Five- and 10-Year Cause-Specific Survival Rates in Carcinoma of the Minor Salivary Gland

Importance  Previous studies of prognostic factors of carcinoma of the minor salivary gland (MSG) have been limited to single-institution studies and small case series. Thus, limited data are available to guide the head and neck oncologist in counseling patients on the prognosis and management of these malignant neoplasms.

Objective  To examine 5- and 10-year cause-specific survival (CSS) rates of MSG carcinomas across all histologic subtypes and head and neck tumor subsites.

Design, Setting, and Patients  Retrospective, population-based study using National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) data from January 1, 1988, through December 31, 2009. The study included 5334 patients diagnosed as having MSG carcinoma and registered in the SEER database. Patients without follow-up, diagnostic confirmation, and/or race designation were excluded from the analysis (131 [2.4%]). Final follow-up was completed on December 31, 2009, and data were analyzed from August 5, 2013, to July 1, 2014.

Main Outcomes and Measures  Five- and 10-year CSS rates for US patients with MSG carcinoma. Cox proportional hazard models were used to estimate adjusted hazard ratios (HRs) and 95% CIs.

Results  Among the 5334 patients with MSG carcinoma included, the most common histologic subtypes included mucoepidermoid carcinoma (1568 [29.4%]), adenoid cystic carcinoma (1228 [23.0%]), and adenocarcinoma (1313 [24.6%]). The most frequent sites of primary tumor were the oral cavity (3132 [58.7%]) and pharynx (1130 [21.2%]). Five-year CSS rate was significantly worse for MSG malignant neoplasms located in the larynx (HR, 2.42; 95% CI, 1.67-3.50) and nasal cavity and/or paranasal sinus (HR, 1.73; 95% CI, 1.29-2.32). Being older than 75 years was associated with a significantly worse 5-year CSS rate (HR, 2.88; 95% CI, 2.05-4.06). Compared with no surgery, local tumor destruction (HR, 0.44; 95% CI, 0.30-0.64), partial surgery (HR, 0.33; 95% CI, 0.23-0.47), and total surgery (HR, 0.55; 95% CI, 0.41-0.74) were each found to be a significant positive prognostic factor. No differences were observed in the 5-year hazard of death for race/ethnicity, sex, diagnosis year, or socioeconomic status, and 10-year adjusted HRs were similar to the 5-year patterns.

Conclusions and Relevance  This study, to date, represents the largest US survival analysis of carcinoma of the MSG. Prognosis is associated with histologic subtype, tumor subsite, age at diagnosis, grade, and surgical therapy.

Malignant neoplasms of the salivary gland account for 3% to 6% of all head and neck cancers,^1,2 and malignant neoplasms of the minor salivary gland (MSG) account for 10% to 20% of all salivary cancers.^2-6 Most of the 500 to 1000 minor salivary glands are located in the oral cavity and oropharynx.⁵ Investigators^3,4,7 have reported that as many as 80% of MSG tumors are malignant compared with their major salivary gland counterparts. These tumors can present themselves in a very elusive fashion because of the heterogeneity in subsite and histologic subtype. The standard therapy for these tumors is primary surgery with or without adjuvant radiotherapy, depending on factors such as stage, histologic grade, and margin status.^6-13

A large variety of MSG histologic subtypes exist, each with its own distinct clinical behavior. The diverse histologic subtypes and the rarity of these malignant neoplasms have made gathering sufficient prognostic data difficult. Some case series address prognosis; however, these are restricted to 1 to 2 subsites or histologic subtypes and/or have limited case numbers.^2,4,14,15 Survival data are also inconsistent in terms of reported survival end points, which makes comparison and interpretation difficult. Reported 5- and 10-year cause-specific survival (CSS) rates range from 69% to 87% and 49% to 80%, respectively.^14-17 In addition to tumor stage, data suggest that other factors such as age, histologic subtype, and tumor grade and subsite may have implications for CSS rates.^9,12-20 Definitive data to guide the head and neck oncologist in counseling patients about the prognosis of these malignant neoplasms are lacking. The primary objective of the present study was to examine 5- and 10-year CSS rates for MSG carcinoma across all histologic subtypes and head and neck tumor subsites in the United States using the Surveillance, Epidemiology, and End Results (SEER) database from January 1, 1988, to December 31, 2009. A secondary objective was to identify any differences in surgical treatment and radiotherapy trends during the study period.

We selected all cases of MSG malignant neoplasms diagnosed from January 1, 1988, to December 31, 2009, from the National Cancer Institute’s SEER cancer registry using codes from the International Classification of Diseases for Oncology, Third Edition (ICD-O-3)²¹ for the following malignant neoplasms of the salivary gland: mucoepidermoid carcinoma (MEC) (code 8430), adenoid cystic carcinoma (ACC) (code 8200), adenocarcinoma (codes 8140, 8147, 8290, 8310, 8410, 8440, 8480, 8525, and 8550), mixed subtype (codes 8980 and 8981), and other rare carcinomas (codes 8012, 8041, 8082, 8562, and 8982). The adenocarcinoma category included entities such as acinic cell carcinoma, polymorphous low-grade adenocarcinoma, adenocarcinoma not otherwise specified (NOS), and other rare adenocarcinomas. The SEER database collects and publishes cancer incidence and survival data from population-based cancer registries covering approximately 28% of the US population. The database contains information on patient demographics, tumor site, histologic subtype, date and source of diagnosis, date of death, and treatment.^22,23 Because SEER is a publicly available database, which is not considered human subjects research based on the Department of Health and Human Service’s Office of Human Research Protection, institutional review board approval was not required for the study.

Demographic, clinical, and pathologic characteristics of adult patients diagnosed as having MSG carcinoma from January 1, 1988, to December 31, 2009, were extracted from the SEER database. The total analytic cohort included 5334 patients. Patients with no follow-up (75 [1.4%]), missing diagnostic confirmation (11 [0.2%]), or lacking race/ethnicity designation (45 [0.8%]) were excluded from the analyses. Follow-up was completed on December 31, 2009.

Demographic characteristics included age (18-34, 35-44, 45-54, 55-64, 65-74, 75-84, or ≥85 years), sex (male or female), race (white, black, Hispanic, Asian or Pacific Islander, or other or unknown), and socioeconomic status measured by residence in a county with a high level of poverty (no, yes, or missing). Clinical characteristics included year of diagnosis (1988-1994, 1995-1999, 2000-2004, or 2005-2009), SEER stage (local, regional, distant, or missing), surgery (no surgery, total surgery, partial surgery, local tumor destruction or biopsy only, or missing), and radiotherapy (no, yes, or missing). The SEER Summary Staging Manual 2000 was used in all analyses.²⁴ We performed additional analyses to identify any differences in surgical treatment and radiotherapy trends during the study period. Total surgery refers to complete removal of the primary lesion (eg, total parotidectomy, radical parotidectomy, or total removal of salivary glands); partial surgery, partial removal of the primary lesion (eg, subtotal parotidectomy, removal of a superficial salivary gland lobe); and local tumor destruction, minimally ablative approaches using photodynamic therapy, cryosurgery, electrocautery, or laser therapy. These surgical therapies are defined according to the third edition of the SEER Program Code Manual.²⁵

Pathologic features included histologic subtype (MEC, ACC, adenocarcinoma, or other rare carcinomas), subsite (larynx, nasal cavity and/or paranasal sinus, oral cavity, or pharynx), and lymph node examination (LNE) (no, yes, or missing). Data for LNE were included in the database beginning in 1988. Low- and high-grade classification (low indicates well or moderately differentiated; high, poorly differentiated or undifferentiated; or missing) was only applied to adenocarcinoma NOS because this subtype could be graded accurately based on the differentiation scheme available in the SEER data. Grade was not applied to other subtypes of adenocarcinoma, MEC, or ACC because these subtypes have their own specific grading criteria unrelated to differentiation. Histologic diagnoses were based on the 2005 World Health Organization classification of malignant salivary gland tumors.²⁶

Data were analyzed from August 5, 2013, to July 1, 2014. Statistical analysis was performed using SAS software (version 9.2; SAS Institute Inc). We used 2-tailed t tests to evaluate associations between primary site and histologic subtype by sociodemographic characteristics. Statistical significance was defined as α < .05. For the survival analysis, days from diagnosis to death or the end of the study period were used to measure person-time. We used Kaplan-Meier survival curves to assess unadjusted 5- and 10-year CSS rates and log-rank statistics to assess differences in survival curves. We used multivariable Cox proportional hazards regression models to assess factors related to 5- and 10-year CSS rates. All variables were entered into a single model to test for violations in the proportional hazards assumption. Because histologic subtype, radiotherapy, and stage violated the proportional hazards assumption and were adjusted for in our multivariable model, we can make no conclusions regarding the association between these factors and survival. For 10-year survival analyses, grade also violated the proportional hazards assumption and was adjusted for in the STRATA statement.

Of the 5334 patients with MSG carcinoma included in the study, 2662 were men (49.9%) and 2672 were women (50.1%). Most of the patients were white (3426 [64.2%]) and aged 45 to 74 years (3267 [61.2%]). A nearly even distribution of cases from counties with high (2869 [53.8%]) and low (2464 [46.2%]) levels of poverty was found. When the tumor stage was stratified by subsite, the proportion of patients with regional disease was the highest for carcinomas of the nasal cavity and/or paranasal sinus (495 of 843 [58.7%]) and the pharynx (561 of 1130 [49.6%]). Tumors in the larynx (26 of 229 [11.4%]) and nasal cavity and/or paranasal sinus (111 of 843 [13.2%]) had the highest rate of distant metastases. In terms of histologic subtype, the 1228 patients with ACC had the highest frequency of regional (600 [48.9%]) and distant (156 [12.7%]) disease.

We performed analyses to identify any differences in surgical treatment and radiotherapy trends during the study period. Overall, most of the patients received a form of surgical therapy (3604 [67.6%]). By specifically focusing on surgical trends over time, we found a significant difference between 2005 to 2009 and 1988 to 1994, with adjustment for other factors (odds ratio, 1.35; 95% CI, 1.11-1.67). Therefore, we found more frequent surgical intervention at the end of the study period compared with the initial years. We found no differences in surgical treatment between the other periods and 1988 to 1994. In terms of radiotherapy, we identified no significant differences among the 5-year periods. Overall, patients with ACC were more likely to receive postoperative radiotherapy (816 of 1228 [66.4%]) than were patients with other subtypes.

In terms of pathologic characteristics, MEC (1568 [29.4%]), ACC (1228 [23.0%]), and adenocarcinoma (1313 [24.6%]) were the most common histologic subtypes. The number of cases of adenocarcinoma NOS totaled 682 (12.8%). Low- and high-grade adenocarcinoma NOS totaled 493 (9.2%) and 189 (3.5%) cases, respectively. The most frequent sites of primary tumor were the oral cavity (3132 [58.7%]) and pharynx (1130 [21.2%]). Lymph node examination was not performed in most of the cases (4184 [78.4%]). These data are stratified by histologic subtype in the Table.

Patients with a primary tumor of the oral cavity had the highest 5- and 10-year unadjusted CSS rates at 90.1% and 84.7%, respectively. Those with a primary site in the larynx had the lowest 5- and 10-year CSSs, at 52.6% and 45.3%, respectively. Patients with MEC had the highest unadjusted 5-year CSS (90.7%) compared with patients with ACC (79.1%), adenocarcinoma (80.8%), and other rare carcinomas (70.1%). Ten-year CSS patterns were similar to those of 5-year CSS for patients with MEC (88.6%), adenocarcinoma (75.5%), and other rare carcinomas (62.1%); however, the 10-year CSS for patients with ACC was much lower (62.4%) (Figure 1). In terms of SEER stage, patients with isolated local disease had a 5-year CSS of 95.3% and a 10-year CSS of 93.0%. Five- and 10- year CSSs for patients with regional disease were 74.9% and 63.4%, respectively; for those with distant disease, 44.8% and 28.8%, respectively. Patients with low- and high-grade adenocarcinoma NOS had 5-year CSSs of 87.1% and 47.1%, respectively, and 10-year CSSs of 83.8% and 35.3%, respectively. Five-year CSS among patients who received radiotherapy was 72.9% compared with 89.1% among patients who did not receive radiotherapy. Five-year unadjusted CSSs for patients who received surgery vs surgery with radiotherapy were 93.5% and 79.4%, respectively (Figure 2).

We used multivariable Cox proportional hazards regression models stratified by histologic subtype, radiotherapy, and stage because these variables did not meet the proportional hazards assumption. For 10-year survival analyses, grade also violated the proportional hazard assumption and was adjusted for in the model. Referent categories included oral cavity, white race, male sex, being younger than 45 years, not undergoing surgery, no LNE, low-grade adenocarcinoma NOS, low poverty areas, and a diagnosis year from 1988 to 1994.

The 5-year CSS stratified by histologic subtype was poorest in patients with MEC located in the larynx (hazard ratio [HR], 5.09; 95% CI, 2.13-12.16), followed by tumors in the nasal cavity and/or paranasal sinus (HR, 2.84; 95% CI, 1.59-5.08). Patients 75 years or older also had decreased CSS (HR, 2.98; 95% CI, 1.44-6.18). Compared with patients with MEC who did not undergo surgery, improved CSS was found in patients with MEC who underwent local tumor destruction (HR, 0.20; 95% CI, 0.09-0.47), partial surgery (HR, 0.15; 95% CI, 0.07-0.33), and total surgery (HR, 0.28; 95% CI, 0.15-0.52).

The 5-year CSS for patients with ACC stratified by site was worse in patients with ACC located in the pharynx (HR, 2.08; 95% CI, 1.20-3.61), followed by those with tumors located in the nasal cavity and/or paranasal sinus (HR, 1.85; 1.29-2.65). Patients 75 years or older also had decreased CSS (HR, 2.08; 95% CI, 1.28-3.38). Compared with patients with ACC who did not undergo surgery, improved CSS was found in patients with ACC who underwent local tumor destruction (HR, 0.26; 95% CI, 0.14-0.49), partial surgery (HR, 0.29; 95% CI, 0.17-0.47), and total surgery (HR, 0.38; 95% CI, 0.25-0.58).

The 5-year CSS for patients with adenocarcinoma stratified by site was worse for patients with tumors located in the nasal cavity and/or paranasal sinus (HR, 1.60; 95% CI, 1.03-2.49). Patients 75 years or older also had significantly decreased CSS (HR, 5.07; 95% CI, 2.24-11.45). Compared with patients with adenocarcinoma who did not undergo surgery, improved CSS was found in patients with adenocarcinoma who underwent local tumor destruction (HR, 0.18; 95% CI, 0.08-0.41), partial surgery (HR, 0.19; 95% CI, 0.10-0.35), and total surgery (HR, 0.33; 95% CI, 0.20-0.54).

The 5-year CSS for patients with other rare carcinomas stratified by site was worse for those with tumors located in the larynx (HR, 4.32; 95% CI, 2.53-7.39), followed by those with tumors located in the nasal cavity and/or paranasal sinus (HR, 2.14; 95% CI, 1.11-4.12). Patients 75 years or older also had decreased CSS (HR, 2.15; 95% CI, 1.23-3.75). Compared with patients with other rare carcinomas who did not undergo surgery, improved CSS was found in patients undergoing local tumor destruction (HR, 0.80; 95% CI, 0.52-1.22), partial surgery (HR, 0.56; 95% CI, 0.30-1.02), and total surgery (HR, 0.70; 95% CI, 0.40-1.21).

When we performed multivariable CSS analysis without histologic stratification and included all histologic subtypes and subsites, additional interesting findings were generated. A significantly worse 5-year CSS was portended by malignant neoplasms of the MSG located in the larynx (HR, 2.42; 95% CI, 1.67-3.50) and nasal cavity and/or paranasal sinus (HR, 1.73; 95% CI, 1.29-2.32). Being older than 75 years was associated with significantly worse 5-year CSS (HR, 2.88; 95% CI, 2.05-4.06). Patients aged 65 to 74 years had worse CSS of marginal significance (HR, 1.50; 95% CI, 1.08-2.08). Lymph node examination was also found to be associated with poor CSS (HR, 1.30; 95% CI, 1.05-1.59). Similar to models stratified by histologic subtype, patients who underwent no surgery again had a significantly worse prognosis compared with those who underwent local tumor destruction (HR, 0.44; 95% CI, 0.30-0.64), partial surgery (HR, 0.33; 95% CI, 0.23-0.47), and total surgery (HR, 0.55; 95% CI, 0.41-0.74). We observed no differences in the 5-year hazard of death for race/ethnicity, sex, diagnosis year, or socioeconomic status using the model stratified by histologic subtype or the unstratified model. Patterns for adjusted 10-year hazard of death on multivariable analysis were similar to those observed for 5-year data.

To our knowledge, the present study represents the largest reported survival analysis for MSG carcinoma in the United States. The existing literature on survival of patients with MSG carcinoma includes smaller case-series and/or is restricted to 1 to 2 subsites or histologic subtypes.^2,4,14,15 Our study included 5334 patients with MSG malignant neoplasms of all histologic subtypes collected for 2 decades. Malignant neoplasms located in the larynx and nasal cavity and/or paranasal sinus were associated with worse survival. Being older than 75 years was associated with decreased CSS across all histologic subtypes. High grade was found to be associated with worse CSS for adenocarcinoma NOS. Lymph node examination was also identified as a poor prognostic factor, which will be discussed in further detail below. Surgical therapy was associated with increased survival for all types of MSG carcinoma. Nevertheless, because comorbidity is not addressed in the SEER database, healthier patients may have been more likely to undergo surgery and therefore had improved outcomes. Race/ethnicity, sex, diagnosis year, and socioeconomic status were not found to be of prognostic significance when accounting for other factors.

Numerous studies have found stage and tumor grade to be consistently associated with survival in MSG carcinoma.^9,12,14,15 However, the data regarding the prognostic value of age, histologic subtype, and tumor subsite are equivocal. In a 1991 single-institution study by Spiro et al,⁹ stage and histologic grade were found to be the most important independent prognostic factors for CSS in 378 patients with MSG malignant neoplasms. Histologic subtype, subsite, and age were not found to be of prognostic value on multivariable analysis. Similarly, Anderson et al¹² analyzed 95 patients and demonstrated that age and tumor subsite were not significant prognostic factors. Vander Poorten et al¹⁴ described 55 patients from the Netherlands Cancer Institute from 1973 to 1994 and found stage and subsite but not age to be significant prognostic factors. Loh et al¹⁶ also studied 171 patients and found only histologic grade to be associated with survival on multivariable analysis, but not age, histologic subtype, or subsite. We observed that age and tumor subsite are independent prognostic survival factors.

We did not anticipate the survival results generated based on LNE. On multivariable analysis, LNE was associated with worse CSS at 5 and 10 years. Lymph node examination was associated with a higher likelihood of regional and distant disease than no LNE. In fact, 477 of 811 patients who underwent LNE had regional disease (58.8%) compared with 1452 of 4184 patients who did not undergo LNE (34.7%). We hypothesized that this association could be owing to residual confounding by stage because we were only able to include broad stage grouping.

In terms of survival, a site in the nasal cavity and/or paranasal sinus carried a worse CSS across all histologic subtypes, which is consistent with findings in the existing literature.^16,27,28 Survival data on MSG malignant neoplasms of the larynx are sparse owing to their rarity. Ganly et al²⁹ reported 12 cases of laryngeal MSG carcinoma, with ACC being the most common histologic subtype and as many as 60% of all cases developing local, regional, or distant recurrent disease. Herein, we report 229 cases of laryngeal MSG malignant neoplasms and found that tumors at this subsite were associated with significantly worse 5- and 10-year CSS compared with other subsites. The most common histologic subtypes encountered in the larynx in our study included 44.5% of other rare carcinomas (102 of 229) and 25.3% of adenocarcinoma (58 of 229). We suspect such poor survival is associated with advanced disease at presentation owing to the propensity for submucosal growth with minimal symptoms, recurrent disease, and distant metastases.

In this study, the point estimate for partial surgery (HR, 0.33; 95% CI, 0.23-0.47) was lower than that for total surgery (HR, 0.55; 95% CI, 0.41-0.74). However, because the 95% CIs overlap, this difference (partial vs total) is not statistically significant. Patients undergoing total and partial surgery had better survival compared with those undergoing no surgery. We adjusted for receipt of radiotherapy in our multivariable models; however, we did not have data related to the completion and dose of radiotherapy. This residual confounding could account for point estimate differences between partial and total surgery.

Several limitations to this study are worthy of mention, including those inherent to the use of any administrative database, such as errors in coding and sampling. In addition, no SEER data were found regarding patient comorbidities and intent of surgery. These data would be useful for determining candidacy for surgical intervention. The SEER database also does not contain information on important survival predictors that include health insurance, which may determine access to and use of health care. In terms of surgical therapy, 393 patients (7.4%) were missing data on treatment type. These missing data may have affected the trends we observed in surgical treatment across the study period and possibly the associated effect on survival. In addition, the SEER staging system is not as clinically applicable as the accepted American Joint Committee on Cancer staging system for salivary gland malignant neoplasms. Furthermore, the 1977 SEER staging system underwent a revision in 2000,²⁴ which could allow for differences in stage classification, depending on the year a case was entered in the database. Finally, the differentiation-based grading system limited our analysis to adenocarcinoma NOS because other types of adenocarcinoma, MEC, and ACC cannot be graded accurately based on differentiation alone.

One strength of the SEER database is the large sample size, which provides sufficient power to examine rare malignant neoplasms such as MSG carcinoma and permit multivariable analyses. In addition, the population-based rather than institution-based identification of patients in the SEER database lends external validity to SEER analyses. Institutional studies are usually confined to major referral centers and may not be representative of the patients with MSG cancers who are treated in community hospitals and clinics.

This study, to our knowledge, represents the largest US survival analyses among patients with MSG carcinoma. Prognosis is associated with histologic subtype, tumor subsite, age at diagnosis, tumor grade, and surgical therapy.

Corresponding Author: H. Michael Baddour Jr, MD, Department of Otolaryngology–Head and Neck Surgery, Emory University School of Medicine, Medical Office Tower, Ste 1135, 550 Peachtree St NE, Atlanta, GA 30308 (hbaddou@emory.edu).

Submitted for Publication: June 2, 2015; final revision received September 28, 2015; accepted October 2, 2015.

Published Online: December 3, 2015. doi:10.1001/jamaoto.2015.2805.

Author Contributions: Ms Fedewa and Dr Chen 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: All authors.

Acquisition, analysis, or interpretation of data: Baddour, Fedewa.

Drafting of the manuscript: Baddour, Fedewa.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Fedewa.

Study supervision: Chen.

Conflict of Interest Disclosures: None reported.

Previous Presentation: This paper was presented at the Fifth World Congress of the International Federation of Head and Neck Oncologic Societies and Annual Meeting of the American Head and Neck Society; July 27, 2014; New York, New York.