Treatment Factors Associated With Survival in Early-Stage Oral Cavity Cancer: Analysis of 6830 Cases From the National Cancer Data Base

Importance  Most patients with oral cavity squamous cell cancer (OCSCC) are initially seen at an early stage (I and II). Although patient and tumor prognostic features have been analyzed extensively, population-level data examining how variations in treatment factors impact survival are lacking to date.

Objective  To analyze associations between treatment variables and survival in stages I and II oral cavity squamous cell carcinoma.

Design, Setting, and Participants  Retrospective study of cases in the National Cancer Data Base. Patients diagnosed as having stage I or II OCSCC between January 1, 2003, and December 31, 2006, and treated with surgery were identified. Univariate and multivariable analyses of overall survival based on patient, disease, and treatment characteristics were conducted.

Main Outcomes and Measures  Overall survival and survival at 5 years.

Results  In total, 6830 patients were included. Survival at 5 years was 69.7% (4760 patients). On univariate analysis, treatment factors associated with improved survival included treatment at academic or research institutions, no radiation therapy, no chemotherapy, and negative margins (P < .001 for all). Neck dissection was associated with improved survival (P = .001), reflecting pathologic restaging and elimination of patients with occult nodal disease. Patients treated at academic or research institutions were more likely to receive neck dissection and less likely to receive radiation therapy or have positive margins. On multivariable analysis, neck dissection (hazard ratio [HR], 0.85; 95% CI, 0.76-0.94; P = .003) and treatment at academic or research institutions (HR, 0.88; 95% CI, 1.01-1.26; P = .03) were associated with improved survival, whereas positive margins (HR, 1.27; 95% CI, 1.08-1.49; P = .005), insurance through Medicare (HR, 1.45; 95% CI, 1.25-1.69; P < .001) or Medicaid (HR, 1.96; 95% CI, 1.60-2.39; P < .001), and adjuvant radiation therapy (HR, 1.31; 95% CI, 1.16-1.49; P < .001) or adjuvant chemotherapy (HR, 1.34; 95% CI, 1.03-1.75; P = .03) were associated with compromised survival.

Conclusions and Relevance  Prognostic impacts of treatment factors in early OCSCC are presented. Overall survival for early OCSCC varies with demographic and tumor characteristics but also varies with treatment and system factors, which may represent targets for improving outcomes in this disease.

Oral cavity squamous cell cancer (OCSCC) is a significant health burden in the United States, with approximately 28 000 cases and 5500 deaths in 2014 predicted.¹ Prognosis in early OCSCC depends on many factors, including patient age, stage at diagnosis, and primary site of disease, with a mean 5-year relative survival of 80%.² Approximately 60% of patients with OCSCC are initially seen with early-stage disease (I or II).³

Treatment of early OCSCC has not changed substantially in several decades, and improvement in outcomes has been slow.² Furthermore, the optimal treatment of patients with early-stage OCSCC remains subject to debate. Most of these patients are treated by peroral tumor resection with or without elective neck dissection (END).⁴Quiz Ref ID Up to 40% of patients with clinically N0 disease have subclinical nodal neck metastases,^5,6 but multiple reports have demonstrated no survival benefit from END in these patients.⁴ As a result, the role of neck dissection in early OCSCC remains a subject of controversy.⁷Quiz Ref ID Adjuvant radiation therapy and chemoradiotherapy are indicated in early OCSCC for certain high-risk pathologic features; however, these recommendations have not been validated by comprehensive therapeutic trials for early-stage OCSCC, and any survival benefit of these interventions is unknown.^4,7,8 The impact of health care system factors, including type of treating facility, case volume, and insurance status, on survival in early OCSCC is also unknown and may illustrate differences in quality of care. Finally, although the goal of surgery in early OCSCC is eradication of tumor, positive tumor margins have been reported in up to 21% of oral cancer resections.^9-11 Data on the association between positive margins and OCSCC outcomes have been inconsistent.^12,13

Data from the Commission on Cancer’s National Cancer Data Base (NCDB) were examined to evaluate population-level associations between treatment factors and survival in early OCSCC. We hypothesized that survival is independently associated with variations in treatment factors. These associations may validate current therapy norms, highlight differences in quality of care, and direct future research.

The NCDB is a nationwide, hospital-based cancer registry sponsored by the American College of Surgeon’s Commission on Cancer and the American Cancer Society. It is the world’s largest oncology outcomes database, capturing approximately 70% of all cancer cases in the United States.¹⁴ Data reported to the NCDB are retrospective and compliant with the requirements of the 1996 Health Insurance Portability and Accountability Act.¹⁵ The Yale University Institutional Review Board determined that this study was exempt from review.

Patients diagnosed as having stage I or II OCSCC between January 1, 2003, and December 31, 2006, were identified using International Classification of Diseases for Oncology, Third Edition (ICD-O-3) topography codes 8052, 8070 through 8078, 8083, and 8084. Patients who did not undergo surgery, those with multiple malignant primary tumors, and those with undocumented pathologic stage were excluded from analysis. Overall survival and 5-year survival were investigated as outcomes.

Factors investigated for the association with overall survival included treatment factors (radiation therapy, chemotherapy, neck dissection, surgical margins, facility type, facility case volume, and insurance) and nontreatment factors (patient age, race, sex, and Charlson Comorbidity Index and tumor primary site, pathologic stage, and grade). Radiation therapy, chemotherapy, and neck dissection were reported as received or not received, and regional lymph node surgery as reported by the NCDB was considered to constitute neck dissection. Surgical margin status was divided into positive margins (reported as microscopic residual tumor, macroscopic residual tumor, or residual tumor not otherwise specified) and negative margins (no residual tumor). Quiz Ref IDResidual tumor was considered present by the NCDB if it was within 5 mm of the margin, and final margin status was reported in cases of reresection. Facility type was based on the Commission on Cancer’s accreditation criteria and was divided into academic or research programs and nonacademic programs, which included community cancer programs, comprehensive community cancer programs, and other programs.¹⁶ Facility case volume was calculated as the mean number of oral cancer cases reported by an institution to the NCDB per year from 1998 to 2011 and was divided into categories of 10 or fewer cases and more than 10 cases per year. Insurance was reported as private, Medicare, Medicaid, other government insurance (including active military personnel, TRICARE, and Veterans Affairs), and no insurance. Race was reported as white, African American or black, and other. Comorbid conditions were analyzed using the Charlson Comorbidity Index modified by Deyo et al¹⁷ and were divided into indexes of 0 and 1 or more. Tumor primary site was divided into categories of tongue, lip, floor of mouth, gum or hard palate, retromolar trigone, buccal mucosa, and other mouth (including unspecified or overlapping sites and tumors of the vestibule of mouth) based on ICD-O-3 topography codes. Tumors were staged according to the American Joint Committee on Cancer’s seventh edition of the Cancer Staging Manual guidelines for pathologic staging. Tumor grade was reported as low grade (well differentiated), intermediate grade (moderately differentiated), high grade (poorly differentiated or anaplastic), or unknown.

All statistical analyses were performed using a software package (SPSS for Windows, version 20; IBM). Pearson χ² tests were used to determine the association between categorical variables and overall survival. Overall survival and 5-year survival were calculated by the Kaplan-Meier method. Cases were entered by the reported date of their last contact or death in months from diagnosis. Significance in survival differences was determined by the log-rank test. Multivariable analysis was conducted by Cox logistic regression. Survival effects of covariates were reported as hazard ratios (HRs). Significance was set at P < .05.

Patient, disease, and treatment characteristics of 6830 cases of surgically treated early OCSCC are listed in Table 1. Age was normally distributed, with a mean of 61.7 years. In total, 61.5% of participants were male, and 90.4% self-identified as being of white race. Survival at 5 years was 69.7% (4760 patients). Univariate analysis of survival revealed that neck dissection, treatment at academic or research cancer centers, and treatment at high case volume centers were associated with improved survival, whereas positive margins, radiation therapy, and chemotherapy were associated with compromised survival (P < .001 for all) (Table 2). The difference in 5-year survival between patients who received vs did not receive neck dissections was greater in stage II disease (63.9% vs 49.1%, P < .001) than in stage 1 disease (78.3% vs 74.2%, P = .001). The difference in 5-year survival between patients who received vs did not receive radiation therapy was greater in stage I disease (63.5% vs 77.6%, P < .001) than in stage II disease (53.5% vs 61.2%, P = .002).

Quiz Ref IDMultivariable analysis revealed that positive margins (HR, 1.27; 95% CI, 1.08-1.49), radiation therapy (HR, 1.31; 95% CI, 1.16-1.49), and chemotherapy (HR, 1.34; 95% CI, 1.03-1.75) were associated with reduced survival, while neck dissection (HR, 0.85; 95% CI, 0.76-0.94) was associated with improved survival (Table 3). Treatment at nonacademic cancer centers (HR, 1.13; 95% CI, 1.01-1.26) and insurance through Medicaid (HR, 1.96; 95% CI, 1.60-2.39) and Medicare (HR, 1.45; 95% CI, 1.25-1.69) were also associated with compromised survival. Facility case volume was not independently associated with survival on multivariable analysis. Patients treated at academic or research cancer centers were more likely to receive neck dissection and were less likely to receive radiation therapy or have positive margins than those treated at nonacademic centers (P < .001 for all) (Table 4). Pathologic characteristics, including stage (P = .87) and grade (P = .18), did not vary between patients at academic and nonacademic cancer centers.

Analyzing differences in survival associated with treatment factors may reveal opportunities to improve outcomes through systems-based approaches. Five-year survival in the NCDB data was consistent with reports over the last several decades at approximately 70%^2,18,19 and was associated with multiple treatment and nontreatment factors. Our data revealed associations between survival and health care delivery factors, such as insurance and type of treating facility, suggesting potential differences in quality of care that may be viable targets for quality improvement efforts. For example, care at academic or research cancer centers was associated with improved survival compared with care at nonacademic centers, consistent with prior reports on head and neck and other cancer outcomes.^20,21 This may be because of increased health care provider expertise, as evidenced by the lower incidence of positive margins and greater oral cancer case volume among these academic centers, or because of increased surgical aggressiveness, as evidenced by the greater proportion of patients receiving neck dissection. However, high case volume was not associated with increased survival in the NCDB data, despite traditionally being viewed as a surrogate marker for high-quality surgical care.²² Patients with Medicare or Medicaid coverage experienced compromised outcomes compared with similar patients with private insurance. Previous reports have similarly shown that lack of insurance and federal insurance coverage are independently associated with compromised outcomes in head and neck cancers.²³ This troubling trend may reflect inconsistent treatment and follow-up because of tenuous access to health care or worse baseline health, which has been reported in these patients.²⁴ Identifying the underlying causes of survival differences associated with these health care delivery factors could enable improvement of outcomes through dissemination of the optimal care practices.

The association of positive margins with reduced survival is widely reported, but whether this is a direct result of residual disease or a consequence of other correlated clinicopathologic or surgical features is unclear.^10,13 Tremendous variation also exists in the reported strength of the association between positive margins and outcomes, with associated decreases in survival ranging from 0% to more than 80%.^9,25,26 As a consequence, there has been ongoing debate regarding the optimal level of aggressiveness of surgery for oral cancer, with some reports recommending surgical de-escalation.²⁷ Our population-level analysis provides compelling evidence about the importance of margin status in oral cancer and confirms the association between positive margins and poor outcomes. These findings support use of aggressive surgery to achieve negative margins in early OCSCC. Furthermore, differences in margin positivity among treatment institutions suggest that more complete resection to obtain negative margins is possible in some patients.

Quiz Ref IDThe National Comprehensive Cancer Network currently recommends selective neck dissection of levels I through III (at a minimum) for tumors of at least 4-mm depth and at the discretion of the surgeon for tumors of at least 2-mm depth.⁸ However, these guidelines are based on consensus, with no supporting high-level evidence,²⁸ and the optimal treatment of a clinically N0 neck in early OCSCC with no adverse features remains a subject of controversy.⁷ The present analysis revealed the association between neck dissection and increased survival. This trend is likely owing to the fact that patients with occult neck disease who received a neck dissection are more likely to have had their cancer pathologically restaged to stage III or higher and consequently removed from our sample than patients who did not receive a neck dissection. This is supported by the greater difference in survival among patients with stage II disease compared with those with stage I disease, reflecting a greater prevalence of occult neck disease in this group. Although prospective data are lacking, a 2011 meta-analysis²⁹ of limited existing data reported improved survival among patients with stage I or II oral cancer and no additional high-risk features treated with END compared with observation of the neck. Further study with prospective trials is necessary to elucidate the role of END in early OCSCC.

Conversely, radiation therapy and chemotherapy were associated with decreased survival. Although positive margins are the most common indication for adjuvant chemoradiotherapy in early-stage OCSCC, other high-risk pathologic features (eg, lymphovascular or perineural invasion) also result in the recommendation for adjuvant therapies.⁸ We were unable to adjust for perineural and lymphovascular invasion in the present analysis because these characteristics were not reported in the NCDB. Although these factors are inconsistently reported and their prognostic impact is debatable, they could confound the effect of radiation therapy and chemotherapy on survival.^30-32 In addition, use of radiation therapy or chemotherapy could be a surrogate marker for less aggressive resection in very localized disease. This could explain our finding that radiation therapy was associated with a greater decrease in survival in stage I disease than in stage II disease.

The present study, to our knowledge, is the largest contemporary analysis to date of prognostic factors in early-stage OCSCC. However, several limitations should be considered. Owing to the retrospective nature of this study, the physician and patient decisions involved in surgery and adjuvant therapy are unknown. Significant proportions of the NCDB pool were excluded either because patients did not undergo surgery or tumors were not pathologically staged. Additional cases were excluded from multivariable analysis because data of 1 or more covariates were missing. The reasons for these missing data and whether their exclusion altered the sample pool are unknown. The list of variables included in our multivariable analysis was not exhaustive, and some potentially important factors (eg, patient smoking status, tumor thickness, and perineural invasion) were unavailable in the NCDB. Finally, we used overall oral cancer caseload to determine hospital case volume, which is an imperfect surrogate marker for surgeon case volume and surgical expertise.

This study demonstrates significant associations between treatment factors and overall survival in early OCSCC. Positive margins are independently associated with decreased survival and should continue to inform adjuvant therapy decisions. Patients who underwent END had greater survival compared with those who did not, reflecting restaging and elimination of patients with occult nodal disease and highlighting the need for prospective studies on END in early OCSCC. Positive margins, not performing END, and decreased overall survival were associated with treatment at a type of facility other than an academic or research center, suggesting that health care system variations affect these outcomes. Identification of the underlying causes of these differences could reveal valuable targets for improvement of outcomes in early OCSCC.

Submitted for Publication: December 18, 2014; final revision received March 18, 2015; accepted March 29, 2015.

Corresponding Author: Benjamin L. Judson, MD, Department of Surgery, Yale University School of Medicine, 333 Cedar St, PO Box 208041, New Haven, CT 06520 (benjamin.judson@yale.edu).

Published Online: May 14, 2015. doi:10.1001/jamaoto.2015.0719.

Author Contributions: Mr Luryi and Dr Judson had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Chen, Mehra, Roman, Sosa, Judson.

Acquisition, analysis, or interpretation of data: Luryi, Roman, Sosa, Judson.

Drafting of the manuscript: Luryi.

Critical revision of the manuscript for important intellectual content: Chen, Mehra, Roman, Sosa, Judson.

Statistical analysis: Luryi.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the William U. Gardner Memorial Research Fund at Yale University School of Medicine.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.