistent with the borderline correlation between VEGF levels and locoregional relapse observed in this study. In addition, VEGF expression likely promotes distant metastatic disease by at least 2 mechanisms. First, angiogenesis may promote metastatic disease by exposing tumors to a greater endothelial surface area, thus increasing the likelihood of hematogenous dissemination.25 In addition, once a tumor has formed a distant micrometastasis, it must recruit a vascular supply to proliferate to a clinically significant size. These hypotheses are consistent with the strong relationship between VEGF expression and distant metastatic disease observed in this study.
Ultimately, the growing literature on molecular markers in head and neck cancer should contribute to the management of patients with HNSCC by enabling more precise staging and more elegant selection of therapeutic modalities. For example, certain markers may enable prediction of the tumor response to radiation therapy, with obvious clinical implications. In addition, certain molecular defects are amenable to pharmacological modification or possibly replacement via gene therapy. Continued efforts to evaluate the influence of molecular lesions on tumor behavior will prove important in the identification of certain pathways for which intervention will result in clinically significant advances in the treatment of patients with HNSCC.
Accepted for publication November 14, 2000.
Presented at the annual meeting of the American Head and Neck Society, Fifth International Conference on Head and Neck Cancer, San Francisco, Calif, August 1, 2000.
Corresponding author and reprints: Bruce G. Haffty, Department of Therapeutic Radiology, Yale University School of Medicine, PO Box 208040, New Haven, CT 06520-8040 (e-mail: firstname.lastname@example.org).