Racial and Ethnic Disparities in Salivary Gland Cancer Survival

Abstract

Importance

Several recent studies have documented disparities in head and neck cancer outcomes for Black patients in the United States. However, few studies have been conducted to evaluate for differences in long-term survival from salivary gland cancer (SGCA) for racial/ethnic minorities compared to Whites.

Objective

To determine if patient race or ethnicity impact SGCA survival.

Design, Setting, and Participants

Retrospective survival analysis of all patients with SGCA from 1988–2010 in the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database, a population-level database of cancer registries covering approximately 28 percent of the US population.

Intervention

None

Main Outcomes and Measures

The primary outcome measured was disease-specific survival. Patients with SGCA were grouped according to race and ethnicity. The end points assessed in each group included age at diagnosis, gender, tumor grade, tumor size at diagnosis, extension at diagnosis, lymph node involvement at diagnosis, and treatment modalities. The racial/ethnic groups were further analyzed by histologic subtype of the SGCAs.

Results

Of 11,007 patients with SGCA, 1,073 (9.7%) and 1,068 (9.7%) were Black and Hispanic, respectively. The mean age at diagnosis for Whites was 63 years old compared to 53 and 52 years old for Blacks and Hispanics, respectively (p < 0.0001). Twenty-year disease-specific survival rates for all SGCA histologies combined for Whites, Blacks, and Hispanics were 78%, 79%, and 81%, respectively. The log-rank test of the unadjusted survival curves showed no significant difference in survival between Black and White patients and an apparent survival advantage for Hispanic compared to White patients. However, using multivariable Cox regression models to control for patient, tumor, and treatment characteristics, we demonstrated that Black patients actually have significantly poorer disease-specific survival for SGCA compared to White patients, while Hispanic patients have no significant difference in disease-specific survival compared to White patients. Further analyses of the individual SGCA histological subtypes identified poorer disease-specific survival for Black compared to White patients with mucoepidermoid and squamous cell carcinomas as the source of the overall poorer disease-specific survival for Black compared to White patients with SGCA. Less surgical treatment for Black compared to White patients was a significant source of the survival disparity for squamous cell SGCA, but not for mucoepidermoid SGCA.

Conclusions and Relevance

This is the largest study to date to explore racial and ethnic disparities in SGCA survival. Our results show that for patients diagnosed with SGCA, Black race is a risk factor for poorer disease-specific survival for those with mucoepidermoid or squamous cell carcinoma, while Hispanic ethnicity has no effect on disease-specific survival for any SGCA histology. Differences in treatment regimens between Black and White patients play a significant role in the disparity for squamous cell SGCA survival, but not in the disparity for mucoepidermoid SGCA survival. Differences in chemotherapy treatment, comorbidities, socioeconomic status, tumor genetic factors, and environmental exposures are potential but unproven additional sources of the racial survival disparities for both mucoepidermoid and squamous cell SGCA.

Introduction

Salivary gland cancers (SGCA) comprise a large variety of histopathologies, many of uncertain etiology, and as a whole are relatively uncommon; the aggregate incidence is estimated at 0.8 to 1.2 per 100,000 in the US population.¹ These aspects of disease make SGCA both a clinical and epidemiological challenge. In the past decade there has been increasing interest in the impact race and ethnicity play in cancer survival. While several studies^2–4 have documented disparities in head and neck cancer outcomes for Black patients in the United States, few studies have evaluated the effect race and ethnicity have on disease-specific survival for patients with SGCA. We therefore conducted this study with the primary objective of determining if racial and/or ethnic disparities exist for SGCA survival.

Materials and Methods

Data source

Completely de-identified data for all cases of SGCA for the years 1988 to 2010 were acquired from the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database on limited-use terms. The follow-up cutoff date for the November 2012 submission used in this analysis was December 31, 2010. The flow chart for the cohort selection is shown in Figure 1. Patients who were diagnosed at death or autopsy and those with incomplete demographic information were excluded from the analysis. Patients were also excluded if the salivary gland cancer was not the patient’s first cancer. Included primary sites as defined by the American Joint Committee on Cancer (AJCC) were identified by the following site codes: Parotid Gland (C079), Submandibular Gland (C080), Sublingual Gland (C081), Overlapping Lesion of Major Salivary Glands (C088), and Major Salivary Gland Not Otherwise Specified (C089). Histology codes were based on the SEER Histology Recode-Broad Groupings: epithelial, not otherwise specified (NOS) (01); squamous cell (02); adenomas and adenocarcinomas (05); mucoepidermoid neoplasms (07); cystic, mucinous and serous (08); ductal and lobular (09); acinar cell (10); complex epithelial (11); complex mixed and stromal (21). Information for a unique patient identifier, sex, age at diagnosis, year of diagnosis, sequence number, tumor size, tumor extension, lymph node involvement, surgical treatment, radiation use and method, race, Hispanic ethnicity, survival time, cause of death, vital status at last follow-up, histology, grade, and source of report were extracted for each case. These parameters were determined by the following SEER variables: Patient ID number, Sex, Age at Diagnosis, Year of Diagnosis, Sequence number-Central, EOD-Tumor Size, CS Tumor Size, EOD-Extension, CS Extension, EOD-Lymph Node Involv, CS Lymph Nodes, RX Summ-Surg Prim Site, RX Summ-Surgery Type, RX Summ-Radiation, Race Recode (W, B, AI, API), Origin Recode NHIA (Hispanic, Non-Hispanic), Survival Months, Cause of Death to SEER Site Recode, Vital Status Recode, Histology Recode-Broad Groupings, Grade, and Type of Reporting Source.

Figure 1.

Schematic flow chart detailing inclusion/exclusion criteria for cohort selection.

Data analysis

Data extraction and preparation were performed in SAS 9.3, (SAS Institute, Cary, North Carolina). Cases were separated into three racial/ethnic groups as White, Black, and Hispanic patients. Histology groups were classified as mucoepidermoid carcinoma, adenocarcinoma, adenoid cystic carcinoma, squamous cell carcinoma, acinar cell carcinoma, and other malignant histologies. For tumor characteristics, tumor grade was reclassified as either low-grade or high-grade. The low-grade group included tumors reported as Grade I, Grade II, well differentiated, moderately differentiated, and intermediate differentiation. The high-grade group included tumors reported as Grade III, Grade IV, poorly differentiated, undifferentiated, and anaplastic. Extension was reclassified as intraglandular, extraglandular, or metastatic. Lymph node involvement was reclassified into either no lymph node involvement or positive nodes. Additionally, for treatment characteristics, surgery and radiation therapy were dichotomized into yes or no.

All statistical analyses and figure generation were performed in R and SAS 9.3. Comparisons of Kaplan-Meier survival estimates for White, Black, and Hispanic populations were performed using the log-rank test, and confidence intervals were calculated using the log-transform method. For further investigating the magnitude of disparity in survival between racial groups, Cox regression models were built to provide hazard ratios with accompanying 95% confidence intervals for minority (Hispanic or Black) versus White by controlling patient characteristics (model 2), patient and tumor characteristics (model 3), and patient, tumor and treatment characteristics (model 4) for all histologies. When the number of events (death from SGCA) was adequate for analysis, a separate set of Cox regression models were built for each of the individual SGCA histologies.

Results

Overall, 11,007 patients met criteria for inclusion in the study. Of these, 8,866 (80.6%), 1,073 (9.7%), and 1,068 (9.7%) were White, Black, and Hispanic, respectively (see Table 1). The patients were 59.3% male in the White population compared to 48.1% and 48.9% in the Black and Hispanic populations (p < 0.0001). Notably, the mean age at diagnosis for Whites was 63 years old compared to 53 and 52 years old for Blacks and Hispanics, respectively (p < 0.0001). Mean tumor size at diagnosis for Whites (29.2 mm) was smaller compared to Blacks (33.5 mm) and Hispanics (31.2 mm; p < 0.0001). Examining tumor grade revealed a larger proportion (33.0%) of high-grade tumors in Whites compared to Blacks (24.3%) and Hispanics (20.7%; p < .0001). Whites had the greatest percentage of patients with positive lymph nodes (32.1%) while Blacks were intermediate (27.7%) and Hispanics had the least (24.6%; p < .0001). Tables 2 through 6 show how these characteristics are distributed for the individual SGCA histologies. For all SGCA histologies combined, unadjusted Kaplan-Meier five-year survival rates for Whites, Blacks, and Hispanics were 83%, 84%, and 89%, respectively, and 78%, 79%, and 81%, respectively, for twenty-year survival (see Table 7 and Figure 2). Using the log-rank test to compare the unadjusted survival curves, Blacks had no significant difference in disease-specific survival compared to Whites (p = 0.178), while Hispanics had significantly better disease-specific survival than Whites (p = 0.003).

Table 1.

Characteristics for Cases of All Histologies of SGCA Stratified by Race

	Race/Ethnicity
Variables	White	Black	Hispanic	P Value
Total (N=11,007)	8,866	1,073	1,068
Age at Diagnosis (yrs) 2				< 0.00011
Mean ± S.D.	62.57 ± 17.67	53.01 ± 17.97	52.34 ± 19.53
Median (I.Q.R)	64 (51 to 76)	53 (41 to 66)	53 (38 to 67)
Gender (%) 2				< 0.0001
Male	59.26	48.09	48.88
Female	40.74	51.91	51.12
Tumor Grade (%)				< 0.0001
Low Grade	30.27	35.41	39.61
High Grade	33.03	24.31	20.69
Unknown	36.70	40.26	39.70
Tumor Size at Diagnosis (mm) 2 (N=8,846)				< 0.00011
Mean ± S.D.	29.17 ± 20.53	33.47 ± 37.55	31.23 ± 21.21
Median (I.Q.R)	25 (16 to 35)	28 (18 to 40)	25 (18 to 40)
Lymph Nodes (%) 2 (N=9,519)				< 0.0001
No Lymph Node Involvement	67.94	72.34	75.42
Positive Nodes	32.06	27.66	24.58
Radiation (%) 2 (N=10,761)				0.0002
Yes	57.54	51.53	53.83
No	42.46	48.47	46.17
Surgery (%) 2 (N=10,968)
Yes	86.54	84.14	90.63	< 0.0001
No	13.64	15.86	9.37
Extension (%) (N=10,236)				0.032
Intraglandular	59.41	62.86	61.83
Extraglandular	37.32	32.93	35.21
Metastatic	3.27	4.20	2.96
Histology (%)				< 0.0001
Mucoepidermoid Carcinoma	21.10	28.70	28.56
Adenocarcinoma	15.20	14.63	11.89
Adenoid Cystic Carcinoma	11.02	14.35	13.76
Squamous Cell Carcinoma	21.01	10.90	13.58
Acinar Cell Carcinoma	12.54	12.77	14.42
Other Malignant Histologies	19.12	18.64	17.79
Diagnosis Year (%)				< 0.0001
1988 to 1995	17.97	16.68	13.01
1996 to 2003	36.42	35.04	29.87
2004 to 2010	45.61	48.28	57.12

¹P values were based on results of Kruskal-Wallis test.

²For continuous and binary categorical variables with a statistically significant difference on race, the results of multiple comparisons between race were summarized as White versus Hispanic (W/H) or/and White versus Black (W/B) or/and Hispanic versus Black (H/B) with p values < 0.05. For age and tumor size at diagnosis, both were W/H and W/B. For gender, lymph nodes and radiation, all were W/H and W/B. For surgery, the differences were W/H and H/B.

S.D. = standard deviation; I.Q.R. = interquartile range

Table 2.

Characteristics for Cases of Mucoepidermoid Carcinoma Stratified by Race

	Race/Ethnicity
Variables	White	Black	Hispanic	P Value
Total (N=2,484)	1,871	308	305
Age at Diagnosis (yrs) 3				< 0.0001 1
Mean ± S.D.	56.73 ± 19.01	47.51 ± 18.67	45.04 ± 19.56
Median (I.Q.R)	58 (44 to 71)	47 (35 to 61)	44 (28 to 59)
Gender (%) 3				0.002
Male	53.39	46.75	43.93
Female	46.61	53.25	56.07
Tumor Grade (%) 3 (N=2,203)				< 0.0001
Low Grade	64.42	77.62	83.15
High Grade	35.58	22.38	16.85
Tumor Size at Diagnosis (mm) 3 (N=2,097)				0.001 1
Mean ± S.D.	24.05 ± 16.67	27.11 ± 16.93	24.50 ± 13.95
Median (I.Q.R)	20 (13 to 30)	24 (15 to 33)	21 (15 to 30)
Lymph Nodes (%) 3 (N=2,195)				0.081
No Lymph Node Involvement	76.07	78.99	81.75
Positive Nodes	23.93	21.01	18.25
Radiation (%) 3 (N=2,446)				0.016
Yes	48.16	44.37	39.60
No	51.84	55.63	60.40
Surgery (%) 3 (N=2,482)				0.049
Yes	92.83	95.13	96.07
No	7.17	4.87	3.93
Extension (%) (N=2,369)				0.410 2
Intraglandular	69.62	74.40	73.22
Extraglandular	28.52	24.57	25.42
Metastatic	1.85	1.02	1.36
Diagnosis Year (%)				< 0.0001
1988 to 1995	19.72	19.48	11.15
1996 to 2003	38.05	31.49	27.87
2004 to 2010	42.22	49.03	60.98

¹P values were based on results of Kruskal-Wallis test.

²P values were based on results of Fisher Exact test.

³For continuous and binary categorical variables with a statistically significant difference on race, the results of multiple comparisons between race were summarized as White versus Hispanic (W/H) or/and White versus Black (W/B) or/and Hispanic versus Black (H/B) with p values < 0.05. For age at diagnosis, the differences were W/H and W/B. For tumor size at diagnosis, the difference was W/B. For gender and tumor grade, both were W/H and W/B. For radiation and surgery, both were W/H.

S.D. = standard deviation; I.Q.R. = interquartile range

Table 6.

Characteristics for Cases of Acinar Cell Carcinoma Stratified by Race

	Race/Ethnicity
Variables	White	Black	Hispanic	P Value
Total (N=1,403)	1,112	137	154
Age at Diagnosis (yrs) 3				< 0.0001 1
Mean ± S.D.	51.33 ± 19.16	42.34 ± 17.19	42.82 ± 17.50
Median (I.Q.R)	52 (37 to 66)	42 (30 to 54)	43 (30 to 56)
Gender (%) 3				0.004
Male	43.17	28.47	40.26
Female	56.83	71.53	59.74
Tumor Grade (%)				0.024
Low Grade	29.05	23.36	27.92
High Grade	5.94	3.65	0.65
Unknown	65.02	72.99	71.43
Tumor Size at Diagnosis (mm) 3 (N=1,245)				0.005 1
Mean ± S.D.	23.42 ± 13.91	27.38 ± 16.00	26.24 ± 15.14
Median (I.Q.R)	20 (15 to 30)	22 (17 to 35)	23 (15 to 35)
Lymph Nodes (%) 3 (N=1,236)				0.004 2
No Lymph Node Involvement	88.16	91.94	96.45
Positive Nodes	11.84	8.06	3.55
Radiation (%) (N=1,376)				0.147
Yes	40.29	32.59	35.57
No	59.71	67.41	64.43
Surgery (%) (N=1,402)				0.673
Yes	96.04	95.62	97.40
No	3.96	4.38	2.60
Extension (%) (N=1,345)				0.607 2
Intraglandular	84.04	87.02	88.59
Extraglandular	15.30	12.98	11.41
Metastatic	0.66	0.00	0.00
Diagnosis Year (%)				0.351
1988 to 1995	19.42	17.52	12.99
1996 to 2003	35.16	33.58	35.71
2004 to 2010	45.41	48.90	51.30

¹P values were based on results of Kruskal-Wallis test.

²P values were based on results of Fisher Exact test.

³For continuous and binary categorical variables with a statistically significant difference on race, the results of multiple comparisons between race were summarized as White versus Hispanic (W/H) or/and White versus Black (W/B) or/and Hispanic versus Black (H/B) with p values < 0.05. For age at diagnosis, the differences were W/H and W/B. For tumor size at diagnosis, the difference was W/B. For gender, the difference was W/B. For tumor grade, the difference was W/H. For lymph nodes, the differences were W/H and W/B.

S.D. = standard deviation; I.Q.R. = interquartile range

Table 7.

Kaplan-Meier Survival Rates and Log-Rank Multiple Comparisons on Race for Years 1988 to 2010

	Survival Rate				Log-Rank Test
	5-year (95% CI)	10-year (95% CI)	15-year (95% CI)	20-year (95% CI)	P Value 1
All Histologies					0.002
White	0.83 (0.82 to 0.84)	0.81 (0.80 to 0.82)	0.79 (0.78 to 0.80)	0.78 (0.77 to 0.80)	–
Black	0.84 (0.82 to 0.87)	0.82 (0.79 to 0.85)	0.81 (0.77 to 0.84)	0.79 (0.75 to 0.84)	0.178	–
Hispanic	0.89 (0.86 to 0.91)	0.85 (0.82 to 0.88)	0.83 (0.79 to 0.87)	0.81 (0.76 to 0.86)	0.003	0.071
Mucoepidermoid Carcinoma					0.020
White	0.89 (0.87 to 0.91)	0.87 (0.86 to 0.89)	0.87 (0.85 to 0.89)	0.87 (0.85 to 0.89)	–
Black	0.88 (0.84 to 0.93)	0.88 (0.84 to 0.93)	0.88 (0.84 to 0.93)	0.88 (0.84 to 0.93)	0.351	–
Hispanic	0.95 (0.92 to 0.98)	0.93 (0.89 to 0.97)	0.93 (0.89 to 0.97)	0.93 (0.89 to 0.97)	0.022	0.186
Adenocarcinoma					0.311
White	0.74 (0.71 to 0.77)	0.68 (0.65 to 0.72)	0.67 (0.63 to 0.71)	0.67 (0.63 to 0.71)	–
Black	0.74 (0.67 to 0.83)	0.73 (0.65 to 0.82)	0.69 (0.60 to 0.81)	–	0.815	–
Hispanic	0.81 (0.73 to 0.89)	0.77 (0.68 to 0.87)	0.77 (0.68 to 0.88)	–	0.357	0.550
Adenoid Cystic Carcinoma					0.969
White	0.89 (0.86 to 0.91)	0.84 (0.81 to 0.87)	0.78 (0.74 to 0.83)	0.75 (0.70 to 0.81)	–
Black	0.91 (0.86 to 0.96)	0.81 (0.73 to 0.90)	0.78 (0.69 to 0.89)	0.72 (0.58 to 0.89)	0.999	–
Hispanic	0.89 (0.83 to 0.96)	0.84 (0.76 to 0.93)	0.76 (0.65 to 0.90)	0.76 (0.65 to 0.90)	0.990	0.968
Squamous Cell Carcinoma					0.097
White	0.81 (0.79 to 0.83)	0.79 (0.77 to 0.82)	0.79 (0.77 to 0.82)	0.79 (0.77 to 0.82)	–
Black	0.70 (0.59 to 0.83)	0.70 (0.59 to 0.83)	0.70 (0.59 to 0.83)	0.70 (0.59 to 0.83)	0.086	–
Hispanic	0.76 (0.66 to 0.87)	0.69 (0.57 to 0.83)	0.69 (0.57 to 0.83)	–	0.346	0.664
Acinar Cell Carcinoma					0.061
White	0.95 (0.93 to 0.96)	0.93 (0.91 to 0.95)	0.92 (0.90 to 0.94)	0.91 (0.88 to 0.94)	–
Black	0.97 (0.93 to 1.00)	0.94 (0.88 to 1.00)	0.94 (0.88 to 1.00)	0.94 (0.88 to 1.00)	0.128	–
Hispanic	0.99 (0.98 to 1.00)	0.99 (0.98 to 1.00)	0.99 (0.98 to 1.00)	0.92 (0.78 to 1.00)	0.049	0.798

¹P values for multiple comparisons of the Log-rank test were based on the Tukey-Kramer adjustment.

Figure 2.

Kaplan-Meier estimates of disease-specific survival for White, Black, and Hispanic patients. The survival probability and the corresponding 95% confidence interval are represented by solid lines and dashed lines, respectively.

The multivariable Cox regression models provide further insight as to how patient, tumor, and treatment characteristics affect disease-specific survival in Whites, Blacks, and Hispanics for all SGCA histologies combined (see Table 8). The models were built starting with the crude, unadjusted model (model 1), then progress to controlling for patient characteristics (age and gender; model 2), then to controlling for patient characteristics and tumor characteristics (histology, tumor grade, tumor size, extension, and lymph node status; model 3), and finally progress to controlling for patient characteristics, tumor characteristics, and treatment characteristics (surgery and radiation; model 4). For all SGCA histologies, the unadjusted Cox model (model 1) showed the same results at the unadjusted Kaplan-Meier estimates: no significant difference in disease-specific survival for Blacks compared to Whites (HR: 0.97, p = 0.74) but an apparent survival advantage for Hispanics compared to Whites (HR: 0.74, p = 0.001). However, when controlling for patient age and gender (model 2), there was no significant difference in disease-specific survival for Hispanics compared to Whites (HR: 0.97, p = 0.76), and a statistically worse disease-specific survival for Blacks compared to Whites (HR: 1.34, p = 0.001) was revealed. Controlling for tumor characteristics in addition to patient characteristics (model 3) and for patient, tumor, and treatment characteristics (model 4) yielded similar results to model 2: no significant difference in disease-specific survival between Hispanic and White patients, but a statistically worse disease-specific survival for Black compared to White patients (see Table 8). Additionally, older age at diagnosis, high tumor grade, larger tumor size, extraglandular or metastatic extension, positive lymph nodes, and no surgical treatment each had a statistically significant detriment on survival for all patients.

Table 8.

Multivariable Cox Regression Modeling Analysis on Survival Disparities of Race/Ethnicity for All Histologies of SGCA ¹

	Model 1 3	Model 2 3	Model 3 3	Model 4 3
	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Race/Ethnicity 4
White	Ref.	Ref.	Ref.	Ref.
Hispanic	0.70 (0.58 to 0.86)	0.97 (0.79 to 1.19)	0.94 (0.77 to 1.15)	0.97 (0.79 to 1.19)
Black	0.97 (0.82 to 1.15)	1.34 (1.13 to 1.60)	1.29 (1.08 to 1.54)	1.22 (1.03 to 1.46)
Gender
Male		Ref.	Ref.	Ref.
Female		0.73 (0.65 to 0.81)	1.00 (0.90 to 1.12)	1.00 (0.89 to 1.11)
Age at Diagnosis
≤ 49		Ref.	Ref.	Ref.
49 to 65		2.19 (1.84 to 2.61)	1.46 (1.22 to 1.74)	1.40 (1.17 to 1.67)
65 to 75		3.43 (2.87 to 4.10)	2.11 (1.76 to 2.53)	2.03 (1.69 to 2.44)
> 75		4.92 (4.13 to 5.85)	2.76 (2.31 to 3.31)	2.58 (2.15 to 3.09)
Histology
Mucoepidermoid Carcinoma			Ref.	Ref.
Adenocarcinoma			1.52 (1.28 to 1.80)	1.47 (1.24 to 1.74)
Adenoid Cystic Carcinoma			1.16 (0.93 to 1.44)	1.11 (0.89 to 1.34)
Squamous Cell Carcinoma			0.95 (0.80 to 1.14)	0.85 (0.71 to 1.01)
Acinar Cell Carcinoma			0.69 (0.52 to 0.92)	0.75 (0.56 to 0.99)
Other Malignant Histologies			1.32 (1.11 to 1.56)	1.21 (1.02 to 1.44)
Tumor Grade 2
Low			Ref.	Ref.
High			2.45 (2.07 to 2.90)	2.22 (1.87 to 2.63)
Tumor Size (mm) 2
≤ 17			Ref.	Ref.
17 to 25			1.78 (1.41 to 2.24)	1.75 (1.39 to 2.21)
25 to 37			2.14 (1.70 to 2.70)	2.07 (1.65 to 2.61)
> 37			3.03 (2.43 to 3.78)	2.74 (2.19 to 3.41)
Extension 2
Intraglandular			Ref.	Ref.
Extraglandular			2.55 (2.24 to 2.91)	2.40 (2.11 to 2.74)
Metastatic			6.29 (5.09 to 7.77)	4.29 (3.43 to 5.36)
Lymph Nodes 2
No Involvement			Ref.	Ref.
Positive			2.34 (2.06 to 2.66)	2.27 (2.00 to 2.57)
Surgery
Yes				Ref.
No				2.66 (2.32 to 3.06)

¹In Cox regression analysis, only patients with complete treatment characteristics were considered (N=10,745). Age at diagnosis and tumor size were divided into quartiles. For age at diagnosis, the second quartile, 62, was replace by 65. Except for model 1 (crude model), diagnosis year was included into the models (not shown).

²For carrying out the analyses for all models with the same number of patients, the patients with unknown tumor grade, tumor size, extension, or lymph nodes were included.

³When fitting the Cox regression models, the assumption of proportionality was assessed by the significance of a term of the predictor associated with the logarithm of survival time, and an appropriate stratified analysis was adopted if needed. In model 2 (patient characteristics) and model 3 (patient and tumor characteristics), there were no violations of this assumption. In model 4 (patient, tumor and treatment characteristics), a severe violation appeared on radiation (yes/no) and a stratified Cox model on radiation was built; therefore, while not shown, radiation is included in model 4.

⁴The corresponding p values of hazard ratios (HR) on the disparity of Hispanic versus White (H/W) and Black versus White (B/W) were shown as follows. In model 1, H/W and B/W were 0.001 and 0.74, respectively; in model 2, H/W and B/W were 0.76 and 0.001, respectively; in model 3, H/W and B/W were 0.55 and 0.005, respectively and in model 4, H/W and B/W were 0.77 and 0.03, respectively. Bold values highlight statistically significant results (p < 0.05).

We next analyzed the survival data by individual SGCA histologies. For patients with mucoepidermoid carcinoma, unadjusted Kaplan-Meier five-year survival rates for Whites, Blacks, and Hispanics were 89%, 88%, and 95%, respectively, and 87%, 88%, and 93%, respectively, for twenty-year survival (see Table 7 and Figure 2). Using the log-rank test to compare the unadjusted survival curves, Blacks had no significant difference in disease-specific survival compared to Whites (p = 0.351), while Hispanics had significantly better disease-specific survival compared to Whites (p = 0.022). The unadjusted Cox model (Table 9, model 1) for mucoepidermoid carcinoma also showed no significant difference in disease-specific survival between Black and White patients and an apparent survival advantage for Hispanic compared to White patients. Controlling for patient age and gender (model 2) showed no significant difference in disease-specific survival for Black compared to White patients and eliminated the apparent survival advantage for Hispanic compared to White patients. However, in model 2 (patient characteristics) and model 3 (patient and tumor characteristics), a trend towards poorer disease-specific survival for Black compared to White patients was noted but did not reach statistical significance. Progressing further to controlling for patient, tumor, and treatment characteristics (model 4), when compared to White patients, there remained no statistical difference in disease-specific survival for Hispanic patients (HR: 0.76, p = 0.36), but the trend towards worse survival for Black patients became statistically significant (HR: 1.56, p = 0.03).

Table 9.

Multivariable Cox Regression Modeling Analysis on Survival Disparities of Race/Ethnicity for Mucoepidermoid Carcinoma ¹

	Model 1 3	Model 2 3	Model 3 3	Model 4 3
	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Race/Ethnicity 4
White	Ref.	Ref.	Ref.	Ref.
Hispanic	0.45 (0.26 to 0.79)	0.68 (0.39 to 1.20)	0.73 (0.41 to 1.30)	0.76 (0.43 to 1.36)
Black	0.94 (0.63 to 1.40)	1.42 (0.95 to 2.12)	1.47 (0.98 to 2.21)	1.56 (1.04 to 2.36)
Gender
Male		Ref.	Ref.	Ref.
Female		0.55 (0.41 to 0.72)	0.91 (0.68 to 1.21)	0.95 (0.71 to 1.26)
Age at Diagnosis
≤ 49		Ref.	Ref.	Ref.
49 to 65		3.48 (2.22 to 5.45)	2.23 (1.42 to 3.51)	2.22 (1.41 to 3.49)
65 to 75		6.00 (3.77 to 9.54)	3.20 (2.00 to 5.14)	3.31 (2.06 to 5.32)
> 75		8.81 (5.61 to 13.8)	3.64 (2.29 to 5.79)	3.30 (2.07 to 5.27)
Tumor Grade 2
Low			Ref.	Ref.
High			5.69 (3.59 to 9.02)	4.94 (3.11 to 7.85)
Tumor Size (mm) 2
≤ 17			Ref.	Ref.
17 to 25			2.87 (1.57 to 5.27)	2.66 (1.45 to 4.89)
25 to 37			1.94 (1.02 to 3.71)	1.77 (0.93 to 3.37)
> 37			4.97 (2.71 to 9.11)	3.66 (1.98 to 6.74)
Extension 2
Intraglandular			Ref.	Ref.
Extraglandular			2.45 (1.74 to 3.45)	2.52 (1.79 to 3.56)
Metastatic			5.77 (3.09 to 10.8)	2.45 (1.26 to 4.80)
Lymph Nodes 2
No Involvement			Ref.	Ref.
Positive			2.28 (1.66 to 3.14)	2.15 (1.56 to 2.95)
Surgery
Yes				Ref.
No				4.70 (3.12 to 7.08)
Radiation
Yes				Ref.
No				0.57 (0.40 to 0.81)

¹In Cox regression analysis, only patients with complete treatment characteristics were considered (N=2,445). Except for model 1 (crude model), diagnosis year was included into the model (not shown).

²For carrying out the analyses for all models with the same number of patients, the patients with unknown tumor grade, tumor size, extension, or lymph nodes were included.

³In the Cox regression models, no significant violation of the assumption of proportionality was detected. Therefore, in model 2 (patient characteristics), model 3 (patient and tumor characteristics), and model 4 (patient, tumor, and treatment characteristics), no stratified analyses were carried out.

⁴The corresponding p values of hazard ratios (HR) on the disparity of Hispanic versus White (H/W) and Black versus White (B/W) were shown as follows. In model 1, H/W and B/W were 0.006 and 0.77, respectively; in model 2, H/W and B/W were 0.18 and 0.09, respectively; in model 3, H/W and B/W were 0.29 and 0.06, respectively and in model 4, H/W and B/W were 0.36 and 0.03, respectively. Bold values highlight statistically significant results (p < 0.05).

For patients with adenocarcinoma, unadjusted Kaplan-Meier five-year survival rates for Whites, Blacks, and Hispanics were 74%, 74%, and 81%, respectively, and 67%, 69%, and 77%, respectively, for fifteen-year survival (see Figure 2 and Table 7). There were an insufficient number of cases to calculate reliable twenty-year survival rates for Black or Hispanic patients. Using the log-rank test to compare the unadjusted survival curves, there was no statistically significant difference in disease-specific survival between Black and White patients (p = 0.815) or Hispanic and White patients (p = 0.357). The multivariable Cox regression analysis presented in Table 10 shows that when controlling for patient, tumor, and treatment characteristics, there remained no statistically significant difference in disease-specific survival for Hispanic compared to White patients (HR: 0.88, p = 0.58) or Black compared to White patients (HR: 1.15, p = 0.45).

Table 10.

Multivariable Cox Regression Modeling Analysis on Survival Disparities of Race/Ethnicity for Adenocarcinoma ¹

	Model 1 3	Model 2 3	Model 3 3	Model 4 3
	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Race/Ethnicity 4
White	Ref.	Ref.	Ref.	Ref.
Hispanic	0.69 (0.44 to 1.09)	0.86 (0.55 to 1.36)	0.92 (0.58 to 1.47)	0.88 (0.55 to 1.40)
Black	1.01 (0.70 to 1.45)	1.16 (0.81 to 1.67)	1.33 (0.92 to 1.92)	1.15 (0.79 to 1.68)
Gender
Male		Ref.	Ref.	Ref.
Female		0.65 (0.52 to 0.81)	0.90 (0.71 to 1.13)	0.89 (0.71 to 1.13)
Age at Diagnosis
≤ 49		Ref.	Ref.	Ref.
49 to 65		1.46 (1.02 to 2.11)	1.12 (0.78 to 1.62)	1.10 (0.76 to 1.58)
65 to 75		2.16 (1.49 to 3.12)	1.55 (1.07 to 2.25)	1.44 (0.99 to 2.10)
> 75		2.23 (1.53 to 3.24)	1.75 (1.20 to 2.56)	1.55 (1.06 to 2.28)
Tumor Grade 2
Low			Ref.	Ref.
High			3.17 (2.19 to 4.60)	3.06 (2.09 to 4.46)
Tumor Size (mm) 2
≤ 17			Ref.	Ref.
17 to 25			1.27 (0.82 to 1.97)	1.26 (0.81 to 1.96)
25 to 37			1.84 (1.21 to 2.80)	1.69 (1.10 to 2.58)
> 37			1.87 (1.24 to 2.83)	1.68 (1.10 to 2.55)
Extension 2
Intraglandular			Ref.	Ref.
Extraglandular			2.26 (1.73 to 2.95)	2.19 (1.67 to 2.88)
Metastatic			4.73 (3.15 to 7.09)	3.38 (2.22 to 5.15)
Lymph Nodes 2
No Involvement			Ref.	Ref.
Positive			2.80 (2.11 to 3.71)	2.76 (2.08 to 3.65)
Surgery
Yes				Ref.
No				3.14 (2.35 to 4.21)
Radiation
Yes				Ref.
No				0.94 (0.73 to 1.21)

¹In Cox regression analysis, only patients with complete treatment characteristics were considered (N=1,593). Except for model 1 (crude model), diagnosis year was included into the models (not shown).

²For carrying out the analyses for all models with the same number of patients, the patients with unknown tumor grade, tumor size, extension, or lymph nodes were included.

³In the Cox regression models, no significant violation of the assumption of proportionality was detected. Therefore, in model 2 (patient characteristic), model 3 (patient and tumor characteristics), and model 4 (patient, tumor, and treatment characteristics), no stratified analyses were carried out.

⁴The corresponding p values of hazard ratios (HR) on the disparity of Hispanic versus White (H/W) and Black versus White (B/W) were shown as follows. In model 1, H/W and B/W were 0.12 and 0.96, respectively; in model 2, H/W and B/W were 0.52 and 0.42, respectively; in model 3, H/W and B/W were 0.73 and 0.13, respectively and in model 4, H/W and B/W were 0.58 and 0.45, respectively.

For patients with adenoid cystic carcinoma, unadjusted Kaplan-Meier five-year survival rates for Whites, Blacks, and Hispanics were 89%, 91%, and 89%, respectively, and 75%, 72%, and 76%, respectively, for twenty-year survival (see Table 7 and Figure 2). There was no statistically significant difference in disease-specific survival between Black and White patients (p = 0.999) or Hispanic and White patients (p = 0.990). The multivariable Cox regression analysis presented in Table 11 shows that when controlling for patient, tumor, and treatment characteristics, there remained were no statistically significant difference in disease-specific survival for Hispanic compared to White patients (HR: 1.05, p = 0.86) or Black compared to White patients (HR: 1.15, p = 0.59).

Table 11.

Multivariable Cox Regression Modeling Analysis on Survival Disparities of Race/Ethnicity for Adenoid Cystic Carcinoma ¹

	Model 1 3	Model 2 3	Model 3 3	Model 4 3
	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Race/Ethnicity 4
White	Ref.	Ref.	Ref.	Ref.
Hispanic	0.88 (0.52 to 1.51)	1.05 (0.70 to 1.85)	1.03 (0.59 to 1.80)	1.05 (0.60 to 1.85)
Black	1.04 (0.64 to 1.69)	1.14 (0.61 to 1.82)	1.09 (0.66 to 1.79)	1.15 (0.69 to 1.90)
Gender
Male		Ref.	Ref.	Ref.
Female		0.82 (0.60 to 1.13)	1.08 (0.77 to 1.50)	1.06 (0.76 to 1.48)
Age at Diagnosis
≤ 49		Ref.	Ref.	Ref.
49 to 65		1.53 (1.03 to 2.26)	1.37 (0.92 to 2.02)	1.34 (0.91 to 1.99)
65 to 75		2.16 (1.40 to 3.33)	1.78 (1.13 to 2.79)	1.68 (1.06 to 2.66)
> 75		2.46 (1.40 to 4.33)	1.96 (1.11 to 3.46)	1.77 (0.99 to 3.15)
Tumor Grade 2
Low			Ref.	Ref.
High			3.68 (1.84 to 7.36)	3.48 (1.74 to 6.99)
Tumor Size (mm) 2
≤ 17			Ref.	Ref.
17 to 25			2.21 (1.18 to 4.17)	2.23 (1.20 to 4.25)
25 to 37			2.89 (1.52 to 5.50)	2.88 (1.52 to 5.48)
> 37			3.62 (1.95 to 6.73)	3.46 (1.87 to 6.46)
Extension 2
Intraglandular			Ref.	Ref.
Extraglandular			2.38 (1.63 to 3.47)	2.35 (1.61 to 3.44)
Metastatic			8.76 (4.48 to 17.1)	6.81 (3.25 to 14.3)
Lymph Nodes 2
No Involvement			Ref.	Ref.
Positive			2.75 (1.89 to 4.01)	2.50 (1.70 to 3.67)
Surgery
Yes				Ref.
No				2.02 (1.20 to 3.42)
Radiation
Yes				Ref.
No				0.81 (0.53 to 1.22)

¹In Cox regression analysis, only patients with complete treatment characteristics were considered (N=1,242). Except for model 1 (crude model), diagnosis year was included into the model (not shown).

²For carrying out the analyses for all models with the same number of patients, the patients with unknown tumor grade, tumor size, extension, or lymph nodes were included.

³In the Cox regression models, no significant violation of the assumption of proportionality was detected. Therefore, in model 2 (patient characteristics), model 3 (patient and tumor characteristics), and model 4 (patient, tumor, and treatment characteristics), no stratified analyses were carried out.

⁴The corresponding p values of hazard ratios (HR) on the disparity of Hispanic versus White (H/W) and Black versus White (B/W) were shown as follows. In model 1, H/W and B/W were 0.65 and 0.87, respectively; in model 2, H/W and B/W were 0.85 and 0.60, respectively; in model 3, H/W and B/W were 0.93 and 0.74, respectively and in model 4, H/W and B/W were 0.86 and 0.59, respectively.

For patients with squamous cell carcinoma, unadjusted Kaplan-Meier five-year survival rates for Whites, Blacks, and Hispanics were 81%, 70%, and 76%, respectively, and 79%, 70%, and 69%, respectively, for fifteen-year survival (see Figure 2 and Table 7). There were an insufficient number of cases to calculate reliable twenty-year survival rates for Hispanic patients. Using the log-rank test to compare the unadjusted survival curves, there was no statistically significant difference in survival between Black and White patients (p = 0.086) or Hispanic and White patients (p = 0.346). However, the unadjusted Cox model (Table 12, model 1) for squamous cell carcinoma showed no significant difference in disease-specific survival between Hispanic and White patients but a statistically significant worse disease-specific survival for Black compared to White patients. Progressing through the Cox models to control for patient characteristics (model 2), patient and tumor characteristics (model 3), and patient, tumor, and treatment characteristics (model 4), the trends of equivalent disease-specific survival for Hispanic and White patients but statistically significant worse disease-specific survival for Black compared to White patients were maintained, though in model 4, the result for Black compared to White patients became marginally statistically significant (HR:1.58, p = 0.05).

Table 12.

Multivariable Cox Regression Modeling Analysis on Survival Disparities of Race/Ethnicity for Squamous Cell Carcinoma ¹

	Model 1 3	Model 2 3	Model 3 3	Model 4 3
	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Race/Ethnicity 4
White	Ref.	Ref.	Ref.	Ref.
Hispanic	1.20 (0.78 to 1.86)	1.27 (0.82 to 1.97)	1.13 (0.72 to 1.75)	1.16 (0.75 to 1.81)
Black	1.65 (1.07 to 2.56)	1.99 (1.27 to 3.13)	1.77 (1.12 to 2.80)	1.58 (0.99 to 2.51)
Gender
Male		Ref.	Ref.	Ref.
Female		1.14 (0.87 to 1.48)	1.19 (0.91 to 1.55)	1.19 (0.91 to 1.56)
Age at Diagnosis
≤ 49		Ref.	Ref.	Ref.
49 to 65		0.94 (0.54 to 1.64)	0.93 (0.53 to 1.63)	0.89 (0.51 to 1.55)
65 to 75		1.55 (0.91 to 2.64)	1.74 (1.01 to 2.98)	1.71 (1.00 to 2.93)
> 75		2.24 (1.34 to 3.75)	2.42 (1.43 to 4.09)	2.39 (1.41 to 4.03)
Tumor Grade 2
Low			Ref.	Ref.
High			1.09 (0.83 to 1.44)	1.06 (0.81 to 1.40)
Tumor Size (mm) 2
≤ 17			Ref.	Ref.
17 to 25			1.84 (0.98 to 3.46)	1.85 (0.98 to 3.48)
25 to 37			2.08 (1.11 to 3.90)	2.12 (1.13 to 3.99)
> 37			3.30 (1.82 to 5.99)	3.06 (1.68 to 5.55)
Extension 2
Intraglandular			Ref.	Ref.
Extraglandular			2.36 (1.72 to 3.22)	2.27 (1.66 to 3.11)
Metastatic			3.64 (2.01 to 6.58)	2.73 (1.49 to 5.00)
Lymph Nodes 2
No Involvement			Ref.	Ref.
Positive			1.41 (1.08 to 1.86)	1.45 (1.10 to 1.90)
Surgery
Yes				Ref.
No				2.19 (1.63 to 2.94)

¹In Cox regression analysis, only patients with complete treatment characteristics were considered (N=2,055). Except for model 1 (crude model), diagnosis year was included into the model (not shown).

²For carrying out the analyses for all models with the same number of patients, the patients with unknown tumor grade, tumor size, extension, or lymph nodes were included.

³When fitting the Cox regression models, the assumption of proportionality was assessed by the significance of a term of the predictor associated with the logarithm of survival time, and an appropriate stratified analysis was adopted if needed. In model 2 (patient characteristics) and model 3 (patient and tumor characteristics), there were no violations of this assumption. In model 4 (patient, tumor and treatment characteristics), a severe violation appeared on radiation (yes/no) and a stratified Cox model on radiation was built; therefore, while not shown, radiation is included in model 4.

⁴The corresponding p values of hazard ratios (HR) on the disparity of Hispanic versus White (H/W) and Black versus White (B/W) were shown as follows. In model 1, H/W and B/W were 0.40 and 0.02, respectively; in model 2, H/W and B/W were 0.28 and 0.003, respectively; in model 3, H/W and B/W were 0.60 and 0.02, respectively and in model 4, H/W and B/W were 0.50 and 0.05, respectively. Bold values highlight statistically significant results (p < 0.05).

For patients with acinar cell carcinoma, unadjusted Kaplan-Meier five-year survival rates for Whites, Blacks, and Hispanics were 95%, 97%, and 99%, respectively, and 91%, 94%, and 92%, respectively, for twenty-year survival (see Table 7 and Figure 2). Using the log-rank test to compare the unadjusted survival curves, Blacks had no difference in disease-specific survival compared to Whites (p = 0.128), while Hispanics had marginally statistically significant better disease-specific survival compared to Whites (p = 0.049). When attempting to perform multivariable Cox regression modeling for this histologic subgroup, due to the very good survival rates for all patient groups, the proportion of censoring was as high as 93% and the standard error on the parameter estimates was too wide, which raised concern for the validity of any results obtained from the modeling; therefore, multivariable analysis was not performed for acinar cell carcinoma. However, given that such an analysis was prevented by very good survival in all racial/ethnic groups with this SGCA histology, any disparity in survival would likely be exceedingly small.

Discussion

Significant disparities in survival have been demonstrated for Black patients with cancers of the head and neck within the last decade.^2–5 Molina and colleagues² found that for all locations of squamous cell carcinoma of the head and neck, median survival time for Black patients was significantly lower (21 months) compared to White patients (40 months, p < 0.001). Likewise, Nichols and colleagues³ demonstrated that for glottic squamous cell carcinoma, both mean overall survival and disease-specific survival were significantly worse in Black patients compared to matched White controls. The authors of these studies found that a large portion of the disparity in survival for Black patients is explained by more advanced disease at presentation, poorer socioeconomic status, comorbidities, and less surgical treatment.^{2, 3} However, even when controlling for these variables, poorer survival persisted in the Black populations, leading the authors to speculate that other environmental, and perhaps genetic, factors play a role in the survival disparity.^{2, 3}

While numerous studies^6–15 in the literature have evaluated prognostic factors in SGCA survival, very few^{6, 14, 15} have evaluated the role race and ethnicity contributes to survival in these cancers. In 1984, Spitz and Batsakis noted improved overall survival for Black patients compared to White patients in their study of 498 patients with SGCA from their institution; however, this finding was not subjected to multivariate analysis; additionally, no meaningful comparisons of survival could be made for Hispanic patients, as only 7 of the 498 patients were Hispanic.⁶ In 2011, Cheung and colleagues found that race and ethnicity had no effect (as determined by multivariable Cox regression analysis) on overall survival in a study of 968 patients with SGCA from a Florida cancer registry; this study included Hispanic patients and data on individual patient comorbidities.¹⁴ In 2012, Ellington and colleagues studied incidence and survival trends for adenoid cystic SGCA in 3,026 patients from the SEER database from 1973 to 2007; multivariable Cox regression models showed that race (defined as White, Black, or Other) had no effect on observed survival; ethnicity was not addressed in the study.¹⁵

Given the sparse literature on the topic, we conducted this study to determine if racial/ethnic disparities exist for SGCA survival. To our knowledge, this is the first study to examine racial and ethnic disparities in SGCA survival as the primary objective, and it is the largest study to date to examine effects of race/ethnicity on SGCA survival when considering studies that have addressed race and/or ethnicity as a prognostic factor.^{6, 14, 15} Overall, when looking at all SGCA histologies combined, our results show that there is an apparent survival advantage in SGCA for Hispanic patients compared to White patients that is explained entirely by younger age at diagnosis, whereas for Black patients compared to White patients, there is an underlying statistically significant worse survival for SGCA that is revealed by controlling for patient characteristics (age and gender) and maintained when additionally controlling for tumor and treatment characteristics (Table 8).

Looking closer at the survival data for the individual SGCA histologies (Tables 7, 9–12), we see that this disparity is driven by poorer disease-specific survival for Black patients with mucoepidermoid (Table 9) and squamous cell (Table 12) carcinomas. However, the nature of the disparity differs between these two SGCA subtypes. For mucoepidermoid carcinoma, while a trend towards worse disease-specific survival was noted as patient and patient and tumor variables were controlled (Table 9, models 2 and 3), a statistically significant poorer disease-specific survival for Black patients was not observed until controlling for patient, tumor, and treatment variables (surgery and radiation, model 4), indicating that the poorer survival for Black compared to White patients with mucoepidermoid carcinoma is due to other variables that were not controlled. This is in contradistinction to the pattern we see for squamous cell carcinoma, where the unadjusted Cox model (Table 12, model 1) shows a baseline statistically significant poorer disease-specific survival for Black compared to White patients. As we progress through the Cox models for squamous cell carcinoma, controlling for patient, tumor, and treatment variables, this poorer disease-specific survival for Black compared to White patients is maintained until treatment variables are added (surgery and radiation, model 4). At this point the difference in disease-specific survival becomes marginally statistically significant, implying that the poorer survival for Black compared to White patients with squamous cell carcinoma is largely due to differences in treatment regimens between the two groups. As shown in Table 5, the difference is that Black patients received significantly less surgical treatment compared to White patients (57.26% vs. 76.94%, p < 0.0001); this finding is consistent with several previously published studies on racial disparities in survival from squamous cell carcinoma of the head and neck.^2–4 However, because the hazard ratio is marginally statistically significant in model 4 (Table 12), it remains possible that other factors beyond those we controlled contributed to the observed survival disparity.

Table 5.

Characteristics for Cases of Squamous Cell Carcinoma Stratified by Race

	Race/Ethnicity
Variables	White	Black	Hispanic	P Value
Total (N=2,125)	1,863	117	145
Age at Diagnosis (yrs) 3				< 0.0001 1
Mean ± S.D.	72.00 ± 13.28	61.40 ± 15.11	67.99 ± 14.94
Median (I.Q.R)	74 (63 to 82)	62 (50 to 72)	70 (60 to 77)
Gender (%) 3				0.0004
Male	78.26	63.25	72.41
Female	21.74	36.75	27.59
Tumor Grade (%)				0.244
Low Grade	32.85	23.08	35.17
High Grade	40.69	47.86	40.00
Unknown	26.46	29.06	24.83
Tumor Size at Diagnosis (mm) 3 (N=1,520)				< 0.0001 1
Mean ± S.D.	35.26 ± 19.73	56.36 ± 102.44	43.32 ± 28.11
Median (I.Q.R)	30 (20 to 45)	45 (30 to 60)	38 (20 to 58)
Lymph Nodes (%) 3 (N=1,785)				0.016
No Lymph Node Involvement	51.09	36.89	53.33
Positive Nodes	48.91	63.11	46.67
Radiation (%) 3 (N=2,060)				0.849
Yes	64.25	66.37	62.94
No	35.75	33.63	37.06
Surgery (%) 3 (N=2,109)
Yes	76.94	57.26	74.48	< 0.0001
No	23.06	42.74	25.52
Extension (%) (N=1,875)
Intraglandular	42.08	35.24	37.50	0.162 2
Extraglandular	54.38	57.14	57.81
Metastatic	3.53	7.62	4.69
Diagnosis Year (%)				0.016
1988 to 1995	14.92	14.53	15.17
1996 to 2003	35.53	37.61	22.07
2004 to 2010	49.54	47.86	62.76

¹P values were based on results of Kruskal-Wallis test.

²P values were based on results of Fisher Exact test.

³For continuous and binary categorical variables with a statistically significant difference on race, the results of multiple comparisons between race were summarized as White versus Hispanic (W/H) or/and White versus Black (W/B) or/and Hispanic versus Black (H/B) with p values < 0.05. For age at diagnosis, the differences were W/H, W/B, and H/B. For tumor size at diagnosis, the difference was W/B. For gender, the differences were W/H and W/B. For lymph nodes, the differences were W/B and H/B. For surgery, the differences were W/B and H/B, and the difference was W/H.

S.D. = standard deviation; I.Q.R. = interquartile range

The recent study by Cheung and colleagues provides further insight into possible sources of the poorer disease-specific survival for Black patients demonstrated in our study.¹⁴ The dataset in their study was obtained from the Florida Cancer Data System (FCDS), which contains chemotherapy data in addition to surgery and radiation treatment data; in their study, the FCDS was also linked to the Florida Agency for Health Care Administration to obtain data on patient comorbidities, and ultimately, through the U.S. Census Bureau, data on community poverty levels for the patients. Therefore, while Cheung and colleagues found that race and ethnicity had no effect on survival, unlike our study, they were able to control for patient comorbidities, poverty level, and chemotherapy treatment, which implies that the source of survival disparity for Black patients compared to White patients seen in our study could be due to differences in chemotherapy treatment, comorbidities, and/or poverty levels. While this comparison is helpful for potentially directing future research, it is not conclusive, as their study and ours are not directly comparable—their study evaluated overall survival while ours evaluated disease-specific survival, a key distinction. It would be expected that comorbidities would have a much greater impact on overall survival than on disease-specific survival, with an important exception being those cases where the presence of certain comorbidities influenced treatment decisions, which would then allow comorbidities to indirectly influence disease-specific survival. The larger dataset in our study also provides increased power to detect smaller differences.

Tumor genetic factors are another potential source for the poorer disease-specific survival for Black compared to White patients in our study. Within the last two years, it has been shown that overexpression and increased gene copy number of EGFR and HER2 are correlated with high-grade malignancy and unfavorable survival for patients with SGCA.^{16, 17} Additionally, gain or loss of MET gene copies and PTEN deletions correlate strongly with high-grade malignancy, increased rate of lymph node metastases, and poorer overall survival in SGCA.^{18, 19} PTEN deletions are also strongly correlated with increased EGFR and HER2 expression, and are potentially responsible for the poor response that has been observed when treating patients with SGCA with increased EGFR and HER2 expression with anti-EGFR and anti-HER2 therapies.¹⁹ We have previously shown that the disparity in survival for Black compared to White patients with squamous cell carcinoma of the head and neck can be explained entirely by greater human papilloma virus (HPV) negative disease in the oropharynx for Black compared to White patients.⁵ It is therefore tempting to speculate that genetic variations in EGFR, HER2, MET, and PTEN in SGCA between Black and White populations could explain at least some of the survival disparities we observed in this study; further research is certainly warranted in this area.

Finally, environmental exposures are also a possible source for the poorer disease-specific survival for Black patients in our study. Looking at differences in occupational exposures between Black and White patients with SGCA, Wilson and colleagues performed a case-control study in 2004 of 2,505 patients to evaluate possible risk factors for death from SGCA.²⁰ Some of their notable findings included increased risk of death from SGCA for Black female cooks (OR: 6.0; 95% CI: 1.47–24.13), Black male janitors (OR: 2.2; 95% CI: 1.01–4.61), White female food service supervisors (OR: 6.7; 95% CI 1.86–23.79), and White male architects (OR: 5.2; 95% CI: 1.38–19.60). Interestingly, in contrast to Black male janitors, White male janitors experienced decreased risk of death from SGCA (OR: 0.6; 95% CI: 0.41–0.87). Looking closer at the differences between Black and White male janitors, the authors found that almost half (49%) of Black male janitors were employed in manufacturing industries, whereas almost all (97%) of White male janitors were employed in industries related to professional services and public administration, suggesting that differences in occupational exposures exist according to race, which could translate into differences in SGCA survival according to race.

This study has several important limitations that must be considered when interpreting the results. While the SEER database provides the opportunity to study even rare diseases at the population level, it lacks data on chemotherapy, has a limited description of the character of surgical and radiation treatments, and includes data from many institutions. Therefore, our results could be biased by different racial and ethnic groups attending different institutions and/or receiving differing types or qualities of treatment that are coded as identical. Furthermore, the SEER database does not contain data on comorbidities, occupation, or socioeconomic status.

In summary, our results show that for patients diagnosed with SGCA, Black race is a risk factor for poorer disease-specific survival for those with mucoepidermoid or squamous cell carcinoma, while Hispanic ethnicity has no effect on disease-specific survival for any SGCA histology. Differences in treatment regimens between Black and White patients—particularly less surgical treatment for Black patients—plays a significant role in the disparity in squamous cell SGCA survival for Black compared to White patients, but not in the disparity in mucoepidermoid SGCA survival. Differences in chemotherapy treatment, comorbidities, socioeconomic status, tumor genetic factors, and environmental exposures are potential but unproven additional sources of the racial survival disparities for mucoepidermoid and squamous cell SGCA, indicating the need for further research into each of these areas.

Table 3.

Characteristics for Cases of Adenocarcinoma Stratified by Race

	Race/Ethnicity
Variables	White	Black	Hispanic	P Value
Total (N=1,632)	1,348	157	127
Age at Diagnosis (yrs) 3				< 0.0001 1
Mean ± S.D.	65.48 ± 14.82	58.96 ± 15.71	56.62 ± 17.73
Median (I.Q.R)	67 (56 to 77)	59 (51 to 70)	56 (43 to 71)
Gender (%)				0.722
Male	61.72	59.24	59.06
Female	38.28	40.76	40.94
Tumor Grade (%)				0.097
Low Grade	26.34	35.03	29.13
High Grade	43.55	28.03	36.45
Unknown	30.12	36.94	36.22
Tumor Size at Diagnosis (mm) 3 (N=1,288)				0.02 1
Mean ± S.D.	29.83 ± 19.97	34.38 ± 22.06	33.73 ± 19.83
Median (I.Q.R)	25 (18 to 35)	30 (20 to 45)	30 (20 to 40)
Lymph Nodes (%) (N=1,400)				0.869
No Lymph Node Involvement	57.07	56.49	59.48
Positive Nodes	42.93	43.51	40.52
Radiation (%) (N=1,599)				0.207
Yes	59.82	55.63	66.13
No	40.18	44.37	33.87
Surgery (%) 3 (N=1,625)				0.0003
Yes	86.51	74.52	85.71
No	13.49	25.48	14.29
Extension (%) (N=1,526)				0.504 2
Intraglandular	53.85	48.63	51.24
Extraglandular	40.59	43.15	41.32
Metastatic	5.56	8.22	7.44
Diagnosis Year (%)				0.378
1988 to 1995	18.47	15.92	15.75
1996 to 2003	38.72	36.94	33.07
2004 to 2010	42.80	47.13	51.18

¹P values were based on results of Kruskal-Wallis test.

²P values were based on results of Fisher Exact test.

³For continuous and binary categorical variables with a statistically significant difference on race, the results of multiple comparisons between race were summarized as White versus Hispanic (W/H) or/and White versus Black (W/B) or/and Hispanic versus Black (H/B) with p values < 0.05. For age at diagnosis, the differences were W/H and W/B. For surgery, the differences were W/B and H/B.

S.D. = standard deviation; I.Q.R. = interquartile range

Table 4.

Characteristics for Cases of Adenoid Cystic Carcinoma Stratified by Race

	Race/Ethnicity
Variables	White	Black	Hispanic	P Value
Total (N=1,278)	977	154	147
Age at Diagnosis (yrs) 3				< 0.0001 1
Mean ± S.D.	56.56 ± 15.82	52.77 ± 15.02	48.10 ± 16.18
Median (I.Q.R)	56 (46 to 69)	51 (42 to 63)	46 (37 to 59)
Gender (%) 3				0.011
Male	42.89	32.47	34.01
Female	57.11	67.53	65.99
Tumor Grade (%)				0.187
Low Grade	14.53	13.64	16.33
High Grade	9.42	9.74	3.40
Unknown	76.05	76.62	80.27
Tumor Size at Diagnosis (mm) 3 (N=1,083)				0.0002 1
Mean ± S.D.	26.77 ± 17.99	30.03 ± 16.49	34.06 ± 25.40
Median (I.Q.R)	23 (15 to 34)	26 (16 to 40)	30 (20 to 40)
Lymph Nodes (%) (N=1,120)				0.750
No Lymph Node Involvement	82.50	85.07	82.17
Positive Nodes	17.50	14.93	17.83
Radiation (%) (N=1,242)				0.219
Yes	74.74	68.46	71.33
No	25.26	31.54	28.67
Surgery (%) (N=1,274)				0.867
Yes	92.29	91.56	93.20
No	7.71	8.44	6.80
Extension (%) (N=1,214)				0.624 2
Intraglandular	54.97	59.86	53.90
Extraglandular	42.44	36.73	42.55
Metastatic	2.59	3.40	3.55
Diagnosis Year (%)				0.223
1988 to 1995	20.47	22.08	14.97
1996 to 2003	36.64	31.17	34.01
2004 to 2010	42.89	46.75	5105

¹P values were based on results of Kruskal-Wallis test.

²P values were based on results of Fisher Exact test.

³For continuous and binary categorical variables with the statistically significant difference on race, the results of multiple comparisons between race were summarized as White versus Hispanic (W/H) or/and White versus Black (W/B) or/and Hispanic versus Black (H/B) with p values < 0.05. For age at diagnosis, the differences were W/H, W/B, and H/B. For tumor size at diagnosis, the difference was W/H. For gender, the differences were W/H and W/B.

S.D. = standard deviation; I.Q.R. = interquartile range