Figure 1. Patient selection algorithm.
Figure 2. Kaplan-Meier survival curves: product-limit survival estimates with number of subjects at risk. Disease-specific survival (A) and overall survival (B) of patients undergoing pre-radiotherapy tonsillectomy vs biopsy only, as estimated by means of the Kaplan-Meier method.
Holliday MA, Tavaluc R, Zhuang T, Wang H, Davidson B. Oncologic Benefit of Tonsillectomy in Stage I and II Tonsil CancerA Surveillance Epidemiology and End Results Database Review. JAMA Otolaryngol Head Neck Surg. 2013;139(4):362-366. doi:10.1001/jamaoto.2013.107
Author Affiliations: Department of Otolaryngology–Head and Neck Surgery, Georgetown University Hospital, Washington, DC (Drs Holliday and Davidson); and Department of Biostatistics, Georgetown University, Washington, DC (Ms Zhuang and Dr Wang). Ms Tavaluc is a student at Georgetown University School of Medicine.
Importance Treatment of low-stage tonsil cancer with radiotherapy is common, but the type of diagnostic procedure prior to radiotherapy varies. This study uses the Surveillance Epidemiology and End Results (SEER) registry to determine whether tonsillectomy or other surgical procedures prior to radiotherapy have an impact on outcome.
Objective To demonstrate whether tonsillectomy adds oncologic advantage over biopsy alone in stage I or II tonsil cancers prior to definitive radiotherapy.
Design Retrospective study of the SEER program.
Participants A total of 524 patients with stage I and II primary tonsil carcinoma diagnosed during the period 1988 through 2006 who received definitive radiation treatment.
Main Outcomes and Measures Hazard ratios, 5-year disease-specific survival (DSS), 5-year overall survival (OS), and Kaplan-Meier survival curves. Subgroup univariate and multivariate analysis of survival compared the following variables: sex, T category (T1 vs T2), year of diagnosis (1988-2003 vs 2004-2006), and surgical status (preradiation tonsillectomy vs biopsy only). A log-rank test compared Kaplan-Meier survival analysis for the surgical status subgroups.
Results Statistically significant contributors to survival (DSS and OS) included age (P < .001, DSS), surgical status (P = .001, DSS), and year of diagnosis (P = .02, DSS). Treatment with radiation after tonsillectomy yielded a 5-year OS of 83% (95% CI, 77%-88%) and a 5-year DSS of 90% (95% CI, 84%-93%). This compares with an OS of 64% (95% CI, 54%-71%) and DSS of 76% (95% CI, 68%-82%) for radiation therapy after biopsy alone.
Conclusions and Relevance SEER data suggest that tonsil resection prior to radiation therapy is associated with improved survival in low-stage tonsil cancer. Although selection bias is a likely factor in this analysis, the effect remains after controlling for age and year of treatment.
Squamous cell carcinoma of the tonsil is the most common cancer of the oropharynx, with increasing incidence during the past 30 years.1 Early-stage tumors have traditionally been treated with either surgery or radiotherapy (RT) after several retrospective studies showed equivalent oncologic outcomes with either modality.2- 4 Current National Comprehensive Cancer Network guidelines for low-stage tonsil cancers recommend either RT or surgical resection with or without elective neck dissection.5
During the workup of tonsil cancer, otolaryngologists generally perform either incisional biopsy or diagnostic tonsillectomy, and many of these patients go on to receive RT. While diagnostic tonsillectomy prevents false-negative results by allowing for the assessment of all tonsillar tissue, the procedure adds increased morbidity in the form of pain, bleeding risk, and delay of definitive RT to allow for healing of the wound bed. In addition, diagnostic tonsillectomy rarely leads to negative oncologic margins. Radical tonsillectomy, in contrast, can lead to negative margins at resection but is not the chosen approach for a diagnostic procedure. The present study was designed to determine whether tonsillectomy—as a diagnostic or therapeutic procedure—adds oncologic advantage over biopsy alone in low-stage tonsil cancers when definitive RT is used.
We performed a review of the 17 registries of the National Cancer Institute's Surveillance Epidemiology and End Results (SEER) program, version SEER*Stat 6.5.2, from 1973 through 2006.6 Because this study used a database available in the public domain, it was exempted from full review by the Georgetown University Hospital institutional review board. We limited our search to those registrants with tonsillar cancer and included only those with a single primary tumor. Tumor histologic subtype was limited to squamous or epithelial cell malignancy. We selected only patients with T1 and T2 tumors and N0 neck disease, according to the American Joint Committee on Cancer's (AJCC’s) 6th edition TNM staging guidelines,7 and those who had completed definitive external beam RT (Figure 1).
Of note, the database has used the AJCC TNM staging criteria since 2004. From 1988 through 2004, the SEER summary staging system was used. For this era, we selected only those patients with localized disease for which the size of the tumor was given in millimeters. We then calculated the T category on the basis of the size of the tumor using current AJCC staging guidelines, with tumors up to 2 cm in greatest dimension classified as T1 and tumors greater than 2 cm and up to 4 cm in greatest dimension classified as T2.7 The staging system used prior to 1988 was incompatible with the current staging system, so these cases were excluded. We excluded patients who were registered in the database after 2006 to allow for adequate oncologic follow-up. We included only those patients who had negative results for neck disease. Also included in the analyses was a small group of patients who underwent neck dissections whose results were negative for disease.
Our null hypothesis states that tonsillectomy will not change survival for patients who eventually receive definitive RT. Our alternative hypothesis states that tonsillectomy or another procedure involving excision of the primary tumor will indeed affect survival in this selected group of patients. The procedures included, based on the SEER coding parameters, are listed in Table 1. Those patients who underwent incisional, needle, or aspiration biopsy of the primary site were placed in the control group. Those patients who underwent diagnostic tonsillectomy or more extensive procedures were placed in the surgery group. Other data elements collected and analyzed were age, sex, year of diagnosis, and tumor size (T1 vs T2). The year of diagnosis subgroups were divided between 2003 and 2004 to obtain approximately balanced sample sizes, to align with the more recent use of the AJCC TNM staging criteria, and to satisfy Cox proportional hazards assumptions.
The disease-specific survival (DSS) and overall survival (OS) probabilities of tonsil cancer were estimated by means of the Kaplan-Meier method for the surgical and nonsurgical groups. (Local-regional recurrence information is not available from the SEER database.) The OS time was measured from the date of registration to the date of death or last contact. The log-rank test was used to compare the survival distributions among surgical subgroups. Hazard ratios (HRs) and the corresponding 95% confidence intervals are estimated by means of the Cox proportional hazards model for each of the subgroups, in both univariate and multivariate fashion. The SAS 9.2 software package (SAS Institute Inc) was used for the statistical analysis. P < .05 was considered statistically significant.
We identified a total of 524 patients with T1 or T2 tonsillar squamous cell or epithelial cell tumors to analyze. Subgroup distribution is listed in Table 2. T2 cancers (322 [61%]) were more common than T1 cancers. The year of diagnosis subgroups and the surgical vs nonsurgical groups were approximately equally divided. Of the 283 patients who underwent surgical resection of the primary tumor, 249 (88%) underwent procedures listed as local excision /destruction; excisional biopsy; limited pharyngectomy; or pharyngectomy, NOS , according to SEER coding parameters. Few patients had extensive procedures such as more complete pharyngectomies and mandibulectomies (Table 1).
Univariate analysis revealed that older age (P < .001), surgery status (P < .001), and year of diagnosis (P = .03) significantly affected DSS and OS (Table 3). Multivariate analysis showed that age (P < .001), surgical status (P = .001), and year of diagnosis (P = .02, DSS; P = .03, OS) retained significance after controlling for each individual parameter. We found that the nonsurgical group was 2.5 times more likely to die of tonsil cancer over 5 years than was the surgical group. Analysis of the year of diagnosis favored those patients registered between 2004 and 2006 compared with those patients registered between 1988 and 2003 (HR = 2.55, DSS; HR = 2.03, OS; Table 3). Increasing age decreased DSS for each subsequent year of life (HR = 1.04; Table 3).
Kaplan-Meier survival curves comparing the surgical and nonsurgical groups are shown in Figure 2. The 5-year DSS was 76% in the control group and 90% in the surgical group (Table 4).
This analysis of low-stage tonsil cancers included in the SEER registry indicates that patients who underwent tonsillectomy prior to RT had better survival than those who underwent RT after biopsy alone. In addition, there was an apparent advantage for patients treated more recently (2004-2006 vs 1988-2003). Younger age was also associated with survival benefit. Each of these parameters significantly affected OS and DSS, and significance was maintained in multivariate analysis.
Yildirim et al8 recently investigated the outcomes of patients with tonsil cancer of all stages who underwent tonsillectomy followed by definitive RT during the period 1979 through 2004. Of 120 total patients in their review, only 20 had stage I or II disease. The 5-year actuarial survival rates for stage I and II tumors were 100% and 77%, respectively. Most patients (83%) had stage III or IV disease, which highlights the rarity of patients presenting with T1 and T2 tumors in the absence of nodal disease. Similarly, our inclusion criteria of localized, early-stage tumors drastically reduced our sample size compared with the overall SEER population of patients with tonsil cancers (Figure 1). Only 51% of patients in the study by Yildirim et al8 had confirmed negative surgical margins, and nearly all patients were treated with a definitive RT dose. Our study differs from the study by Yildirim et al8 in that we focused exclusively on stage I and II cancers treated with RT and compared those patients who underwent diagnostic tonsillectomy or wider resection prior to RT with those patients who underwent only biopsy prior to RT. Our 5-year OS of 83.2% for patients with stage I and II tumors who underwent surgery followed by RT is consistent with the above data for this subgroup.
The SEER database lacks information on margin status and on local recurrence. We compared tonsillectomy with biopsy alone but do not have data regarding margin status at surgery. Oncologic resection may range from radical tonsillectomy to composite resection (transoral robotic surgery was introduced after the period of study in this analysis). The goal in these procedures is generally to obtain negative margins. However, our analysis considers all tonsillectomies, including diagnostic tonsillectomy, and is likely to include a large number of patients with positive margins at surgery. The ability to discern whether margin status determines outcome in this group of irradiated tonsil cancers is beyond the capability of the database. Even if all of the procedures performed are essentially viewed as surgical “debulking” with potentially positive margins, the observed survival benefit remains.
Criteria that led to patient selection for tonsillectomy over biopsy alone are not provided in the database. One element that would be particularly useful is a record of comorbidities that might have played a role in selection for surgical resection. Comorbidities may indicate a surgical bias that may have played a role in the improved survival of those who underwent tonsillectomy. The healthier patient who was selected to receive surgery may be more likely to be alive after 5 years, regardless of intervention, simply as a result of better overall health. The database does provide patient age, however, and we used this as a proxy for overall health status, albeit an imperfect one. After controlling for age, the survival benefit of tonsillectomy prior to RT remains.
There are also unknown tumor parameters in this group that may have led to selection for surgical resection over biopsy alone. In fact, the only tumor characteristic noted in the database was size, from which T category was derived for our analysis. The fact that extensive resections such as total pharyngectomies, mandibulectomies, and laryngectomies were performed suggests that the tumors may have had characteristics that would require upstaging, such as extension to the larynx or invasion of the mandible, despite being smaller than 4 cm in greatest diameter. We were concerned that censoring these more invasive tumors would bias the surgical group in favor of the less invasive tumors, so they were left in the analysis.
A recent study by Miller et al9 examined all tonsillar cancers entered in the SEER database during the period 1998 through 2006 specifically to compare cancer survival over time. They found that DSS has increased for tonsillar cancers in recent years. They attributed this survival benefit to earlier diagnosis, an increase in the incidence of less aggressive squamous cell cancer subtypes related to an association with human papilloma virus (HPV) infection, and more effective treatment modalities in more recent years. The latter may be attributed to more advanced RT technology, including intensity-modulated RT, and more refined RT protocols used in the more recent group. Our analysis focused on only early-stage tonsil cancers but confirms their findings over a longer time frame.
In addition, the increasing incidence of tonsillar cancers in the United States and abroad has also been reported and has been attributed to increasing HPV prevalence in the oropharynx.10- 13 Although our data set shows an increase in the number of cases during recent years, it is beyond the scope of this study to analyze tonsillar cancer incidence over time. In fact, the large number of recent cases in our data set likely reflects an increased number of cancer registries reporting to SEER in recent years. Furthermore, the number of cases in the earlier years was limited by strict inclusion criteria that used size to stage the tumor.
In another study of SEER data from 1974 through 2003, Nguyen et al14 found that young age is associated with an excellent prognosis in tonsil carcinoma. The patients in their study likely overlapped with those in our sample, but whereas our analysis treated age as a continuous variable, their analysis grouped patients into age groups of older than 50 years and 50 years or younger. They found a 40% reduction in risk of cancer death in the younger group compared with the older group. They believe that this improved survival is a result of a higher incidence of HPV-associated tumors in younger patients and that young age is essentially a proxy for HPV positivity. In addition, the possibility that younger patients may have better overall health statuses was considered. We similarly speculate that younger patients in our study had better outcomes because of either HPV positivity, better overall health status, or both. In fact, whereas Nguyen et al14 used age as a proxy for HPV status, our analysis used age as a proxy for overall health status.
In conclusion, whether to treat patients with early-stage tonsil cancer using surgery or RT remains controversial, and with the advent of new procedures to address the primary tumor, the debate is far from settled. This study provides, to our knowledge, the largest sample and the only direct comparison of pre-RT excisional surgery of the primary site vs simple biopsy prior to RT. Our analysis revealed that RT after tonsillectomy is associated with better OS and DSS than is RT after biopsy alone. Opportunities for future study would include assessment of local-regional control, surgical selection bias, and the impact of comorbidities in this clinical question. Ultimately, a prospective, controlled study would be required to determine causation for this particular clinical question.
Correspondence: Michael A. Holliday, MD, Georgetown University Hospital, Department of Otolaryngology–Head and Neck Surgery, 1 Gorman, 3800 Reservoir Rd NW, Washington, DC 20016 (firstname.lastname@example.org).
Submitted for Publication: October 12, 2012; final revision received November 23, 2012; accepted December 19, 2012.
Author Contributions: Dr Davidson 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: Davidson. Acquisition of data: Holliday, Tavaluc, and Davidson. Analysis and interpretation of data: Holliday, Zhuang, Wang, and Davidson. Drafting of the manuscript: Holliday, Tavaluc, Zhuang, Wang, and Davidson. Critical revision of the manuscript for important intellectual content: Holliday, Zhuang, Wang, and Davidson. Statistical analysis: Zhuang and Wang. Administrative, technical, and material support: Tavaluc and Davidson. Study supervision: Davidson.
Conflict of Interest Disclosures: None reported.
Previous Presentation: This study was presented at the American Head and Neck Society Eighth International Conference on Head and Neck Cancer; July 24, 2012; Toronto, Ontario, Canada.