Abstract
Background:
Gastrointestinal (GI) cancers represent a diverse group of diseases. We assessed differences in geographic and racial disparities in cancer-specific mortality across subtypes, overall and by patient characteristics, in a geographically and racially diverse US population.
Methods:
Clinical, sociodemographic, and treatment characteristics for patients diagnosed during 2009–2014 with colorectal cancer (CRC), pancreatic cancer, hepatocellular carcinoma (HCC), or gastric cancer in Georgia were obtained from the Surveillance, Epidemiology, and End Results Program database. Patients were classified by geography (rural or urban county) and race and followed for cancer-specific death. Multivariable Cox proportional hazards models were used to calculate stratified hazard ratios (HR) and 95% confidence intervals (CIs) for associations between geography or race and cancer-specific mortality.
Results:
Overall, 77% of the study population resided in urban counties and 33% were non-Hispanic Black (NHB). For all subtypes, NHB patients were more likely to reside in urban counties than non-Hispanic White patients. Residing in a rural county was associated with an overall increased hazard of cancer-specific mortality for HCC (HR=1.15, 95% CI=1.02–1.31), pancreatic (HR=1.11, 95% CI=1.03–1.19), and gastric cancer (HR=1.17, 95% CI=1.03–1.32) but near-null for CRC. Overall racial disparities were observed for CRC (HR=1.18, 95% CI=1.11–1.25) and HCC (HR=1.12, 95% CI=1.01–1.24). Geographic disparities were most pronounced among HCC patients receiving surgery. Racial disparities were pronounced among CRC patients receiving any treatment.
Conclusion:
Geographic disparities were observed for the rarer GI cancer subtypes, and racial disparities were pronounced for CRC. Treatment factors appear to largely drive both disparities.
Keywords: gastrointestinal cancer mortality, racial disparities, geographic disparities, cancer mortality disparities
INTRODUCTION
Cancers of the gastrointestinal (GI) system represent a diverse range of diseases with different etiologies, and overlapping, nonspecific symptomology.1,2 Colon and rectal cancer (CRC) is detectable through regular screening, preventable through removal of precancerous lesions and dietary and lifestyle changes, and has 90% 5-year survival if diagnosed early.3 In contrast, pancreatic cancer has largely unknown etiology, is often diagnosed at late stages, and has only 37% 5-year survival if diagnosed early.4,5 Following CRC and pancreatic cancer, the most commonly diagnosed GI malignancies are liver and gastric cancer.5
While geographic heterogeneity in GI cancer mortality has been previously reported,6–9 the operationalization of geography and the unit of spatial area have been inconsistent, making it difficult to compare findings across studies and cancer subtypes. Racial disparities in GI cancers have also been separately documented in various populations across subtypes.6,9–19 However, these disparities have not been examined together in the same population both across subtype and within strata of patient characteristics. Understanding how geographic and racial disparities are distributed by patient characteristics will identify patient groups in which inequities are pronounced and where interventions should therefore be targeted. Comparison of race and geographic disparities across GI subtypes, each with different screening opportunities, treatment protocols, and subspecialty requirements, may provide insight into underlying causes of disparities.
The rural underserved and racial minority groups are both recognized as populations that experience adverse health outcomes when compared with their counterparts.20 However, these factors are not modifiable and therefore reducing geographic or racial disparities requires innovative, and often complex interventions. Geographic and racial disparities may reflect differences in socioeconomic status, access to quality or timely care, presence of comorbid conditions, and even health provider training or health literacy levels.21,22 Given complexities associated with preventing, diagnosing, and treating cancer, rural communities may be particularly at risk due to lack of resources and specialized facilities and other structural barriers (e.g., transportation).23–25 In addition, racial disparities result from upstream societal factors, such as persistent discrimination and inequitable housing or education opportunities.26–28
The primary aims of this analysis were to assess geographic and racial disparities in cancer-specific mortality for the four most common GI cancers in a well-defined population with both geographic and racial heterogeneity, and to examine clinical, sociodemographic, and treatment factors associated with these disparities. We also investigated the interaction between geography and race among CRC patients. Georgia is an ideal place to study both geographic and racial disparities because it has a large and diverse population and setting—approximately 25% of residents live in rural areas29 and nearly one-third identify as non-Hispanic Black (NHB).30
METHODS
Data source and variable definitions
Demographic and clinical information for NHB and non-Hispanic White (NHW) men and women 18 years or older diagnosed in Georgia during 2009–2014 with a first primary cancer of the colon and rectum, pancreas, stomach or liver and intrahepatic bile duct according to the site recode ICD-O-3 variable was obtained from the Surveillance, Epidemiology, and End Results Program (SEER) database.31 SEER is a population-based registry program of the National Cancer Institute (NCI) with about 35% coverage in the U.S. Three non-overlapping registries (Atlanta, rural Georgia, and greater Georgia) cover the entire state of Georgia.32 Death certificate or autopsy only cases (1.9% of eligible cases) and cases with incomplete follow-up dates (0.1%) were excluded. CRC, pancreatic cancer, and gastric cancer were identified by corresponding ICD-O-3 codes. Liver cancers were restricted to HCC cases (the most common subtype), identified by ICD-O-3 site code in combination with HCC histology. Liver and intrahepatic bile duct cases with alternative histologies (22%) were excluded.
Sociodemographic characteristics included race (NHB or NHW), sex (male or female), insurance (any Medicaid, insured, or uninsured), marital status (single, married/domestic partner, divorced/separated, or widowed), and county-level socioeconomic status (SES) indices for the percent of families below poverty and the percent of adults 25 years and older with less than high school education (0–<20% or 20–50%). These two SES variables, available in the SEER database, were derived from the 5-year American Community Survey33 datafile centered on year of diagnosis for each patient.
Clinical and treatment characteristics included age at diagnosis (18–49, 50–59, 60–69, or 70+ years), summary stage (localized/regional or distant/nodes involved), tumor size (<2, 2–<5, or 5+ cm), surgery of the primary site (yes or no), and lymph node removal/biopsy (yes or no). The summary stage variable is available for the majority of patients and allows for broad comparisons across subtypes. Surgery of the primary site was determined by site-specific surgical codes34 in combination with clinical consultation (MCR). Specifically, for patients indicated parenthetically, codes corresponding to local tumor excision (CRC, pancreatic, and regional or distant gastric), local tumor destruction (CRC and regional gastric) or surgery not otherwise specified (CRC, pancreatic, and gastric) were classified as no surgery. Additional first course treatment information was obtained from the SEER radiation/chemotherapy database, in which receipt is categorized as yes or no/unknown. Although this data is limited by inability to distinguish between patients truly not receiving treatment and those with unknown status, overall positive predictive values were >90% for GI cancers examined in a large study.35
Geographic assessment
Patients were classified as residing in an urban or rural county at diagnosis according to 2013 rural-urban continuum (RUC) codes.36 Counties with codes of 1–3 were designated as urban (metro) and counties with codes of 4–9 were designated as rural (non-metro). Codes with corresponding descriptions are available in Supplemental Table 1. A map of the distribution of RUC codes across Georgia counties is shown in Supplemental Figure 1.
Outcome assessment
Cancer-specific death was determined from the SEER cause-specific death classification variable which aims to account for misclassification by including causes of death that can probably be attributed to initial cancer diagnosis but which were not coded as such.37 Deaths from other causes were censored. Follow-up time was defined as the number of months from date of diagnosis to date of death or study cutoff date (31 December 2016). A sensitivity analysis was performed in which outcome status was alternatively determined by the cause of death (COD) recode variable that provides specific COD based on ICD codes. For this analysis, only patients with a COD corresponding to the same malignancy as initially diagnosed were considered to have the outcome. All others were censored.
Statistical analysis
Descriptive statistics were summarized by geography and race, separately for each cancer, as frequency and percent for categorical variables or median and interquartile range (IQR) for continuous variables. Median follow-up for all patients was calculated using reverse Kaplan-Meier methods, in which events (deaths) are censored.38 Cox proportional hazards models were used to calculate hazard ratios (HR) and corresponding 95% confidence intervals (CIs) associating geography or race with cancer-specific mortality. For each characteristic, an interaction term between exposure and the characteristic were included in the models and stratum-specific effect estimates were reported. The proportional hazards assumption was assessed separately for geography and race using a graphical approach and time-dependent variables by inclusion of an interaction term between exposure and survival time. No gross violations were observed. However, in ln(-ln) plots for pancreatic cancer, the urban and rural curves converged at the end of follow-up (>50 months). All models were adjusted for continuous age at diagnosis. Multivariable models were adjusted for potential confounders, identified a priori through previous literature and graphical representation of relationships between variables.39,40 Multivariable-adjusted models for geography included race, marital status, sex, and stage, while multivariable-adjusted models for race included sex and stage. While not a component of the minimally sufficient adjustment set for either exposure, stage was included in multivariable models because it is a strong predictor of cancer-specific mortality. Observations with missing data for any of the adjustment variables (e.g., stage) were excluded. Additive interaction between geography and race among CRC only was assessed by calculating the relative excess risk due to interaction (RERI),41 adjusting for continuous age at diagnosis. Analyses were conducted using SAS version 9.4 (Cary, NC).
RESULTS
A total of 29,558 patients diagnosed between 2009 and 2014 in Georgia were identified, including 18,782 CRC, 2,618 HCC, 5,455 pancreatic cancer, and 2,703 gastric cancer patients (Table 1). The number of corresponding cancer-specific deaths were 6,043, 1,829, 4,568, and 1,605, respectively. More than three-quarters of the study population resided in urban counties at diagnosis, ranging from 81% (HCC) to 76% (CRC). Overall, and for all cancer subtypes individually, NHB patients were more likely to reside in urban counties than NHW patients (82% vs 74%). Median total follow-up ranged from 52 (gastric cancer, rural) to 42 months (HCC, urban), with the greatest urban-rural difference observed among gastric cancer patients.
Table 1.
Patient clinical, sociodemographic, and treatment characteristics among non-Hispanic White and non-Hispanic Black males and females 18+ years diagnosed with colorectal cancer, hepatocellular carcinoma, pancreatic cancer, or gastric cancer in Georgia 2009–2014 and registered in the Surveillance, Epidemiology, and End Results (SEER) Program (N=29,558) by cancer site and geography of county of residence at diagnosis (urban or rural)
|
Colorectal Cancer (n=18782) |
Hepatocellular
Carcinoma (n=2618) |
Pancreatic Cancer (n=5455) |
Gastric Cancer (n=2703) |
|||||
|---|---|---|---|---|---|---|---|---|
| Urban (n=14299) |
Rural (n=4483) |
Urban (n=2133) |
Rural (n=485) |
Urban (n=4230) |
Rural (n=1225) |
Urban (n=2065) |
Rural (n=638) |
|
| Median (IQR) |
Median (IQR) |
Median (IQR) |
Median (IQR) |
Median (IQR) |
Median (IQR) |
Median (IQR) |
Median (IQR) |
|
| Age at Diagnosis (years) | 63 (54, 73) | 65 (56, 74) | 60 (55, 66) | 62 (56, 70) | 67 (58, 76) | 68 (59, 77) | 65 (56, 75) | 68 (59, 76) |
| Length of Follow-up (months)* | 51 (32, 72) | 50 (32, 71) | 42 (27, 66) | 46 (33, 63) | 47 (31, 66) | 43 (29, 68) | 47 (30, 70) | 52 (31, 73) |
| Time to Death (months) | 13 (3, 27) | 12 (3, 26) | 5 (1, 13) | 5 (1, 15) | 4 (1, 11) | 3 (1, 9) | 6 (2, 14) | 5 (1, 14) |
| N (%) | N (%) | N (%) | N (%) | N (%) | N (%) | N (%) | N (%) | |
| Clinical Characteristics | ||||||||
| Age at Diagnosis | ||||||||
| 18–49 years | 2159 (15) | 491 (11) | 150 (7.0) | 30 (6.2) | 339 (8.0) | 85 (6.9) | 251 (12) | 63 (9.9) |
| 50–59 years | 3550 (25) | 1087 (24) | 853 (40) | 155 (32) | 853 (20) | 232 (19) | 449 (22) | 918 (17) |
| 60–69 years | 3881 (27) | 1312 (29) | 733 (34) | 178 (37) | 1267 (30) | 371 (30) | 560 (27) | 184 (29) |
| 70+ years | 4709 (33) | 1593 (36) | 397 (19) | 122 (25) | 1771 (42) | 537 (44) | 805 (39) | 282 (44) |
| Year of Diagnosis | ||||||||
| 2009–2011 | 6965 (49) | 2188 (49) | 915 (43) | 223 (46) | 1960 (46) | 565 (46) | 1016 (49) | 327 (51) |
| 2012–2014 | 7334 (51) | 2295 (51) | 1218 (57) | 262 (54) | 2270 (54) | 660 (54) | 1049 (51) | 311 (49) |
| Summary Stage | ||||||||
| Localized/regional | 10463 (73) | 3237 (72) | 1645 (77) | 354 (73) | 1698 (40) | 454 (37) | 1117 (54) | 352 (55) |
| Distant sites/nodes involved | 3289 (23) | 1062 (24) | 346 (16) | 96 (20) | 2255 (53) | 677 (55) | 771 (37) | 222 (35) |
| Unknown/unspecified | 547 (3.8) | 184 (4.1) | 142 (6.7) | 35 (7.2) | 277 (6.6) | 94 (7.7) | 177 (8.6) | 64 (10) |
| Tumor Size | ||||||||
| <2 cm | 1818 (13) | 498 (11) | 229 (11) | 37 (7.6) | 241 (5.7) | 65 (5.3) | 207 (10) | 79 (12) |
| 2–<5 cm | 5038 (35) | 1624 (36) | 752 (35) | 165 (34) | 2141 (51) | 569 (46) | 485 (23) | 145 (23) |
| 5+ cm | 4631 (32) | 1402 (31) | 795 (37) | 175 (36) | 884 (21) | 259 (21) | 511 (25) | 157 (25) |
| Unknown | 2812 (20) | 959 (21) | 357 (17) | 108 (22) | 964 (23) | 332 (27) | 862 (42) | 257 (40) |
| Sociodemographic Characteristics | ||||||||
| Race/Ethnicity | ||||||||
| Non-Hispanic White | 9399 (66) | 3341 (75) | 1291 (61) | 375 (77) | 2828 (67) | 956 (78) | 1154 (56) | 413 (65) |
| Non-Hispanic Black | 4900 (34) | 1142 (25) | 842 (39) | 110 (23) | 1402 (33) | 269 (22) | 911 (44) | 225 (35) |
| Sex | ||||||||
| Female | 6884 (48) | 2088 (47) | 426 (20) | 99 (20) | 2133 (50) | 570 (47) | 807 (39) | 256 (40) |
| Male | 7415 (52) | 2395 (53) | 1707 (80) | 386 (80) | 2097 (50) | 655 (53) | 1258 (61) | 382 (60) |
| Insurance Type** | ||||||||
| Any Medicaid | 136 (9.68) | 635 (14) | 435 (20) | 108 (22) | 415 (9.8) | 170 (14) | 227 (11) | 106 (17) |
| Other insurance | 11358 (79) | 3321 (74) | 1419 (67) | 327 (67) | 3411 (81) | 915 (75) | 1621 (79) | 471 (74) |
| Uninsured | 972 (6.8) | 375 (8.4) | 198 (9.3) | 42 (8.7) | 234 (5.5) | 86 (7.0) | 123 (6.0) | 39 (6.1) |
| Marital Status | ||||||||
| Single | 2405 (17) | 668 (15) | 485 (23) | 91 (19) | 627 (15) | 144 (12) | 355 (17) | 95 (15) |
| Married or domestic partner | 7187 (50) | 2321 (52) | 950 (45) | 257 (53) | 2088 (49) | 631 (52) | 1057 (51) | 314 (49) |
| Divorced or separated | 1788 (13) | 462 (10) | 400 (19) | 69 (14) | 568 (13) | 149 (12) | 219 (11) | 74 (12) |
| Widowed | 2020 (14) | 736 (16) | 194 (9.1) | 52 (11) | 749 (18) | 251 (20) | 303 (15) | 120 (19) |
| Unknown | 899 (6.3) | 296 (6.6) | 104 (4.9) | 16 (3.3) | 198 (4.7) | 50 (4.1) | 131 (6.3) | 35 (5.5) |
| SES Index – families below poverty*** | ||||||||
| 0%–<20% | 12927 (90) | 2967 (66) | 1895 (89) | 329 (68) | 3868 (91) | 814 (66) | 1852 (90) | 420 (66) |
| 20%–50% | 1372 (9.6) | 1516 (34) | 238 (11) | 156 (32) | 362 (8.6) | 411 (34) | 213 (10) | 218 (34) |
| SES Index - less than high school education*** | ||||||||
| 0%–<20% | 12144 (85) | 1556 (35) | 1808 (85) | 179 (37) | 3618 (86) | 460 (38) | 1766 (86) | 228 (36) |
| 20%–50% | 2155 (15) | 2927 (65) | 325 (15) | 306 (63) | 612 (14) | 765 (62) | 299 (14) | 410 (64) |
| Treatment Characteristics | ||||||||
| Surgery of the Primary Site** | ||||||||
| Yes | 10702 (75) | 3307 (74) | 404 (19) | 80 (16) | 802 (19) | 183 (15) | 892 (43) | 270 (42) |
| No | 3575 (25) | 1164 (26) | 1726 (81) | 405 (84) | 3419 (81) | 1034 (84) | 1171 (57) | 366 (57) |
| Lymph Node Removal or Biopsy** | ||||||||
| Yes | 10369 (73) | 3201 (71) | 122 (5.7) | 23 (4.7) | 829 (20) | 194 (16) | 679 (33) | 213 (33) |
| No | 3847 (27) | 1228 (27) | 2008 (94) | 460 (95) | 3374 (80) | 1020 (83) | 1371 (66) | 420 (66) |
| Radiation | ||||||||
| Yes | 1967 (14) | 702 (16) | 172 (8.1) | 45 (9.3) | 629 (15) | 165 (13) | 455 (22) | 149 (23) |
| No/unknown | 12332 (86) | 3781 (84) | 1961 (92) | 440 (91) | 3601 (85) | 1060 (87) | 1610 (78) | 489 (77) |
| Chemotherapy | ||||||||
| Yes | 6001 (42) | 1882 (42) | 941 (44) | 203 (42) | 2174 (51) | 530 (43) | 1028 (50) | 264 (41) |
| No/unknown | 8298 (58) | 2601 (58) | 1192 (56) | 282 (58) | 2056 (49) | 695 (57) | 1037 (50) | 374 (59) |
Calculated using reverse Kaplan-Meier method, where events (deaths) are censored
Frequencies and percents may not sum to total due to missing data and small cell counts
SES index is an attribute of the county of residence at diagnosis and is coded using the American Community Survey (ACS) file centered around the year of diagnosis (e.g., 2008–2012 ACS file used for cases diagnosed in 2010)
Geographic disparities
Patient characteristics by geography are provided in Table 1. Differences between urban- and rural-residing patients were similar across cancers. Briefly, for all cancer subtypes, patients in urban counties were more likely to be younger, single, and have localized disease (except gastric cancer). Patients in rural counties were more likely to be older at diagnosis, have Medicaid, and be widowed. Across all cancer subtypes, the majority (85–91%) living in urban areas resided in counties with low poverty and high educational attainment. Overall, about 56% of both urban and rural patients received surgery, but proportions varied by cancer subtype (e.g., among HCC: 19% vs 16%).
Overall geographic disparities in cancer-specific mortality were observed in multivariable-adjusted models for HCC (HR=1.15, 95% CI=1.02–1.31), pancreatic cancer (HR=1.11, 95% CI=1.03–1.19), and gastric cancer (HR=1.17, 95% CI=1.03–1.32) but were near-null for CRC (HR=1.05, 95% CI=0.98–1.11) (Figure 1). Associations stratified by clinical, sociodemographic, and treatment characteristics are available in Table 2. For both pancreatic and gastric cancer, the disparity was most pronounced among the youngest age group and decreased with increasing age. Among gastric cancer only, the disparity was greater among males (HR=1.27, 95% CI=1.10–1.48) than females (HR=1.00, 95% CI=0.81–1.23). For HCC, the disparity was most pronounced among patients receiving surgery (HR=1.58, 95% CI=1.09–2.31) or lymph node removal/biopsy (HR=2.83, 95% CI=1.37–5.88). Rurality was also associated with increased mortality among CRC patients receiving radiation and HCC patients receiving chemotherapy. Age-adjusted models are available in Supplemental Table 2. Most associations were similar to multivariable-adjusted results but some gastric cancer associations were attenuated.
Figure 1.
Overall geographic and racial disparities by gastrointestinal cancer subtype
Table 2.
Multivariable Cox proportional hazard ratios for the association between geography of county of residence at diagnosis (rural vs urban) and cause-specific mortality by patient clinical, sociodemographic, and treatment characteristics among non-Hispanic White and non-Hispanic Black males and females 18+ years diagnosed with colorectal cancer, hepatocellular carcinoma, pancreatic cancer, or gastric cancer in Georgia 2009–2014 and registered in the Surveillance, Epidemiology, and End Results (SEER) Program by cancer site
| Colorectal Cancer | Hepatocellular Carcinoma | Pancreatic Cancer | Gastric Cancer | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No. of deaths | Stratified Effects* HR (95% CI) | No. of deaths | Stratified Effects* HR (95% CI) | No. of deaths | Stratified Effects* HR (95% CI) | No. of deaths | Stratified Effects* HR (95% CI) | |||||
| Urban | Rural | Rural vs Urban (ref) | Urban | Rural | Rural vs Urban (ref) | Urban | Rural | Rural vs Urban (ref) | Urban | Rural | Rural vs Urban (ref) | |
| Clinical Characteristics | ||||||||||||
| Age at Diagnosis | ||||||||||||
| 18–49 years | 602 | 158 | 1.12 (0.94, 1.33) | 89 | 16 | 1.07 (0.63, 1.83) | 215 | 63 | 1.51 (1.14, 2.00) | 115 | 36 | 1.80 (1.24, 2.62) |
| 50–59 years | 876 | 296 | 1.13 (0.99, 1.29) | 511 | 102 | 1.09 (0.88, 1.35) | 622 | 181 | 1.28 (1.08, 1.51) | 233 | 58 | 1.14 (0.86, 1.52) |
| 60–69 years | 1030 | 357 | 1.05 (0.93, 1.18) | 423 | 118 | 1.27 (1.04, 1.56) | 997 | 296 | 1.05 (0.92, 1.20) | 294 | 95 | 1.16 (0.92, 1.47) |
| 70+ years | 1603 | 522 | 1.00 (0.91, 1.11) | 257 | 87 | 1.12 (0.88, 1.43) | 1327 | 396 | 1.04 (0.93, 1.16) | 419 | 157 | 1.09 (0.91, 1.31) |
| Summary Stage | ||||||||||||
| Localized/Regional | 1634 | 541 | 1.06 (0.96, 1.17) | 987 | 243 | 1.21 (1.05, 1.40) | 1195 | 344 | 1.15 (1.02, 1.30) | 441 | 151 | 1.11 (0.92, 1.33) |
| Distant sites/nodes involved | 2477 | 792 | 1.04 (0.96, 1.12) | 293 | 80 | 1.00 (0.78, 1.28) | 1966 | 592 | 1.09 (0.99, 1.19) | 620 | 195 | 1.22 (1.04, 1.44) |
| Tumor Size (cm) | ||||||||||||
| <2 cm | 124 | 39 | 1.07 (0.74, 1.53) | 85 | 14 | NR | 137 | 42 | 1.07 (0.76, 1.51) | 40 | 22 | 1.42 (0.84, 2.40) |
| 2–<5 cm | 1251 | 411 | 1.04 (0.93, 1.16) | 405 | 101 | 1.33 (1.07, 1.65) | 1670 | 471 | 1.14 (1.02, 1.26) | 233 | 76 | 1.19 (0.92, 1.54) |
| 5+ cm | 1694 | 520 | 1.07 (0.97, 1.19) | 590 | 136 | 0.97 (0.80, 1.17) | 718 | 207 | 1.19 (1.02, 1.39) | 266 | 93 | 1.31 (1.04, 1.66) |
| Sociodemographic Characteristics | ||||||||||||
| Race/Ethnicity | ||||||||||||
| Non-Hispanic White | 2633 | 965 | 1.06 (0.98, 1.14) | 779 | 240 | 1.12 (0.97, 1.30) | 2149 | 737 | 1.09 (1.00, 1.19) | 616 | 230 | 1.18 (1.02, 1.38) |
| Non-Hispanic Black | 1478 | 368 | 1.02 (0.91, 1.15) | 501 | 83 | 1.25 (0.99, 1.58) | 1012 | 199 | 1.18 (1.01, 1.37) | 445 | 116 | 1.15 (0.93, 1.41) |
| Sex | ||||||||||||
| Female | 1909 | 592 | 1.04 (0.94, 1.14) | 241 | 71 | 1.20 (0.92, 1.57) | 1568 | 434 | 1.12 (1.01, 1.25) | 365 | 116 | 1.00 (0.81, 1.23) |
| Male | 2202 | 741 | 1.06 (0.97, 1.15) | 1039 | 252 | 1.14 (0.99, 1.31) | 1593 | 502 | 1.10 (1.00, 1.22) | 696 | 230 | 1.27 (1.10, 1.48) |
| Insurance Type | ||||||||||||
| Any Medicaid | 534 | 252 | 1.07 (0.92, 1.24) | 259 | 66 | 0.95 (0.72, 1.24) | 309 | 133 | 1.08 (0.88, 1.32) | 125 | 57 | 1.28 (0.93, 1.76) |
| Other insurance | 3141 | 912 | 1.00 (0.93, 1.07) | 845 | 222 | 1.21 (1.04, 1.40) | 2608 | 720 | 1.09 (1.00, 1.19) | 848 | 258 | 1.11 (0.96, 1.27) |
| Uninsured | 353 | 150 | 1.08 (0.89, 1.30) | 145 | 33 | 1.26 (0.86, 1.85) | 184 | 69 | 1.09 (0.82, 1.44) | 71 | 29 | 1.45 (0.94, 2.23) |
| Marital Status | ||||||||||||
| Single | 843 | 255 | 1.05 (0.91, 1.21) | 321 | 63 | 1.06 (0.81, 1.39) | 469 | 112 | 1.11 (0.90, 1.36) | 199 | 51 | 1.50 (1.10, 2.04) |
| Married or Domestic Partner | 1918 | 650 | 1.03 (0.95, 1.13) | 565 | 170 | 1.17 (0.99, 1.39) | 1664 | 505 | 1.07 (0.97, 1.18) | 581 | 189 | 1.13 (0.96, 1.33) |
| Divorced or Separated | 621 | 171 | 1.11 (0.93, 1.31) | 269 | 46 | 1.12 (0.81, 1.53) | 454 | 121 | 1.22 (1.00, 1.50) | 118 | 40 | 1.16 (0.81, 1.67) |
| Widowed | 729 | 257 | 1.04 (0.90, 1.20) | 125 | 44 | 1.32 (0.94, 1.87) | 574 | 198 | 1.16 (0.98, 1.36) | 163 | 66 | 1.08 (0.81, 1.43) |
| SES Index – families bel poverty | ||||||||||||
| 0%–<20% | 3735 | 893 | 1.03 (0.95, 1.11) | 1151 | 217 | 1.20 (1.03, 1.39) | 2893 | 615 | 1.12 (1.02, 1.22) | 957 | 228 | 1.14 (0.98, 1.31) |
| 20%–50% | 376 | 440 | 1.11 (0.97, 1.27) | 129 | 106 | 1.17 (0.90, 1.51) | 268 | 321 | 0.99 (0.84, 1.16) | 104 | 118 | 1.27 (0.97, 1.65) |
| SES Index – less than high school education | ||||||||||||
| 0%–<20% | 3440 | 451 | 1.00 (0.91, 1.10) | 1088 | 113 | 1.05 (0.86, 1.27) | 2680 | 360 | 1.18 (1.05, 1.32) | 899 | 127 | 1.16 (0.96, 1.40) |
| 20%–50% | 671 | 882 | 1.02 (0.92, 1.13) | 192 | 210 | 1.25 (1.03, 1.52) | 481 | 576 | 0.93 (0.82, 1.05) | 162 | 219 | 1.13 (0.92, 1.39) |
| Treatment Characteristics | ||||||||||||
| Surgery of the Primary Site | ||||||||||||
| Yes | 2613 | 824 | 1.05 (0.97, 1.14) | 109 | 36 | 1.58 (1.09, 2.31) | 456 | 111 | 1.03 (0.84, 1.27) | 293 | 105 | 1.24 (0.99, 1.55) |
| No | 1497 | 506 | 0.98 (0.89, 1.09) | 1170 | 287 | 1.10 (0.96, 1.25) | 2705 | 822 | 1.11 (1.03, 1.20) | 768 | 240 | 1.14 (0.98, 1.32) |
| Lymph Node Removal or Biopsy | ||||||||||||
| Yes | 2541 | 785 | 1.04 (0.96, 1.13) | 21 | 11 | NR | 504 | 125 | 1.07 (0.88, 1.30) | 296 | 108 | 1.21 (0.97, 1.51) |
| No | 1541 | 527 | 0.99 (0.90, 1.09) | 1258 | 310 | 1.12 (0.98, 1.27) | 2647 | 805 | 1.10 (1.01, 1.19) | 758 | 235 | 1.17 (1.01, 1.35) |
| Radiation | ||||||||||||
| Yes | 583 | 239 | 1.19 (1.02, 1.38) | 123 | 35 | 1.23 (0.84, 1.79) | 485 | 133 | 1.11 (0.92, 1.35) | 288 | 95 | 1.03 (0.81, 1.30) |
| No/unknown | 3528 | 1094 | 1.02 (0.95, 1.09) | 1157 | 288 | 1.15 (1.00, 1.31) | 2676 | 803 | 1.11 (1.02, 1.20) | 773 | 251 | 1.22 (1.06, 1.41) |
| Chemotherapy | ||||||||||||
| Yes | 2284 | 739 | 1.10 (1.02, 1.20) | 586 | 147 | 1.23 (1.02, 1.47) | 1765 | 449 | 1.05 (0.95, 1.17) | 632 | 166 | 1.06 (0.89, 1.26) |
| No/unknown | 1827 | 594 | 0.97 (0.88, 1.06) | 694 | 176 | 1.09 (0.92, 1.28) | 1396 | 487 | 1.12 (1.01, 1.24) | 429 | 180 | 1.27 (1.07, 1.52) |
Excluded patients with unknown summary stage or unknown marital status (total N excluded=3,044) because these variables were included as covariates for all models
Multivariable models adjusted for age at diagnosis (continuous variable), race, marital status, sex, and stage
NR=not reported; HRs with numerators <16 are suppressed
Racial disparities
The distribution of patient characteristics by race are available in Supplemental Table 3. As with geography, distributions were similar across cancer subtypes, with NHB patients more likely to be diagnosed at younger ages, have Medicaid or no insurance, and be single. NHW patients were more likely to receive surgery or lymph node biopsy for all cancer subtypes except gastric cancer. Notably, for HCC, NHW patients were approximately 50% more likely to receive surgery (21% vs 14%), despite minimal difference in stage distribution (77% vs 76% localized disease).
Age-adjusted and multivariable-adjusted associations between race and cancer-specific mortality, stratified by patient characteristics, are available in Supplemental Table 4 and Table 3, respectively. In overall multivariable-adjusted models, racial disparities were observed for CRC (HR=1.18, 95% CI=1.11–1.25) and HCC (HR=1.12, 95% CI=1.01–1.24), but not for pancreatic or gastric cancer (Figure 1). For HCC only, the association was stronger among patients with localized/regional disease (HR=1.16, 95% CI=1.03–1.30) than among those with distant disease (HR=1.00, 95% CI=0.81–1.23). Racial disparities were also observed among CRC patients receiving treatment of any modality and among HCC patients receiving chemotherapy (HR=1.20, 95% CI=1.04–1.39). Decreased mortality for NHB patients was observed among divorced/separated pancreatic (HR=0.82, 95% CI=0.69–0.97) and gastric cancer (HR=0.64, 95% CI=0.47–0.87) patients.
Table 3.
Multivariable Cox proportional hazard ratios for the association between race and cause-specific mortality by patient clinical, sociodemographic, and treatment characteristics among non-Hispanic White and non-Hispanic Black males and females 18+ years diagnosed with colorectal cancer, hepatocellular carcinoma, pancreatic cancer, or gastric cancer in Georgia 2009–2014 and registered in the Surveillance, Epidemiology, and End Results (SEER) Program by cancer site
| Colorectal Cancer | Hepatocellular Carcinoma | Pancreatic Cancer | Gastric Cancer | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No. of deaths | Stratified Effects* HR (95% CI) | No. of deaths | Stratified Effects* HR (95% CI) | No. of deaths | Stratified Effects* HR (95% CI) | No. of deaths | Stratified Effects* HR (95% CI) | |||||
| NHW | NHB | NHB vs NHW (ref) | NHW | NHB | NHB vs NHW (ref) | NHW | NHB | NHB vs NHW (ref) | NHW | NHB | NHB vs NHW (ref) | |
| Clinical Characteristics | ||||||||||||
| Age at Diagnosis | ||||||||||||
| 18–49 years | 457 | 341 | 1.22 (1.06, 1.40) | 65 | 43 | 1.16 (0.78, 1.71) | 190 | 105 | 0.84 (0.66, 1.06) | 71 | 90 | 1.12 (0.82, 1.53) |
| 50–59 years | 703 | 537 | 1.26 (1.12, 1.41) | 384 | 266 | 1.15 (0.99, 1.35) | 540 | 303 | 0.93 (0.81, 1.07) | 150 | 150 | 0.99 (0.79, 1.24) |
| 60–69 years | 950 | 507 | 1.18 (1.06, 1.31) | 334 | 231 | 1.19 (1.01, 1.41) | 928 | 412 | 0.97 (0.86, 1.09) | 252 | 148 | 0.96 (0.78, 1.18) |
| 70+ years | 1651 | 574 | 1.12 (1.02, 1.23) | 288 | 74 | 0.98 (0.76, 1.27) | 1345 | 446 | 1.07 (0.96, 1.19) | 401 | 205 | 0.96 (0.81, 1.14) |
| Summary Stage | ||||||||||||
| Localized/Regional | 1578 | 704 | 1.14 (1.05, 1.25) | 820 | 473 | 1.16 (1.03, 1.30) | 1135 | 464 | 1.04 (0.93, 1.16) | 367 | 251 | 1.01 (0.86, 1.19) |
| Distant sites/nodes involved | 2183 | 1255 | 1.20 (1.12, 1.28) | 251 | 141 | 1.00 (0.81, 1.23) | 1868 | 802 | 0.96 (0.88, 1.04) | 507 | 342 | 0.97 (0.84, 1.11) |
| Tumor Size | ||||||||||||
| <2 cm | 122 | 49 | 0.94 (0.67, 1.30) | 65 | 40 | 1.08 (0.73, 1.61) | 141 | 43 | 1.01 (0.72, 1.43) | 46 | 16 | 0.90 (0.51, 1.60) |
| 2–<5 cm | 1162 | 565 | 1.33 (1.20, 1.47) | 361 | 162 | 1.07 (0.88, 1.28) | 1587 | 637 | 1.00 (0.91, 1.10) | 196 | 124 | 1.07 (0.86, 1.35) |
| 5+ cm | 1514 | 800 | 1.09 (1.00, 1.19) | 470 | 295 | 1.12 (0.96, 1.30) | 665 | 301 | 0.85 (0.75, 0.98) | 216 | 156 | 0.81 (0.66, 1.00) |
| Sociodemographic Characteristics | ||||||||||||
| Sex | ||||||||||||
| Female | 1702 | 927 | 1.11 (1.03, 1.20) | 195 | 132 | 0.94 (0.75, 1.17) | 1408 | 678 | 0.93 (0.85, 1.02) | 285 | 214 | 0.82 (0.69, 0.98) |
| Male | 2059 | 1032 | 1.24 (1.15, 1.34) | 876 | 482 | 1.17 (1.05, 1.31) | 1595 | 588 | 1.06 (0.96, 1.16) | 589 | 379 | 1.09 (0.96, 1.24) |
| Insurance Type | ||||||||||||
| Any Medicaid | 400 | 425 | 1.04 (0.91, 1.19) | 185 | 160 | 1.15 (0.93, 1.42) | 241 | 224 | 0.66 (0.55, 0.79) | 65 | 128 | 1.09 (0.81, 1.47) |
| Other insurance | 2985 | 1224 | 1.14 (1.07, 1.22) | 750 | 356 | 1.11 (0.98, 1.26) | 2528 | 914 | 1.04 (0.96, 1.12) | 739 | 397 | 0.91 (0.80, 1.03) |
| Uninsured | 287 | 239 | 1.02 (0.86, 1.22) | 110 | 75 | 0.75 (0.56, 1.01) | 160 | 99 | 0.72 (0.56, 0.93) | 50 | 52 | 0.81 (0.55, 1.20) |
| Marital Status | ||||||||||||
| Single | 473 | 625 | 1.11 (0.98, 1.25) | 156 | 228 | 1.27 (1.04, 1.56) | 280 | 301 | 0.97 (0.83, 1.14) | 100 | 150 | 1.00 (0.77, 1.29) |
| Married or Domestic Partner | 1885 | 683 | 1.24 (1.13, 1.35) | 548 | 187 | 1.00 (0.85, 1.18) | 1703 | 466 | 1.00 (0.90, 1.11) | 523 | 247 | 1.07 (0.91, 1.24) |
| Divorced or Separated | 508 | 284 | 1.04 (0.90, 1.21) | 189 | 126 | 0.93 (0.74, 1.17) | 365 | 210 | 0.82 (0.69, 0.97) | 82 | 76 | 0.64 (0.47, 0.87) |
| Widowed | 732 | 254 | 0.91 (0.79, 1.05) | 126 | 43 | 0.92 (0.65, 1.30) | 538 | 234 | 0.98 (0.84, 1.14) | 141 | 88 | 0.75 (0.57, 0.98) |
| SES Index – families below poverty | ||||||||||||
| 0%–<20% | 3270 | 1585 | 1.22 (1.15, 1.30) | 936 | 496 | 1.14 (1.02, 1.27) | 2633 | 1021 | 0.96 (0.89, 1.03) | 765 | 470 | 0.96 (0.85, 1.07) |
| 20%–50% | 491 | 374 | 0.98 (0.85, 1.12) | 135 | 118 | 1.06 (0.83, 1.36) | 370 | 245 | 1.10 (0.94, 1.29) | 109 | 123 | 1.13 (0.87, 1.46) |
| SES Index – less than high school education | ||||||||||||
| 0%–<20% | 2582 | 1506 | 1.20 (1.13, 1.28) | 747 | 519 | 1.13 (1.01, 1.26) | 2138 | 1031 | 0.99 (0.92, 1.06) | 606 | 464 | 0.97 (0.86, 1.10) |
| 20%– 50% | 1179 | 453 | 1.14 (1.02, 1.27) | 324 | 95 | 1.15 (0.92, 1.45) | 865 | 235 | 1.06 (0.92, 1.22) | 268 | 129 | 1.07 (0.87, 1.32) |
| County designation | ||||||||||||
| Urban | 2733 | 1570 | 1.20 (1.12, 1.27) | 820 | 530 | 1.12 (1.01, 1.25) | 2240 | 1060 | 0.99 (0.92, 1.06) | 636 | 475 | 1.01 (0.90, 1.14) |
| Rural | 1028 | 389 | 1.13 (1.00, 1.27) | 251 | 84 | 1.20 (0.94, 1.54) | 763 | 206 | 1.05 (0.90, 1.22) | 238 | 118 | 0.93 (0.74, 1.16) |
| Treatment Characteristics | ||||||||||||
| Surgery of the Primary Site | ||||||||||||
| Yes | 2408 | 1173 | 1.22 (1.14, 1.31) | 113 | 37 | 0.95 (0.65, 1.37) | 425 | 151 | 0.92 (0.76, 1.11) | 235 | 175 | 0.95 (0.78, 1.16) |
| No | 1350 | 785 | 0.98 (0.90, 1.07) | 957 | 577 | 0.99 (0.89, 1.10) | 2575 | 1115 | 0.97 (0.90, 1.04) | 639 | 417 | 1.07 (0.95, 1.22) |
| Lymph Node Removal or Biopsy | ||||||||||||
| Yes | 2331 | 1133 | 1.21 (1.12, 1.30) | 30 | NR | NR | 466 | 174 | 0.95 (0.80, 1.13) | 235 | 178 | 0.89 (0.73, 1.09) |
| No | 1391 | 814 | 1.05 (0.97, 1.15) | 1038 | 610 | 1.07 (0.97, 1.18) | 2523 | 1090 | 0.97 (0.90, 1.04) | 632 | 411 | 1.07 (0.94, 1.21) |
| Radiation | ||||||||||||
| Yes | 595 | 265 | 1.31 (1.14, 1.52) | 122 | 43 | 0.77 (0.54, 1.09) | 453 | 189 | 1.00 (0.84, 1.18) | 245 | 145 | 1.18 (0.96, 1.45) |
| No/unknown | 3166 | 1694 | 1.16 (1.09, 1.23) | 949 | 571 | 1.16 (1.04, 1.29) | 2550 | 1077 | 0.98 (0.92, 1.06) | 629 | 448 | 0.94 (0.83, 1.06) |
| Chemotherapy | ||||||||||||
| Yes | 2064 | 1101 | 1.19 (1.10, 1.28) | 460 | 310 | 1.20 (1.04, 1.39) | 1641 | 663 | 1.08 (0.94, 1.24) | 503 | 324 | 0.91 (0.79, 1.05) |
| No/unknown | 1697 | 858 | 1.13 (1.04, 1.23) | 611 | 304 | 1.08 (0.94, 1.24) | 1362 | 603 | 1.00 (0.91, 1.11) | 371 | 269 | 1.08 (0.93, 1.27) |
Excluded patients with missing stage (total N excluded = 1,520)
Multivariable models adjusted for age at diagnosis (continuous), sex, and stage
NR=not reported; cells with counts <5 and HRs with numerators <16 are suppressed
Geography-race interaction
In a post-hoc analysis of the interaction between geography and race among CRC patients overall, we did not observe a departure from multiplicative or additive interaction (RERI=0.018, 95% CI=-0.14–0.18).
Outcome sensitivity analysis
Using the SEER COD recode variable resulted in fewer events and more censored observations. The overall proportion of patients whose outcome status changed varied by cancer subtype from 3.3% (CRC) to 16% (gastric cancer). Proportions did not vary by stage for any subtype except pancreatic cancer. Multivariable results (data not shown) for the association between geography and cancer-specific mortality showed a majority of stratified effects did not meaningfully change. A few results were stronger for HCC (e.g., among NHB HR=1.35, 95% CI=1.04–1.75) and attenuated for gastric cancer (e.g., among never married HR=1.26, 95% CI=0.85–1.85). However, point estimates were contained within the 95% CIs of primary results. Multivariable associations between race and cancer-specific mortality were similar for all sites, except gastric cancer, which were stronger and suggestive of racial disparities among several strata. The overall association suggested a 22% increase in mortality among NHB patients (HR=1.22, 95% CI=1.07–1.37) and a more pronounced disparity among males (HR=1.51, 95% CI=1.29–1.76), patients 18–49 years (HR=1.57, 95% CI=1.08–2.28), and patients receiving treatment. The proportion reclassified varied by race for gastric cancer only (NHW: 20% vs NHB: 11%).
DISCUSSION
This population-based study among patients diagnosed with the four most common GI cancers in Georgia provides a comprehensive comparison of geographic and racial disparities in mortality, suggesting disparities vary by cancer subtype and by patient characteristics within subtype. Stratification by patient characteristics highlighted certain groups for whom disparities are most pronounced, identifying possible targets for intervention or increased surveillance. We found geographic disparities exist overall for less prevalent cancers but not CRC. We found overall racial disparities for CRC and HCC patients.
Geographic disparities were strongest for HCC patients receiving surgery, pancreatic and gastric cancer patients younger than 50 years, and uninsured gastric cancer patients. Racial disparities were pronounced among CRC patients who were younger than 60 years or who received treatment of any modality, and HCC patients living in rural counties or who were diagnosed with localized or regional stage disease. Notably, cancer-specific mortality was lower for NHB than NHW patients among pancreatic cancer patients with Medicaid or no insurance and among divorced/separated gastric cancer patients.
Previous studies have provided some evidence of geographic heterogeneity in cancer outcomes; however, comparisons of the magnitude of the disparity and contributions of patient factors have been limited by differences in study populations and geographic specification. In one large study conducted by the CDC covering 97% of the US population and categorizing geography by RUC codes, more pronounced geographic disparities in death rates were observed for CRC than other GI subtypes,42 conflicting with our findings. However, this study calculated overall death rates in the total population, while our study examined mortality rates among cancer patients. Another large study used the North American Association of Central Cancer Registries database, a nationwide population-based registry, to show geographic heterogeneity in incidence rates—with higher incidence in urban areas for gastric, liver, and pancreatic cancers and higher incidence in rural areas for CRC.43
One explanation for overall higher cause-specific mortality in rural compared to urban counties for rarer GI cancers, but not for CRC, may be that rural areas lack specialized expertise in the diagnosis and treatment of rare cancers. There is only one NCI-designated cancer center in Georgia. This center, along with several other large healthcare systems, are located in urban areas, and thus, access to specialized care could contribute to rural-urban disparities for the rarer GI cancers. Lack of specialized care may not be as important for CRC because it is a common cancer (third most commonly diagnosed cancer for both men and women5) and therefore diagnosis and treatment requires less specialized care that is more widely available. Receipt of guideline cancer care has also been posited as a major driver of geographic disparities for many cancer types, and increasing access may reduce observed disparities.44 For example, we observed among the strongest geographic disparities for HCC and gastric cancer patients who received surgery. Even among CRC patients, for whom we did not observe overall geographic disparities, there were disparities among patients receiving either radiation or chemotherapy. These findings suggest that quality of care or treatment adherence may be poor among rural patients. Interventions that ensure equity of care across regions and that increase adherence may reduce observed geographic disparities. In contrast, among patients diagnosed with pancreatic cancer, geographic differences were minimal for those receiving surgery but were pronounced among those who were younger and those who were separated or widowed. Rurality may impact pancreatic cancer patients’ ability to access treatments, but not the efficacy of such treatments that are received. Pancreatic cancer is aggressive, often diagnosed late, and has few treatment options. Younger and non-partnered patients in rural areas may lack social supports that patients in urban areas have through living in more densely populated areas.
Overall racial disparities were observed for CRC and HCC mortality only. Racial disparities likely have more complex and distal drivers operating beyond the individual level, such as neighborhood factors and long-standing policies. Racial disparities among CRC patients were observed regardless of age, stage, and geography of county at diagnosis. Interestingly, CRC disparities were not observed for patients with Medicaid, those with no insurance, and for those living in high poverty counties. Thus, interventions aimed at these groups typically thought of as underserved may have reduced some CRC disparities, however, racial disparities persist for those with private insurance and those living in low poverty areas. Disparities were also observed for CRC patients receiving surgery, potentially indicating differences in treatment factors such as quality or timing of care. In contrast, among HCC patients, disparities were not pronounced among patients receiving surgery but were pronounced among those with Medicaid. For HCC patients, curative treatment includes liver transplantation and NHB patients were less likely to receive surgery. Among NHB patients who did receive surgery, only 29% received a liver transplant, compared to 52% of NHW patients. Racial disparities in liver transplantation are well documented and are an important area for intervention.45 Ensuring equal access to life-saving treatment can reduce mortality disparities. The lack of racial disparities among pancreatic and gastric cancer patients is noteworthy and may be limited to our population, as other studies have reported racial disparities for these cancer subtypes. This further demonstrates the importance of identifying and targeting limited public health resources to patient groups with most pronounced disparities.
One additional and important difference between GI cancers is the availability of an effective screening tool for CRC but not for the rarer cancers. CRC screening reduces mortality through two mechanisms. Detection and removal of pre-cancerous lesions reduces incidence and early detection of malignancies is associated with better survival.46 Greater adherence to screening could result in earlier age and stage at diagnosis. While we did not observe differences in either age or stage at diagnosis by geography, we did observe differences by race; Black patients were more likely to have distant stage disease, however, Black patients were also more likely to be younger at diagnosis. Further, race disparities in CRC mortality persisted for patients who were younger than the recommended screening age (50 years during our study period) and for patients diagnosed with localized or regional disease, suggesting other factors largely account for observed disparities.
The disparate findings for the association between race and mortality among gastric cancer patients when using different outcome variables are noteworthy. In analyses using the cause-specific death classification variable, we found no racial disparities, whereas in analyses using the COD recode variable, we found pronounced racial disparities. The difference in presence of racial disparity was due to differential outcome classification by race. Among gastric cancer patients only, NHW patients were nearly twice as likely to be censored as NHB patients when using the COD variable. Investigating specific CODs revealed nearly 70% of reclassified NHW patients had a COD corresponding to esophageal cancer. In contrast, only about one-third of reclassified NHB patients listed esophageal cancer as COD. In an earlier study of these variables, esophageal cancer was found to be a commonly coded COD for patients initially diagnosed with gastric cancer.47
Strengths of this study include a well-defined source population derived from a population-based cancer registry in a large, diverse state with the same catchment area for all cancers, the inclusion of a breadth of patient characteristics, and operationalization of geography in a manner which can be applied to other settings. RUC codes were developed by the US Department of Agriculture to classify all US counties.36 The lack of information on individual-level socioeconomic characteristics, such as household income, educational attainment, and current employment status is an important limitation. We were able to account for county-level SES, which is likely valid in areas of wide spread; these indices could be mis-specified for densely populated counties surrounding metropolitan areas where large socioeconomic gaps exist. We were also limited to treatment data available in the SEER database and did not have access to more detailed information, including facility type or treatment adherence, which could be indicative of healthcare access especially for recurring chemotherapy or radiation treatments. Last, it is possible that defining geography at the county level was not granular enough. An analysis of CRC-specific mortality using population-based data in Utah found rural-urban differences when classifying geography based on zip code.48 However, it is worth noting that Utah has a larger area but far fewer counties than Georgia (29 vs. 159).
In conclusion, this study provides a comparison of geographic and racial disparities by patient clinical, sociodemographic, and treatment characteristics across the four most common GI cancers. We identified specific cancer subtypes and patient groups for which disparities are more pronounced. Interventions aimed at reducing racial and geographic disparities in GI cancers, such as ensuring equal access to specialized care, should be targeted to groups experiencing a disproportionate burden of cancer-related mortality.
Supplementary Material
Highlights:
Disparities in gastrointestinal (GI) cancer mortality vary by cancer type
Geographic disparities are more pronounced for rarer cancer types
Racial disparities are greatest for colorectal cancer (CRC), the most common GI cancer
Patient factors associated with mortality disparities vary by cancer type and disparity
Racial disparities in CRC mortality persist even among patients receiving treatment
Acknowledgements:
Data for this study was provided by the Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov) Database: Incidence - SEER 18 Regs Research Data + Hurricane Katrina Impacted Louisiana Cases, Nov 2018 Sub (1975-2016 varying) - Linked To County Attributes - Total U.S., 1969-2017 Counties, National Cancer Institute, DCCPS, Surveillance Research Program, released April 2019, based on the November 2018 submission.
Financial Support:
L Collin was funded, in part, by the National Cancer Institute (F31CA239566) and the National Center for Advancing Translational Sciences (TL1TR002540) of the National Institutes of Health.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest statement: The authors declare no potential conflicts of interest.
Ethics approval statement: This study utilizes only secondary, previously collected, data available through the Surveillance, Epidemiology, and End Results Program and adheres to the SEER Research Data Use Agreement.
CRediT authorship contribution statement
Manuscript title: Understanding gastrointestinal cancer mortality disparities in a racially and geographically diverse population
Rebecca Nash: Analysis, Writing-drafting, reviewing, and editing; Maria C. Russell: Conceptualization, Writing-reviewing and editing; Jasmine M. Miller-Kleinhenz: Writing-reviewing and editing; Lindsay J. Collin: Conceptualization, Analysis, Writing-reviewing and editing; Katherine Ross-Driscoll: Conceptualization, Writing-reviewing and editing; Jeffrey M. Switchenko: Methodology, Writing-reviewing and editing; Lauren E. McCullough: Conceptualization, Supervision, Writing-reviewing and editing
REFERENCES
- 1.Koo MM, von Wagner C, Abel GA, et al. The nature and frequency of abdominal symptoms in cancer patients and their associations with time to help-seeking: evidence from a national audit of cancer diagnosis. J Public Health (Oxf). 2018;40(3):e388–e395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pullmer R, Linden W, Rnic K, Vodermaier A. Measuring symptoms in gastrointestinal cancer: a systematic review of assessment instruments. Support Care Cancer. 2014;22(11):2941–2955. [DOI] [PubMed] [Google Scholar]
- 3.Siegel RL, Miller KD, Goding Sauer A, et al. Colorectal cancer statistics, 2020. CA Cancer J Clin. 2020;70(3):145–164. [DOI] [PubMed] [Google Scholar]
- 4.Wolfgang CL, Herman JM, Laheru DA, et al. Recent progress in pancreatic cancer. CA Cancer J Clin. 2013;63(5):318–348. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70(1):7–30. [DOI] [PubMed] [Google Scholar]
- 6.Franco RA, Fan Y, Jarosek S, Bae S, Galbraith J. Racial and Geographic Disparities in Hepatocellular Carcinoma Outcomes. Am J Prev Med. 2018;55(5 Suppl 1):S40–S48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Henry KA, Niu X, Boscoe FP. Geographic disparities in colorectal cancer survival. Int J Health Geogr. 2009;8:48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Naishadham D, Lansdorp-Vogelaar I, Siegel R, Cokkinides V, Jemal A. State disparities in colorectal cancer mortality patterns in the United States. Cancer Epidemiol Biomarkers Prev. 2011;20(7):1296–1302. [DOI] [PubMed] [Google Scholar]
- 9.Rana N, Gosain R, Lemini R, et al. Socio-Demographic Disparities in Gastric Adenocarcinoma: A Population-Based Study. Cancers (Basel). 2020;12(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Aizer AA, Wilhite TJ, Chen MH, et al. Lack of reduction in racial disparities in cancer-specific mortality over a 20-year period. Cancer. 2014;120(10):1532–1539. [DOI] [PubMed] [Google Scholar]
- 11.Ellis L, Canchola AJ, Spiegel D, Ladabaum U, Haile R, Gomez SL. Racial and Ethnic Disparities in Cancer Survival: The Contribution of Tumor, Sociodemographic, Institutional, and Neighborhood Characteristics. J Clin Oncol. 2018;36(1):25–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kish JK, Yu M, Percy-Laurry A, Altekruse SF. Racial and ethnic disparities in cancer survival by neighborhood socioeconomic status in Surveillance, Epidemiology, and End Results (SEER) Registries. J Natl Cancer Inst Monogr. 2014;2014(49):236–243. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Clegg LX, Li FP, Hankey BF, Chu K, Edwards BK. Cancer survival among US whites and minorities: a SEER (Surveillance, Epidemiology, and End Results) Program population-based study. Arch Intern Med. 2002;162(17):1985–1993. [DOI] [PubMed] [Google Scholar]
- 14.Arnold LD, Patel AV, Yan Y, et al. Are racial disparities in pancreatic cancer explained by smoking and overweight/obesity? Cancer Epidemiol Biomarkers Prev. 2009;18(9):2397–2405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Brotherton L, Welton M, Robb SW. Racial disparities of pancreatic cancer in Georgia: a county-wide comparison of incidence and mortality across the state, 2000–2011. Cancer Med. 2016;5(1):100–110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Permuth JB, Clark Daly A, Jeong D, et al. Racial and ethnic disparities in a state-wide registry of patients with pancreatic cancer and an exploratory investigation of cancer cachexia as a contributor to observed inequities. Cancer Med. 2019;8(6):3314–3324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Khawja SN, Mohammed S, Silberfein EJ, Musher BL, Fisher WE, Van Buren G 2nd. Pancreatic cancer disparities in African Americans. Pancreas. 2015;44(4):522–527. [DOI] [PubMed] [Google Scholar]
- 18.Stewart SL, Kwong SL, Bowlus CL, et al. Racial/ethnic disparities in hepatocellular carcinoma treatment and survival in California, 1988–2012. World J Gastroenterol. 2016;22(38):8584–8595. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ashktorab H, Kupfer SS, Brim H, Carethers JM. Racial Disparity in Gastrointestinal Cancer Risk. Gastroenterology. 2017;153(4):910–923. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Alvidrez J, Castille D, Laude-Sharp M, Rosario A, Tabor D. The National Institute on Minority Health and Health Disparities Research Framework. Am J Public Health. 2019;109(S1):S16–S20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Zonderman AB, Ejiogu N, Norbeck J, Evans MK. The influence of health disparities on targeting cancer prevention efforts. Am J Prev Med. 2014;46(3 Suppl 1):S87–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Duran DG, Perez-Stable EJ. Novel Approaches to Advance Minority Health and Health Disparities Research. Am J Public Health. 2019;109(S1):S8–S10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.James CV, Moonesinghe R, Wilson-Frederick SM, Hall JE, Penman-Aguilar A, Bouye K. Racial/Ethnic Health Disparities Among Rural Adults - United States, 2012–2015. MMWR Surveill Summ. 2017;66(23):1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Douthit N, Kiv S, Dwolatzky T, Biswas S. Exposing some important barriers to health care access in the rural USA. Public Health. 2015;129(6):611–620. [DOI] [PubMed] [Google Scholar]
- 25.Goins RT, Williams KA, Carter MW, Spencer M, Solovieva T. Perceived barriers to health care access among rural older adults: a qualitative study. J Rural Health. 2005;21(3):206–213. [DOI] [PubMed] [Google Scholar]
- 26.Colen CG, Ramey DM, Cooksey EC, Williams DR. Racial disparities in health among nonpoor African Americans and Hispanics: The role of acute and chronic discrimination. Soc Sci Med. 2018;199:167–180. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Williams DR, Sternthal M. Understanding racial-ethnic disparities in health: sociological contributions. J Health Soc Behav. 2010;51 Suppl:S15–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.In: Smedley BD, Stith AY Nelson AR, eds. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington (DC) 2003. [PubMed] [Google Scholar]
- 29.US Census Bureau. Georgia 2010: Population and Housing Unit Counts. Issued August 2012. https://www.census.gov/library/publications/2012/dec/cph-2.html.
- 30.US Census Bureau. The Black Population: 2010. Issued September 2011. https://www.census.gov/prod/cen2010/briefs/c2010br-06.pdf.
- 31.National Cancer Institute: Surveillance, Epidemiology, and End Results Program. Site Recode ICD-O-3/WHO 2008 Definition. https://seer.cancer.gov/siterecode/icdo3_dwhoheme/index.html.
- 32.Surveillance, Epidemiology, and End Results (SEER): Overview of the SEER Program. National Cancer Institute. https://seer.cancer.gov/about/overview.html.
- 33.Surveillance, Epidemiology, and End Results (SEER): Static County Attributes. National Cancer Institute. https://seer.cancer.gov/seerstat/variables/countyattribs/static.html.
- 34.Facility Oncology Registry Data Standards. American College of Surgeons, Chicago, IL. 2002. [Google Scholar]
- 35.Noone AM, Lund JL, Mariotto A, et al. Comparison of SEER Treatment Data With Medicare Claims. Med Care. 2016;54(9):e55–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.US Department of Agriculture Economic Research Service. Rural-Urban Continuum Codes Documentation. Last Updated October 2019. https://www.ers.usda.gov/data-products/rural-urban-continuum-codes/documentation/.
- 37.Howlader N, Ries LA, Mariotto AB, Reichman ME, Ruhl J, Cronin KA. Improved estimates of cancer-specific survival rates from population-based data. J Natl Cancer Inst. 2010;102(20):1584–1598. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials. 1996;17(4):343–346. [DOI] [PubMed] [Google Scholar]
- 39.Howards PP, Schisterman EF, Poole C, Kaufman JS, Weinberg CR. “Toward a clearer definition of confounding” revisited with directed acyclic graphs. Am J Epidemiol. 2012;176(6):506–511. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Textor J, van der Zander B, Gilthorpe MS, Liskiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package ‘dagitty’. Int J Epidemiol. 2016;45(6):1887–1894. [DOI] [PubMed] [Google Scholar]
- 41.VanderWeele TJ, Knol MJ. A Tutorial on Interaction. Epidemiol Methods. 2014;3(1):33–72. [Google Scholar]
- 42.Henley SJ, Anderson RN, Thomas CC, Massetti GM, Peaker B, Richardson LC. Invasive Cancer Incidence, 2004–2013, and Deaths, 2006–2015, in Nonmetropolitan and Metropolitan Counties - United States. MMWR Surveill Summ. 2017;66(14):1–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Zahnd WE, James AS, Jenkins WD, et al. Rural-Urban Differences in Cancer Incidence and Trends in the United States. Cancer Epidemiol Biomarkers Prev. 2018;27(11):1265–1274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.McCullough LE, Flowers CR. Identifying and Addressing Disparities in Survival Outcomes for Rural Patients With Cancer. JAMA Netw Open. 2018;1(4):e181243. [DOI] [PubMed] [Google Scholar]
- 45.Nephew LD, Serper M. Racial, Gender, and Socioeconomic Disparities in Liver Transplantation. Liver Transpl. 2021;27(6):900–912. [DOI] [PubMed] [Google Scholar]
- 46.Wolf AMD, Fontham ETH, Church TR, et al. Colorectal cancer screening for average-risk adults: 2018 guideline update from the American Cancer Society. CA Cancer J Clin. 2018;68(4):250–281. [DOI] [PubMed] [Google Scholar]
- 47.Lund JL, Harlan LC, Yabroff KR, Warren JL. Should cause of death from the death certificate be used to examine cancer-specific survival? A study of patients with distant stage disease. Cancer Invest. 2010;28(7):758–764. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Rogers CR, Blackburn BE, Huntington M, et al. Rural-urban disparities in colorectal cancer survival and risk among men in Utah: a statewide population-based study. Cancer Causes Control. 2020;31(3):241–253. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.