2018-19 Section 7-Neoplastic and Inflammatory Diseases of the Head and Neck eBook

Research Original Investigation

Epidemiology and Treatment of Malignant Submandibular Gland Tumors

M alignant primary tumors of the salivary glands ac- count for less than 5% of all cancers of the head and neck. 1 Themost common tumor location in the sali- vary glands is in the parotid gland, whereas tumors in the sub- mandibular gland are less common; only 5% to 15%of all sali- vary gland neoplasms occur in the submandibular gland. 2-7 Submandibular gland tumors have a frequency of malig- nancy of roughly 43%, almost double that of the parotid gland. 8,9 In general, tumors of the submandibular gland are associated with higher rates of malignancy and less favorable prognoses when compared with other primary tumors of the salivary glands. 2,5,10-12 Because of the relatively low incidence of submandibu- lar gland tumors, it has been difficult to evaluate treatment modalities, and limited information is available in the litera- ture regarding survival and prognostic factors. In almost all cases of primary tumors of all salivary glands, surgery is the primary treatment modality. 6 Patients with submandibular gland carcinoma are more likely to undergo radical resection of the affected gland in comparison with patients with other forms of salivary gland cancers; such procedures are more extensive and include the removal of the adjacent muscles and nerves. 6,13-15 However, the role of radiation therapy to provide enhanced locoregional control for the treatment of submandibular gland malignant neoplasms is still unclear. 1,2,7,13,15-17 This study focused on determining the incidence and sur- vival of, to our knowledge, the largest population of patients to date with a diagnosis of primary malignant tumors of the submandibular gland. Data from2626 patients with a diagno- sis of submandibular gland cancer from 1973 through 2011 in the US National Cancer Institute’s Surveillance, Epidemiol- ogy, and End Results (SEER) cancer registrywere used to ana- lyze several patient and disease characteristics to determine possible factors affectingbothoverall survival (OS) anddisease- specific survival (DSS). Much of the existing literature con- sists of small case series and reports from single- and multi- center studies, with limited large-population data. 4,14,15,18-24 A previous SEER database analysis on submandibular gland cancer was performed in 2004 on data from370 patients who received a diagnosis from 1988 through 1998. 2 Our study ex- pands on the previous SEER database analysis through in- creasing the sample size, analyzing DSS, and performing sta- tistical analysis on patient populations stratified by histologic subtype, stage, and tumor size. Use of this database for clini- cal outcomes research has been validated in previous studies investigating several types of cancers of theheadandneck. 25-30

Study Participants A total of 2626 patientswho received a diagnosis between 1973 and 2011 of a primary epithelial malignant tumor of the sub- mandibular gland, confirmed by histological diagnosis, were identified within the SEER database by using the primary site cancer label C8.0 (submandibular gland). Patient data within the SEER database included demographic variables for age at diagnosis, sex, and race. Pathologic variables for this study in- cluded tumor histologic subtype ( International Classification of Diseases,NinthRevision,ClinicalModification code),tumorgrade (categorized into 2 groups: low grade and high grade), tumor extent, lymph node involvement, tumor size (from both ex- tent of disease and collaborative stage coding methods, di- vided into 2 groups: tumors ≤3 and >3 cm), and tumor stage at presentation(AmericanJointCommitteeonCancer[AJCC]).The clinical variables used in our analysis were treatment with ra- diation (yes or no), treatment with surgery (yes or no), OS in months, andDSS inmonths. For thepurposesof this study,well- differentiated and moderately differentiated histological sub- types were grouped together and classified as low grade, and poorly differentiated and undifferentiated histological sub- types were grouped together and classified as high grade. The AJCC stages are recorded in the SEERdatabase for patientswho receivedadiagnosisafter2003;forcasesdiagnosedbefore2003, AJCCstageswere retroactivelydeterminedwherepossibleusing extent of disease and collaborative stage staging codes for tu- mor size, extent, and lymph node involvement according to AJCC protocol. Statistical Analysis Patient outcomes were measured on the basis of OS, the time in months between diagnosis and death from any cause, and DSS, the time in months between diagnosis and death di- rectly caused by the primary tumor as reported in the SEER database. Median survival time was defined as the smallest length of time, in years, inwhich half the patients in the group with a diagnosis of the disease are still alive; if at the end of the period there were not sufficient deaths to reach this num- ber, the median survival time is listed as undefined. Kaplan- Meier survival analysis was used to evaluate differences be- tween survival curves. The log-rank testwas used todetermine statistical differences using the threshold P < .05. Covariates were chosen formultivariate Cox proportional hazards regres- sionmodels with 95%confidence intervals with regards to OS and DSS. Multivariate analysis was conducted for the follow- ing groups of patients: overall patients with submandibular gland tumors, patients grouped by tumor stage (stage I or II and stage III or IV), tumor size (primary tumors ≤3 and >3 cm), and tumor histologic subtype (squamous cell carcinoma, ad- enocarcinoma, adenoid cystic carcinoma, and mucoepider- moid carcinoma). Tumor size, extent, and lymph node in- volvement were not used in themultivariate analysis because these were each used to determine tumor stage at presenta- tion, so using these variables in addition to stage at presenta- tion would violate the principle of excluding linearly code- pendent variables. Using thismethodology, therewere no less than 10 events per covariate for eachmodel. 31 Statistical analy- sis was performed using SPSS software, version 21.0 (SPSS).

Methods Data Source

Data were gathered from the SEER 18 Registries, estimated to encompass 27.8%of the US population, including 40%of His- panics, 23% of blacks, and 20 different US geographic re- gions. Because the SEER database is publicly available and all patient data are deidentified, no institutional reviewboard ap- proval or informed consent was required for this study.

JAMA Otolaryngology–Head & Neck Surgery October 2015 Volume 141, Number 10 (Reprinted)

jamaotolaryngology.com

19