Biological Markers and Prognosis in Recurrent Oral Cancer After Salvage Surgery

Objective  To analyze the prognostic effect of epidermal growth factor receptor (EGFR), matrix metalloproteinases 2 and 9, and vascular endothelial growth factor expression in patients with locally recurrent oral carcinoma after salvage surgery.

Design  Retrospective cohort study.

Settings  Tertiary center cancer hospital.

Patients  The charts of 111 patients with local recurrence of oral carcinomas were retrospectively analyzed. The previous treatment consisted of surgery in 33 patients (30.0%), radiotherapy with or without chemotherapy in 46 patients (41.0%), and surgery with adjuvant radiotherapy in 32 patients (29.0%). The expression of EGFR, matrix metalloproteinases 2 and 9, and vascular endothelial growth factor was analyzed with a tissue microarray immunohistochemical technique.

Main Outcome Measures  Overall survival and cancer-specific survival (CSS).

Results  The recurrences were diagnosed in less than 1 year in 69 patients (62.2%) and in more than 1 year in 42 patients (37.8%). The prognosis was worse in the group with the disease-free interval of less than 1 year (P = .01). Patients with more advanced disease (clinical stage of recurrence, III/IV) had worse rates of CSS (P = .04). Cases that were positive for EGFR had a 3-year CSS of 27.2%, while EGFR-negative cases had a 3-year CSS of 64.3% (P = .001). The expression of matrix metalloproteinases 2 (P = .83) and 9 (P = .15) and vascular endothelial growth factor (P = .86) was not significant in this group. In multivariate analysis, only the disease-free interval and the overexpression of EGFR were associated with a higher risk of cancer death.

Conclusions  Local recurrence in oral carcinomas carries a poor prognosis. A disease-free interval of more than 1 year and a EGFR-negative expression are the main prognostic factors related to better CSS in patients treated with salvage surgery.

Patients with carcinomas of the oral cavity and oropharynx present a high risk of local recurrence after initial treatment with surgery, radiotherapy, surgery with adjuvant radiotherapy, or radiochemotherapy. For clinical stage (CS) III or IV tumors, the rates of local recurrence may be as high as 40%.^1,2 Salvage surgery is considered the only curative therapeutic option in such cases.^3,4 The main prognostic factors are the disease-free interval (DFI) and the CS of recurrence (rCS). A DFI of more than 1 year and rCS I and II are associated with a better survival rate after salvage surgery.⁵ However, such high-cost surgical procedures can result in significant morbidity, and the long-term results are considered to be inadequate by some authors.^6,7 Patient selection based only on clinical factors does not seem to be accurate enough in identifying those patients who might benefit the most from salvage surgery.⁵

The aim of this study was to assess the prognostic impact of biological markers involved in the process of invasion in patients with local recurrence of cancer of the mouth and oropharynx after salvage surgery. Analysis of the expression of (1) matrix metalloproteinases 2 and 9 (MMP-2 and MMP-9), enzymes capable of cleaving extracellular matrix components,⁸ (2) epidermal growth factor receptor (EGFR), a receptor associated with cellular growth and differentiation,⁹ and (3) vascular endothelial growth factor (VEGF), a protein associated with angiogenesis,¹⁰ was performed to identify potentially useful variables for future therapeutic decisions.

We reviewed the charts of 111 patients who underwent salvage surgery for locally recurrent squamous cell carcinomas (SCCs) of the oral cavity or oropharynx from 1990 to 2004 at Hospital A. C. Camargo, São Paulo, Brazil. Only patients with neck stage classified as N0 (cTNM) by clinical examination and computed tomography or magnetic resonance imaging were included in the study. Eighty-one patients (73.0%) were men and 30 were women (27.0%). One hundred one patients were white (90.1%), 8 were Afro-Brazilian (7.2%), and 2 were Asian (1.7%). They ranged in age from 31 to 91 years, with a median of 56 years.

The site of the primary tumor was the lip in 10 cases (9.0%), oral cavity in 68 cases (61.3%), and oropharynx in 33 cases (29.7%). The most commonly affected site in the oral cavity was the tongue, with 27 cases; in the oropharynx, the majority of patients had tumors in the tonsil (22 cases) (Table 1). Thirty-three patients (29.7%) had been previously treated with surgery alone (15 with neck dissection); in 32 patients (28.8%), the previous therapy consisted of surgery (25 with neck dissection) with adjuvant radiotherapy; and in 46 patients (41.4%), the treatment consisted of radiotherapy alone or radiotherapy and cisplatin-based chemotherapy (Table 1). Radiotherapy doses varied from 4400 to 7400 cGy.

The DFI, defined as the period between the date of the end of the initial treatment and the date at diagnosis of the recurrence, ranged from 0.89 to 140.9 months, with a median of 6.87 months. Sixty-nine patients (62.2%) presented with a recurrence within 1 year of the initial treatment (early recurrence), while in 42 cases (37.8%), the recurrence was diagnosed after 1 year (late recurrence).

The 2002 version of the TNM system by the American Joint Committee on Cancer–Union Internationale Contre le Cancer¹¹ was used to classify the tumors by CS. The CS of the primary tumor at the time of previous treatment was identified as initial (CS I or II) in 26 patients (23.4%) and as advanced (CS III or IV) in 85 patients (76.6%). The rCS maintained the same tendency of presentation in the advanced stages of the disease. Eighty patients (72.1%) presented with local recurrence at an advanced stage (rCS III or IV), while 31 patients (27.9%) presented with disease at an initial CS (rCS I or II). The surgical treatment of these patients consisted of resection of the recurrent primary site lesion with 3-dimensional margins, along with a neck dissection (en bloc surgery) in 56 cases (50.5%). Neck dissection was unilateral in 51 cases and bilateral in 5 cases. In 27 patients (48.2%), node metastases were diagnosed during the pathologic examination (pN⁺). The surgical margins after salvage surgery were negative in 79 cases (71.2%) and positive or close in 32 cases (28.8%). Adjuvant radiotherapy (dose range, 4000-7000 cGy) was used in 23 patients (16.2%) who had not previously undergone this modality of treatment. A second course of radiotherapy was administered to 7 patients (6.3%), because most of them had not received a radical dose during the initial treatment. The reirradiation doses ranged from 4200 to 6000 cGy.

Surgical defect reconstruction by primary closure was achieved in 48 patients (44.0%). In all other patients, a flap was necessary to reconstruct the primary site defect. A pectoralis major myocutaneous flap was used in 49 patients (44.1%); a deltopectoral Backanjian flap was used in 1 patient; and a microvascular free flap was used in 13 patients (11.7%). Hospitalization time after the salvage surgery ranged from 1 to 60 days, with a mean of 10.2 days and a median of 7 days. Fifty-three patients (47.7%) had local postoperative complications (minor), the most common of which were wound infection (21.6%) and orocutaneous fistula (19.8%). Ten patients (9.0%) had systemic postoperative complications (major), the most common of which was pulmonary infection. Two patients died of systemic complications during the postoperative period (postoperative mortality rate, 1.8%).

The paraffin blocks and slides of the primary site tumor specimen submitted for each of the 111 cases had been identified in the archives of the pathology department. The slides were reviewed by a single pathologist (C.A.L.P.) to delimit the most significant area of each tumor, which was then used to construct the tissue microarray (TMA) from the paraffin blocks. A TMA paraffin receptor block (Beecher Instruments, Silver Spring, Maryland) was constructed from the samples collected from the original donor block using a 0.6-mm-diameter needle (TMArrayer Punch; Beecher Instruments) after previous selection and circular marking of the area representative of the tumor on the original hematoxylin-eosin–stained slide. The cases were placed on the block receptor in ascending numerical order and in duplicate. One hundred serial 3-μm-thick histologic cuts were made and fixed on glass slides with adhesive film (Microsystems Inc, Downers Grove, Illinois).

The cuts at levels 1, 10, 20, 40, 60, 80, and 100 were stained with hematoxylin-eosin. The slides were examined under an optic microscope by a single observer (C.A.L.P.) to assess the representativeness of the tumor cells. The protein expression of MMP-2, MMP-9, EGFR, and VEGF was immunohistochemically analyzed. The immunohistochemical reactions were performed in duplicate on the cuts obtained from the TMA block (total, 16 slides and 32 colors). The technique included the use of streptavidine-biotin-peroxidase complex (StreptABC; Dako Corp, Carpinteria, California).

All the slides were incubated with primary antibodies diluted in a preestablished titer in accordance with the antibody table. Then, they were incubated with the secondary antibody in the preestablished titer of 1:200. The antibodies were diluted in phosphate-buffered saline. The reactions were revealed with light-sensitive diaminobenzidine (Sigma-Aldrich Corp, St Louis, Missouri) solution and counterstained with Harris hematoxylin (Merck & Co inc, Whitehouse Station, New Jersey). The reactions were accompanied by positive (absence of primary antibody) and negative (absence of secondary antibody) controls. Immunohistochemical reactions were performed at 2 levels for each case, ie, in quadruplicate for each antibody used. The slides were used at levels 4 and 20 for MMP-2, at levels 6 and 7 for MMP-9, at levels 17 and 18 for VEGF, and at levels 26 and 36 for EGFR. Placental tissue was used as positive control. The primary antibody was omitted as negative control. Therefore, patient-specific normal tissue was not necessary for the TMA controls.

The slides were read under an optic microscope (× 400 magnification), and the expression of the antigens was characterized by 2 observers (I.M.G.A. and C.A.L.P.), accompanied by the first author, according to the following criteria: group 1, negative or absence of expression; group 2, weakly positive (irregular or homogeneous marking of the core with weak staining intensity); and group 3, strongly positive (regular or homogeneous marking of the core with strong staining intensity). The expression of the antigens was considered positive in the cases of cytoplasmic and membrane staining for MMP-2, MMP-9, and VEGF; the expression only of membrane staining was considered positive for EGFR.

A commercially available software package (SPSS Version 10.0; SPSS Inc, Chicago, Illinois) was used for statistical analysis. Survival analysis was performed using the Kaplan-Meier method, and the log-rank test was used to analyze the significance between the differences observed for the categories of the same variable. Other statistical correlations were assessed using the χ² test. Overall survival was considered as the time elapsed between salvage surgery and death or the last information received. Cancer-specific survival (CSS) was defined as the interval between the date of salvage surgery and the last objective follow-up information or death caused by the disease. The Cox proportional risk model was used to calculate the multivariate risk of death. Statistical significance was determined at P < .05.

Follow-up after salvage surgery ranged from 0.8 to 140.9 months, with a median of 11.94 months. At the end of the study follow-up, 47 patients (42.3%) were alive with no evidence of recurrent or metastatic cancer. Sixty-four patients (57.7%) presented with recurrences, the majority of which (36.0%) were local. Distant metastases occurred in 10 patients (9.0%), while 19 patients (17.1%) had regional neck recurrences (Table 2). Follow-up information was not available in 12 cases (10.8%). The median recurrence-free survival was 9.6 months. No treatment was possible in 21 of the 64 patients (18.9%) who presented with further recurrence. In 26 patients (23.4%), a second salvage surgery was performed. Five patients (4.5%) were treated with radiotherapy, and 12 (10.8%) received palliative chemotherapy. The overall survival rate was 58.8% at 1 year, 29.4% at 3 years, and 23.4% at 5 years. The CSS rate was 68.3% at 1 year, 37.1% at 3 years, and 31.7% at 5 years (Figure 1).

There were no statistically significant differences in CSS according to the variable of sex (P = .75). At 3 years, the CSS rate was null in patients younger than 45 years, 45.9% in patients between 45 and 60 years of age, and 20.7% in patients older than 60 years (P = .02) (Table 3). The site of the primary tumor also had no impact on the prognosis of these patients. The CSS rates tended to be worse in patients with oropharynx carcinomas: 30.1% at 3 years, compared with 40.2% in patients with oral cavity carcinomas and 37.5% in patients with lip carcinomas (P = .84) (Table 3). The rCS was important in defining prognosis. At 3 years, patients with recurrences in the initial CS (rCS I and II) had a CSS rate of 42.3%; those with more advanced disease (rCS III and IV) had a survival rate of 31.2% (P = .04) (Table 3). The initial treatment had no significant impact on prognosis. Patients who previously underwent surgery alone had a CSS rate of 49.8% at 3 years. When the previous treatment was radiotherapy alone, the survival rate at 3 years was 28.2%. In patients who were previously treated with both therapeutic modalities, the survival rate at 3 years was 28.9% (P = .12).

The DFI was a decisive factor in the results obtained with salvage surgery. The recurrences that occurred beyond 1 year after initial treatment were associated with the highest survival rates (51.2% of the patients included in this subgroup were alive at the end of 3 years of follow-up). Early recurrences were associated with the worst prognosis, with a CSS rate of 29.6% at 3 years (P = .01) (Table 3). The surgical resection margins were negative in more than 70% of the patients; this group had a better survival result: 38.7% were alive at 3 years. The patients who had small or involved surgical margins had a worse prognosis: 26.5% were alive at 3 years. However, the difference was not statistically significant (P = .79) (Table 3).

There was no difference in the expression of biological markers with regard to the primary tumor site in relation to MMP-2, MMP-9, and VEGF. However, there was a higher number of cases that were positive for EGFR in oral cavity carcinomas than in carcinomas of the lip and oropharynx. Fifty percent of the lip carcinomas, 71.0% of the oropharynx carcinomas, and 84.6% of the oral cavity carcinomas were positive for EGFR; this difference was significant (P = .03). The expression of EGFR was not assessed immunohistochemically in 5 cases (4.5%) owing to technical difficulties in constructing the TMA. There was no expression of the marker in 24 cases (21.6%); in 82 cases (73.9%), there was EGFR expression in the cytoplasmic membrane (Figure 2). Patients with tumors that expressed this membrane receptor had a CSS rate of 27.2% at 3 years. When EGFR was absent, the survival rate was 64.3% at 3 years (P < .001) (Table 3 and Figure 3).

The expression of MMP-2 was elevated in the majority of the tumors; it was positive in 88 cases (79.3%) and negative in 20 cases (18.0%). There was no suitable material for assessing the expression of this protein in only 3 cases (2.7%). With regard to the expression of MMP-9, there was no suitable material in 4 cases (3.6%); the majority of the cases (74%) were positive for this protein. Only 16 cases (14.4%) were negative for MMP-9. Tumors that were positive for MMP-2 and MMP-9 were associated with lower survival rates than those that were negative for these proteins. However, the difference was not statistically significant for MMP-2 or MMP-9 (P = .83 and P = .15, respectively) (Table 3).

Immunostaining demonstrated that the great majority of tumors in this sample of patients were positive for VEGF. Eighty-eight tumors (79.3%) expressed VEGF; it was not possible to assess this marker using TMA in only 1 case, because of a lack of material. It was shown that VEGF was not associated with a difference in the treatment results (Table 3).

The paraffin blocks of the primary tumor and recurrent tumor were available in 31 cases. We noted that there was a trend toward higher expression of EGFR in the recurrent tumors compared with the primary ones. Hyperexpression of EGFR was observed in 17 primary tumors (54.8%) and 27 recurrent tumors (87.1%).

In multivariate survival analysis, only the DFI and the expression of EGFR were associated with an independent significantly high risk of death. Patients with a DFI of less than 1 year (hazard ratio, 1.97; 95% confidence interval, 1.05-3.69) and patients with an overexpression of EGFR (hazard ratio, 4.20; 95% confidence interval, 1.56-11.28) had the worst prognosis (Table 4).

We analyzed the survival rates for the patients considered to have the best and the worst prognosis, in accordance with the results obtained by the univariate and multivariate analyses. The patients who presented with late recurrences (DFI >1 year) and were negative for EGFR (n = 11) were considered to have the best prognosis. Among this group, the 3-year CSS rate was 80%. The 3-year CSS rate was 20.7% among the patients who had early recurrences (DFI <1 year) and were positive for EGFR (worst prognosis, 54 patients).Furthermore, when we considered the rCS, which is a significant variable in univariate analysis, and then a group of 41 patients with the worst prognostic factors (DFI <1 year, rCS III or IV, and EGFR positive), we observed a 3-year CSS rate of 19.5%. In this latter group, only 3 patients were alive and disease-free at the end of follow-up (Figure 4).

Patients with recurrent carcinomas of the oral cavity or oropharynx pose a challenge to the medical team with regard to defining the best therapeutic approach. In the majority of cases, the disease presents at an advanced CS at the time of diagnosis; therefore, the prognosis in such cases is usually poor. However, salvage surgery is the only curative treatment option for selected patients with suitable clinical status and resectable tumors.^4,5

The DFI, a variable that was shown to be significant for defining the survival results in a previous study,⁵ was less than 1 year in the majority of cases. In many patients who undergo radiotherapy or radiochemotherapy, this interval may be null; ie, the disease persists owing to a nonresponse to the initial treatment. Other authors have also demonstrated that survival and locoregional control rates are lower in patients with early recurrences or persistence of disease than they are in patients with longer disease-free survival periods.^12,13 In the present study, the DFI was the most important clinical variable in defining prognosis. Patients who developed recurrent disease less than 1 year from initial treatment had the worst prognosis. In multivariate analysis, their risk of death was 97% higher than that of the controls (DFI >1 year). Other clinical and therapeutic variables did not show a significant impact on prognosis, similar to previous reported results.⁵ Despite the lower survival rates for patients with an advanced rCS (rCS III/IV) in the univariate analysis (P = .04), the rCS showed no impact on prognosis in multivariate analysis. This result may have occurred because only patients with local recurrence were included in the present study. However, based on our findings and those of a previous study⁵ in which locoregional recurrences were also included and in which this variable was significant for the survival result, we suggest that the rCS should influence the therapeutic decision.

The majority of tumors in our study were positive for MMP-2 and MMP-9. Patients with tumors that did not express these proteins had better survival rates. However, these differences were not statistically significant. Other authors have found similar results, but the controversy about the prognostic impact of MMPs on head and neck cancer remains. Kim et al,¹⁴ in a retrospective study of 38 patients with tongue carcinomas, concluded that hyperexpression of MMP-2 and MMP-9 occurred in the majority of cases, but there was no correlation with survival and/or locoregional control. Katayama et al¹⁵ also found no association between MMP-2 expression and the prognosis of 53 patients with oral cavity carcinomas. In the same series, patients with high expression of MMP-9 had worse CSS rates than those without expression. Yorioka et al¹⁶ found a correlation between high enzymatic activity of MMP-2 and MMP-9 and disease-free survival in patients with carcinomas of the oral cavity.

Expression of VEGF was found in the majority of studies that analyzed this protein in carcinoma of the head and neck. Kyzas et al¹⁷ assessed VEGF expression in 36 patients with carcinomas of the oral cavity and larynx. Using multivariate analysis, they found that when this protein was expressed, it was associated with a greater risk of recurrence and death, being the only independent variable with prognostic impact. A recently published meta-analysis of studies of VEGF expression in carcinomas of the upper aerodigestive tract¹⁸ confirmed the relationship of high expression of this growth factor with poor survival results. In the present study, there were no significant differences in the results of CSS among patients who did or did not express VEGF. In another study,¹⁸ VEGF expression had a clear association with N stage. In the present study, only patients with CS N0 were included, which is likely the justification for the disagreement with the results published by other authors.

The expression of EGFR was associated with prognosis in this series. Patients whose tumors did not express EGFR presented better treatment results after salvage surgery, with a CSS rate of 64.3% at 3 years, while those whose tumors expressed this protein had a CSS rate of 27.2% at 3 years. In a multivariate analysis, high EGFR expression was associated with an increase of up to 4.2 times in the risk of death.

Ang et al¹⁹ assessed the influence of EGFR immunohistochemical expression in patients with carcinomas of the head and neck who were treated with radiotherapy as part of a phase 3 study conducted by the Radiation Therapy Oncology Group. The study included 155 patients with SCC of the oral cavity, oropharynx, hypopharynx, and larynx. Overexpression of EGFR had a direct association with the result of treatment and was associated with the worst overall survival and disease-free survival results as well as with the worst result in locoregional control. In multivariate analysis, the high expression of this receptor was an independent prognostic factor. Chen et al²⁰ studied the prognostic effect of the expression of EGFR in 59 patients with SCC of the oral cavity who were treated with surgery combined with postoperative radiotherapy. Overexpression of EGFRwas associated with the worst overall survival rates in univariate and multivariate analyses.

The poor prognosis of patients with SCC of the head and neck and high expression of EGFR may be related to resistance to radiotherapy. Based on this hypothesis, a phase 3 study²¹ was conducted to compare radiotherapy alone with radiotherapy plus therapy with cetuximab, an anti-EGFR antibody, in patients with CS III or IV SCC of the oropharynx, larynx, or hypopharynx. There was better locoregional control, better progression-free survival, and better overall survival among patients who underwent both radiotherapy and cetuximab therapy.

Immunohistochemical analysis of EGFR expression in a biopsy specimen from a recurrent tumor seems to be a promising approach to salvage treatment planning. This study showed the poor prognosis of patients with recurrent tumors that overexpressed EGFR. To our knowledge, all previous reported studies involved patients who had no previous treatment. As far as we are aware, this is the first study to demonstrate the prognostic significance of EGFR expression in recurrent tumors. The natural question for future clinical trials involves the possibility of association of salvage surgery and targeted therapy for patients with EGFR-positive tumors.

The present study showed that cases involving locally recurrent carcinomas of the oral cavity and oropharynx can be stratified according to prognosis on the basis of clinical information available for diagnosis (DFI), which, in association with the findings of immunohistochemical analysis of EGFR, may be useful in defining the salvage therapeutic approach. Therefore, we suggest that the algorithm shown in Figure 5 could be useful in the diagnosis and treatment of cases of resectable local recurrence of SCC of the oral cavity and oropharynx.

Correspondence: Luiz Paulo Kowalski, MD, PhD, Department of Head and Neck Surgery–Otorhinolaryngology, Hospital A. C. Camargo, Rua Professor Antonio Prudente, 211, 01509-900 São Paulo, Brazil (lp_kowalski@uol.com.br).

Submitted for Publication: May 25, 2007; final revision received June 19, 2007; accepted July 31, 2007.

Author Contributions: Dr Agra had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Agra, Carvalho, Pinto, Soares, and Kowalski. Acquisition of data: Agra, Martins, and Filho. Analysis and interpretation of data: Agra, Carvalho, Filho, Soares, and Kowalski. Drafting of the manuscript: Agra, Carvalho, and Martins. Critical revision of the manuscript for important intellectual content: Kowalski, Carvalho, Pinto, Filho, Soares, and Kowalski. Statistical analysis: Carvalho, Filho, and Soares. Obtained funding: Soares and Kowalski. Administrative, technical, and material support: Pinto, Martins, and Kowalski. Study supervision: Carvalho, Pinto, and Kowalski.

Financial Disclosure: None reported.

Previous Presentation: This study was presented in part at the annual meeting of the American Head and Neck Society; April 29, 2007; San Diego, California.