Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Oct 1.
Published in final edited form as: J Gastrointest Surg. 2023 Aug 8;27(10):2155–2165. doi: 10.1007/s11605-023-05764-z

Association between Rural Residence and Processes of Care in Pancreatic Cancer Treatment

Krisha Amin 1, Hamza Khan 2, Larry R Hearld 3, Daniel I Chu 4, Victoria Prete 1, Krista R Mehari 5, Martin J Heslin 1, Annabelle L Fonseca 1
PMCID: PMC10731615  NIHMSID: NIHMS1946455  PMID: 37553515

Abstract

Background:

Pancreatic adenocarcinoma (PDAC) is an aggressive malignancy associated with poor outcomes. Surgical resection and receipt of multimodal therapy have been shown to improve outcomes in patients with potentially resectable PDAC, however treatment and outcome disparities persist on many fronts. The aim of this study was to analyze the relationship between rural residence and receipt of quality cancer care in patients diagnosed with non-metastatic PDAC.

Methods:

Using the National Cancer Database, patients with non-metastatic pancreatic cancer were identified from 2006–2016. Patients were classified as living in metropolitan, urban, or rural areas. Multivariable logistic regression was used to identify predictors of cancer treatment and survival.

Results:

A total of 41,786 patients were identified: 81.6% metropolitan, 16.2% urban, and 2.2% rural. Rural residing patients were less likely to receive curative-intent surgery (p = 0.037) and multimodal therapy (p <0.001) compared to their metropolitan and urban counterparts. On logistic regression analysis, rural residence was independently associated with decreased surgical resection [OR 0.82; CI 95% 0.69–0.99; p=0.039] and multimodal therapy [OR 0.70; CI 95% 0.38–0.97; p=0.047]. Rural residence independently predicted decreased overall survival [OR 1.64; CI 95% 1.45–1.93; p<0.001] for all patients that were analyzed. In the cohort of patients who underwent surgical resection, rural residence did not independently predict overall survival [OR 0.97; CI 95% 0.85–1.11; p=0.652].

Conclusions:

Rural residence impacts receipt of optimal cancer care in patients with non-metastatic PDAC but does not predict overall survival in patients who receive curative-intent treatment.

Keywords: pancreatic cancer, healthcare disparities, rural disparities, access to care, pancreatectomy, multimodal therapy

Introduction

Pancreatic cancer has an estimated five-year survival rate of just over 11% 1 and is predicted to become the second highest cause of cancer-associated mortality in the US by 2030 2, 3. In patients with potentially resectable pancreatic ductal adenocarcinoma (PDAC), surgical resection and receipt of multimodal therapy have been shown to improve outcomes 4–9. However, receipt of oncologic treatment in PDAC varies with respect to nonclinical factors such as race, gender, insurance, and socioeconomic status 10.

A growing body of literature suggests that rural-residing patients have lower quality of care, later stage diagnosis, and worse survival outcomes compared to their urban counterparts, even after controlling for confounding factors 10–14. In addition, rural patients face many inherent challenges to receiving quality care, such as longer travel times, lack of access to specialists, fewer opportunities for involvement in clinical trials, and limited financial resources 15. Disparities in oncologic treatment have been documented along the rural-urban continuum for various cancers. Associations between rural residence and survival outcomes have been observed both domestically and internationally in patients with PDAC 10, 12, 16, 17, with urban residence linked to longer survival rates. 12, 16

Although previous studies provide insight into the influence of social determinants of health and rurality on outcomes in patients with PDAC, 10–12, 16, 17 there is a paucity of literature examining the impact of rural residence on receipt of curative-intent therapy and on access-related disparities associated with PDAC. The purpose of this study was to use a large national cancer database to determine if rural residence predicts receipt of optimal cancer care. Specifically, we hypothesized that patients living in rural areas would have lower rates of curative-intent surgical resection and decreased receipt of multimodal therapy compared to their urban and metro counterparts. To test this hypothesis, we evaluated the association of rural residence with the following: 1) cancer care processes, including curative-intent surgical resection and multimodal chemotherapy and 2) outcomes, including short term post-operative outcomes, and overall survival.

Methods

Database

This retrospective study was conducted using data from the National Cancer Database (NCDB), a hospital-based registry sponsored by both the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. It is a clinical oncology database sourced from over 1,500 CoC-accredited facilities and captures over 72% of newly-diagnosed cancer cases in the US 18.This study was approved by the Institutional Review Board (IRB) at The University of South Alabama.

Study Population

Inclusion criteria consisted of patients diagnosed with non-metastatic PDAC (Stage I-III) between 2006 and 2016. Patients were identified using the following International Classification of Disease (ICD) codes including diagnostic codes: C20.0 - C25.3 and C25.7 - C25.9 and tumor histology: 8000, 8010, 8020–8022, 8140, 8141, 8211, 8230, 8500, 8521, 8050, 8260, 8441, 8450, 8453, 8470–8473, 8480, 8481, 8503. Patients excluded from the study included those with Stage IV exocrine PDAC, pancreatic endocrine neoplasms, patients with PDAC diagnosed before 40 years of age (due to incomplete NCDB facility information on these patients). Other exclusion criteria included patients with a history of prior cancers, and patients with a Charlson-Deyo comorbidity score of 3 or higher, since a higher comorbidity profile could preclude surgical resection.

Study Variables

NCBD defines metropolitan (metro), urban, and rural based on the National Center for Health (NCHS) 2013 urban-rural classification scheme 19. Levels of the NCHS scheme are used to specifically study associations between urbanization level of residence and health and to monitor health differences across the urban-rural continuum.

Metropolitan counties were defined by the population size of their metro area; whereas, non-metro counties were defined by the degree of their urbanization and adjacency to a metro area. The metro category included subjects living in the following regions: counties in metro areas of 1 million in population or more, counties in metro areas of 250,000 to 1 million population, and counties in metro areas of fewer than 250,000 population. The Urban group was defined as urban populations of 20,000 or more (adjacent to metro areas), urban populations of 20,000 or more (not adjacent to metro areas), urban populations of 2,500 to 19,999 (adjacent to metro areas) and urban populations of 2,500 to 19,999 (not adjacent to metro areas). The Rural category was defined as completely rural or less than 2,500 urban population (adjacent to a metro area) and completely rural or less than 2,500 urban population (not adjacent to a metro area).

The demographic variables included patient age, gender, race, zip code-based median income and education level, U.S. regional location and geographic residence. Age was grouped into the following age groups: 40 to 59 years, 60 to 79 years, and 80+ years. Median income was reported based on the median household income for the zip code of the patient’s residence at the time of diagnosis. Education level was reported based on the percentage of individuals without a high school diploma in the zip code of the patient’s residence at the time of diagnosis. The data on median income and education level were aggregated into quartiles.

Illness and treatment-related variables included disease stage, Carlson-Deyo comorbidity score, Crowfly distance (the “great circle” or longitudinal distance in miles between patients’ residence and address of treatment facility where they received care), type of treatment facility, fragmented care, treatment received, insurance status, and outcome variables including 30- and 90-day survival as well as overall mortality. Stage was defined based on Tumor, Node, and Metastasis (TNM) staging system defined by the American Joint Committee on Cancer (AJCC) 8th edition staging manual. High-volume facility (HVF) was defined using Leapfrog criteria and included those institutions that performed at least 20 pancreatic operations per year 20–22. Fragmented care was defined as receipt of oncologic care at more than one facility. Curative-intent surgical resection was defined by set NCDB codes for partial pancreatectomy (30), pancreaticoduodenectomy (35–37), total pancreatectomy (40), and extended pancreatectomy (70) 23. Multimodal therapy was defined as curative-intent surgical resection plus chemotherapy, given in either a neoadjuvant or adjuvant fashion.

Statistical Analysis

Sociodemographic variables for the three geographic areas were compared using Chi square test for categorical variables and ANOVA test for continuous variables. Multiple logistic regression models were then performed to examine the extent to which access to care (receipt of curative-intent resection and multimodal therapy) and post-operative outcomes (30- and 90- day mortality) were independently associated with rural residence. Factors controlled for included age, gender, race, AJCC stage, Charlson-Deyo score, median income, insurance status, regional location, geographic residence, type of treating facility, and treatment at HVF. Educational level was excluded from the model due to its collinearity with median income on the basis of zip code.

A Cox regression analysis was performed to evaluate the effect of geographic residence on overall survival (OS). OS was defined as the time in months from diagnosis to the time of death. Patients who did not die were censored at the date of last contact. As with the logistic regressions, the model also included age, gender, race, AJCC stage, Charlson-Deyo score, median income, insurance status, regional location, geographic residence, type of treating facility, and treatment at HVF.

Variables with p-values less than 0.05 were considered statistically significant. All statistics were estimated using Stata version 16.1 software (Stata Corporation, College Station, TX).

Results

Patient profile

Overall, 41,786 patients met study criteria. The patient distribution by area of residence consisted of metro (n=33,298, 81.6%), urban (n=6,599, 16.2%), and rural (n=887, 2.2%) (Table 1). Although White patients constituted the majority group in all areas, there was a higher percentage of Black patients in metro regions compared to rural regions (13.2% vs. 5.7%, p < 0.001). Urban and rural groups consisted of a greater percentage of male patients compared to patients in the metro group (51.6% and 54.5% vs. 49.6%, p<0.001). The average income of rural patients was the lowest amongst the three groups with 50.53% of subjects in the lowest quartile, whereas, 48.9% of subjects in the metro group fell within the highest quartile (p<0.001). Metro residents were more likely to hold private insurance (39.7%) versus rural residents (30.4%) (p<0.001). Rural residents were more likely to receive fragmented care (32.0%) compared to patients in metro and urban groups (27.9% and 30.1%, respectively; p<0.001). Most of the patients in the rural group resided in either the Midwest or the South (44.1% and 44.7%, respectively). Baseline patient characteristics are summarized in Table 1.

TABLE 1.

Patient demographics and treatment characteristics

Variable Metro Urban Rural p-value
Age n (%) n (%) n (%) <0.001*
40–59 yrs 9,019 (27.4) 1,750 (26.8) 211 (24.1)
60–79 yrs 20,096 (61.1) 4,168 (63.9) 575 (65.6)
80+ yrs 3,787 (11.5) 609 (9.3) 91 (10.4)
Gender <0.001*
Male 16,524 (49.6) 3,402 (51.6) 483 (54.5)
Female 16,774 (50.4) 3,197 (48.5) 404 (45.6)
Race <0.001*
White 27,214 (83.5) 5,958 (91.7) 816 (93.4)
Black 4,287 (13.2) 456 (7.0) 50 (5.7)
Asian 1,019 (3.1) 41 (0.6) 2 (0.2)
Native American 62 (0.2) 41 (0.6) 6 (0.7)
Median income <0.001*
Lowest quartile 3,775 (12.8) 1,904 (35.3) 381 (50.5)
Second quartile 2,820 (9.5) 1,479 (27.4) 197 (26.1)
Third quartile 8,552 (28.9) 1,623 (30.0) 151 (20.0)
Highest quartile 14,470 (48.9) 396 (7.3) 25 (3.3)
Insurance status <0.001*
Private 12,861 (39.7) 2,234 (35.0) 259 (30.4)
Medicaid 1,870 (5.8) 347 (5.4) 67 (7.9)
Medicare 16,655 (51.4) 3,593 (56.4) 504 (59.1)
Uninsured 1,006 (3.1) 202 (3.2) 23 (2.7)
Charlson Comorbidity Score 0.1
0 21,973 (69.0) 4,234 (66.9) 585 (68.3)
1 8,002 (25.1) 1,699 (26.8) 226 (26.4)
2 1,891 (5.9) 411 (6.5) 45 (5.3)
Regional Location <0.001*
West 5,266 (16.0) 38 (11.3) 72 (8.2)
Northeast 7,351 (22.3) 681 (10.4) 26 (3.0)
Midwest 8,242 (25.1) 2,294 (35.2) 387 (44.1)
South 12,043 (36.6) 2,814 (43.1) 392 (44.7)
AJCC Stage 0.47
1 5,946 (17.9) 1,162 (17.6) 147 (16.6)
2 10,323 (31.0) 1,996 (30.3) 270 (30.4)
3 17,029 (51.1) 3,441 (52.1) 470 (53.0)
Treatment at academic center <0.001*
Yes 17,107 (51.4) 3,158 (47.9) 374 (42.2)
No 16,191 (48.6) 3,441 (52.1) 513 (57.8)
Treatment at high- volume facility <0.002
Yes 13,894 (41.7) 2,797 (57.6) 320 (36.1)
No 19,404 (58.3) 3,802 (42.4) 567 (63.9)
Crowfly distance <0.001*
11 (5-25.6) 46.6 (28.2-76.3) 75.8 (42.7-126.9)
Fragmented care <0.001*
Yes 9,294 (27.9%) 2,029 (30.1%) 284 (32.0%)
No 24,004 (72.1%) 4,570 (69.3%) 603 (68.0%)
Open in a new tab

Impact of rural residence on receipt of cancer care processes

Impact of rural residence on surgical resection

Curative intent surgical resection was received by 20,231 patients: 49.3% in metro, 48.1% in urban, and 46.0% in rural regions (p=0.037) (Table 2). In the logistic regression models, compared to residence in metro locations, residence in urban (OR 0.90, 95% CI 0.83–0.97; p=0.008) and rural locations (OR 0.82, 95% CI 0.69–0.99; p=0.039) were associated with decreased receipt of surgical resection. Other factors independently associated with decreased surgical resection included older age, Black and Native American races, greater Charlson comorbidity scores, uninsured status, treatment at non-academic centers, and treatment at non-HVFs. While uninsured status (OR 0.78, 95% CI 0.61–0.93; p=0.035) was also predictive of decreased receipt of surgical resection, this association did not appear to be impacted by residence in a rural location. In a subset analysis of uninsured patients, residence in rural locations was not significantly associated with receipt of surgical resection: 47.8% in patients living in rural areas versus 46.0% and 49.1% in patients living in urban and metro locations respectively (p= 0.7).

TABLE 2.

Univariate and multivariate analysis of factors associated with curative-intent surgical resection.

Univariate analysis Multivariate analysis
Variable No surgery Yes surgery p-value OR (95% CI) p-value
Age n (%) n (%) <0.001*
40–59 yrs 5,603 (49.8) 5,648 (50.2) Ref
60–79 yrs 12,689 (49.7) 12,855 (50.3) 0.97 (0.91–1.04) 0.46
80+ yrs 2,869 (62.4) 1,728 (37.6) 0.61 (0.55–0.68) <0.001*
Gender 0.188
Male 10,622 (50.7) 10,345 (49.3)
Female 10,727 (51.3) 10,182 (48.7)
Race 0.001*
White 17,760 (50.6) 17,257 (49.4) Ref
Black 2,577 (52.9) 2,297 (47.1) 0.84 (0.77–0.91) <0.001*
Asian 604 (55.0) 494 (45.0) 0.93 (0.79–1.09) 0.374
Native American 57 (50.4) 56 (49.6) 0.78 (0.63–0.87) <0.001*
Insurance status <0.001*
Private 7,763 (49.2) 8,005 (50.8) Ref
Medicaid 1,270 (54.3) 1,068 (45.7) 0.89 (0.79–1.01) 0.066
Medicare 11,055 (51.9) 10,265 (48.2) 0.99 (0.93–1.06) 0.839
Uninsured 643 (51.4) 608 (48.6) 0.78 (0.61–0.93) 0.035*
Median income of zip code <0.001*
First quartile 3,103 (50.64) 3,025 (49.36) Ref
Second quartile 2,275 (49.66) 2,306 (50.34) 1.04 (0.95–1.14) 0.413
Third quartile 5,590 (52.91) 4,975 (47.09) 0.94 (0.87–1.02) 0.127
Fourth quartile 7,766 (50.57) 7,591 (49.43) 0.98 (0.91–1.07) 0.712
Geographic location 0.037*
Metro 16,876 (50.7) 16,422 (49.3) Ref
Urban 3,424 (51.9) 3,175 (48.1) 0.90 (0.83–0.97) 0.008*
Rural 479 (54.0) 408 (46.0) 0.82 (0.69–0.99) 0.039*
Regional Location <0.001*
West 3,502 (55.7) 2,781 (44.3) Ref
Northeast 4,503 (54.1) 3,826 (45.9) 0.87 (0.80–0.95) 0.002*
Midwest 5,562 (50.1) 5,545 (49.9) 1.13 (1.04–1.23) 0.004*
South 7,594 (48.5) 8,079 (51.6) 1.21 (1.11–1.31) <0.001*
AJCC Stage <0.001*
1 1,849 (24.8) 5,615 (75.2) Ref
2 3,380 (26.2) 9,513 (73.8) 0.93 (0.86–1.00) 0.053
3 16,120 (74.91) 5,399 (25.1) 0.09 (0.08–0.10) <0.001*
Charlson Comorbidity Score <0.001*
0 14,988 (53.2) 13,192 (46.8) Ref
1 4,864 (47.0) 5,485 (53.0) 1.39 (1.31–1.48) <0.001*
2 1,093 (44.0) 1,336 (56.0) 1.53 (1.38–1.71) <0.001*
Facility type <0.001*
Academic 12,239 (59.0) 8,519 (41.0) Ref
Non-academic 9,110 (43.1) 12,008 (56.9) 0.63 (0.59–0.67) <0.001*
High-volume facility <0.001*
No 14,702 (60.2) 9,734 (39.8) Ref
Yes 6,647 (38.1) 10,793 (61.9) 2.41 (2.26–2.56) <0.001*
Open in a new tab

Impact of rural residence on receipt of multimodal therapy

Among patients (n=20,231) who received surgical resection, 12,781 (63.2%) received multimodal therapy: 79.8% in metro, 58.1% in urban, and 51.2% in rural regions (p<0.001) (Table 3). On logistic regression, residence in rural regions (OR 0.70, 95% CI 0.38–0.97; p=0.047) was independently associated with decreased receipt of multimodal therapy. Other independent predictors of decreased receipt of multimodal therapy included older age, non-private insurance, residence in the Northeast or Midwest, and treatment at non-HVFs.

TABLE 3.

Univariate and multivariate analysis of factors associated with receipt of multimodal therapy.

Univariate analysis Multivariate analysis
Variable Multimodal therapy p-value OR (95% CI) p-value
No; n (%) Yes; n (%)
Age p<0.001*
40–59 yrs 1,761 (31.2) 3,887 (68.8) Ref
60–79 yrs 4,739 (36.9) 8,116 (63.1) 0.88 (0.81–0.97) p=0.007*
80+ yrs 1,091 (63.1) 637 (36.9) 0.30 (0.26–0.35) p<0.001*
Gender p=0.317
Male 3,879 (37.4) 6,476 (62.6)
Female 3,877 (38.1) 6,305 (61.9)
Race p=0.080
White 6,431 (37.3) 10,826 (62.7) Ref
Black 891 (38.8) 1,406 (61.2) 1.04 (0.93–1.16) p=0.499
Asian 208 (42.1) 286 (57.9) 0.82 (0.66–1.00) p=0.054
Native American 23 (41.1) 33 (58.9) 1.20 (0.59–2.43) p=0.626
Insurance status p<0.001*
Private 2,493 (31.1) 5,512 (68.9) Ref
Medicaid 374 (35.0) 694 (65.0) 0.88 (0.75–0.91) p=0.014*
Medicare 4,357 (42.5) 5,908 (57.6) 0.74 (0.68–0.80) p<0.001*
Uninsured 262 (43.1) 346 (56.9) 0.66 (0.55–0.81) p<0.001*
Median income of zip code p<0.001*
First quartile 1,270 (42.0) 1,755 (58.0) Ref
Second quartile 1,058 (45.9) 1,248 (54.1) 0.83 (0.74–0.94) p=0.003*
Third quartile 1,866 (37.5) 3,109 (62.5) 1.12 (1.01–1.25) p=0.036*
Fourth quartile 2,664 (35.1) 4,927 (64.9) 1.20 (1.07–1.34) p=0.001*
Geographic location p<0.001*
Metro 6,063 (36.9) 10,359 (63.1) Ref
Urban 1,331 (41.9) 1,844 (58.1) 0.90 (0.68–0.94) p=0.001*
Rural 199 (48.8) 209 (51.2) 0.70 (0.38–0.97) p=0.047*
Regional Location p<0.001*
West 1,021 (36.7) 1,760 (63.3) Ref
Northeast 1,274 (33.3) 2,552 (66.7) 1.25 (1.11–1.41) p<0.001*
Midwest 1,762 (31.8) 3,783 (68.2) 1.31 (1.17–1.47) p<0.001*
South 3,534 (43.7) 4,545 (56.3) 0.80 (0.72–0.88) p<0.001*
AJCC Stage p<0.001*
1 2,404 (42.8) 3,211 (57.2) Ref
2 3,496 (36.8) 6,017 (63.3) 1.28 (1.19–1.39) p<0.001*
3 1,846 (34.2) 3,553 (65.8) 1.34 (1.22–1.46) p<0.001*
Charlson Comorbidity Score p=0.001
0 4,928 (37.4) 8,264 (62.6) Ref
1 2,050 (37.4) 3,435 (62.6) 1.02 (0.95–1.10) p=0.582
2 535 (40.0) 801 (60.0) 0.89 (0.78–1.02) p=0.087
Facility type p=0.158
Academic 4,483 (37.3) 7,525 (62.7)
Non-academic 3,263 (38.3) 5,256 (61.7)
High-volume facility p<0.001
No 3,888 (39.9) 5,846 (60.1) Ref
Yes 3,858 (35.8) 6,935 (64.3) 1.24 (1.15–1.33) p<0.001*
Open in a new tab

Impact of rural residence on post-operative outcomes

Among patients who received surgical resection, overall 30-day mortality was 3.28% and 90-day mortality was 6.46%. 30-day mortality was 2.50% for rural patients (vs. 3.31% in metro and 3.42% in urban; p= 0.626) and 90-day mortality was 4.75% for rural patients (vs. 6.44% in metro and 7.03% in urban; p= 0.169). While the overall 30-day and 90-day post-operative mortality was lower for patients residing in rural areas compared to urban and metro areas, this was not significantly different on univariable analysis. Rural residence was not independently associated with 30-day and 90-day mortality. Independent factors predicting a greater likelihood of 30-day and 90-day mortality included older age, male gender, and a Charlson Comorbidity score of 2. Higher income quartiles and female gender predicted a lower likelihood of both 30-day and 90-day survival. Additionally, non-private insurance status and stage 2 and 3 disease was independently associated with increased 90-day mortality (Supplementary table 1 and 2).

Impact of rural residence on fragmentation of care

Overall, rural residents were more likely to receive fragmented care (32.0%) compared to patients in metro and urban groups (27.9% and 30.1%, respectively; p<0.001). In the cohort of patients who received curative-intent resection, residence in a rural location was not associated with care fragmentation (24.8% versus 25.5% in urban and 23.82% in metro locations; p= 0.12).

Impact of rural residence on survival

Survival data were available for 33,506 patients diagnosed with PDAC, 27,508 of whom were deceased at the last follow up evaluation. The median follow-up period was 14.23 months (interquartile range [IQR], 7.06– 26.68 months). Of all patients diagnosed with PDAC, rural residence was independently associated with decreased overall survival; (hazard ratio [HR], 1.64; 95 CI 1.45– 1.93). Additional factors independently associated with decreased overall survival included older age, non-private insurance, lower income quartiles, higher Charlson comorbidity score and higher AJCC stage. (Table 4).

TABLE 4.

Multivariate analysis of factors associated with overall survival and overall survival in pancreatic cancer surgery patients.

Variable Overall Survival Survival in patients who underwent surgical resection
Hazard Ratio (95%) CI p-value Hazard Ratio (95%) CI p-value
Age
40–59 yrs Ref Ref
60–79 yrs 1.16 (1.12–1.19) <0.001* 1.17 (1.11–1.23) <0.001*
80+ yrs 1.89 (1.81–1.99) <0.001* 1.66 (1.54–1.80) <0.001*
Gender
Male Ref Ref
Female 0.98 (0.96–1.00) 0.076 0.94 (0.91–0.98) 0.002*
Race
White Ref Ref
Black 1.03 (0.99–1.08) 0.092 0.96 (0.91–1.02) 0.227
Asian 0.92 (0.85–1.01) 0.068 0.85 (0.75–0.97) 0.015*
Native American 1.01 (0.80–1.27) 0.937 0.96 (0.68–1.36) 0.812
Insurance status
Private Ref Ref
Medicaid 1.23 (1.16–1.30) <0.001* 1.31 (1.20–1.44) <0.001*
Medicare 1.16 (1.12–1.19) <0.001* 1.21(1.15–1.26) <0.001*
Uninsured 1.13 (1.05–1.22) 0.001* 1.11 (0.99–1.24) 0.069
Median income of zip code
Lowest quartile Ref Ref
Second quartile 1.03 (0.98–1.08) 0.213 0.95(0.88–1.01) 0.118
Third quartile 0.92 (0.89–0.96) <0.001* 0.89 (0.84–0.94) <0.001*
Highest quartile 0.85 (0.82–0.89) <0.001* 0.81 (0.76–0.86) <0.001*
Geographic Location
Metro Ref Ref
Urban 1.12 (1.09–1.66) 0.043* 1.03 (.97–1.08) 0.371
Rural 1.64 (1.45–1.93) <0.001* 0.97 (.85–1.11) 0.652
Location
West Ref Ref
Northeast 0.86 (0.83–0.90) <0.001* 0.90 (0.84–0.96) 0.002*
Midwest 0.99(0.95–1.03) 0.502 1.05 (0.98–1.11) 0.151
South 1.00 (0.97–1.04) 0.828 1.01 (0.95–1.07) 0.704
AJCC Stage
1 Ref Ref
2 1.52 (1.46–1.58) <0.001* 1.66 (1.59–1.74) <0.001*
3 2.84 (2.74–2.94) <0.001* 2.28 (2.17–2.40) <0.001*
Charlson Comorbidity Score
0 Ref Ref
1 1.07(1.04–1.10) <0.0001* 1.04 (0.99–1.08) 0.092
2 1.21 (1.15–1.28) <0.001* 1.19 (1.11–1.29) <0.001*
Facility type
Academic Ref Ref
Non-academic 1.05 (1.02–1.08) 0.001* 1.09 (1.05–1.14) <0.001*
High-volume facility
No Ref Ref
Yes 0.85 (0.83–0.88) <0.001* 0.84 (0.80–0.87) <0.001*
Open in a new tab

In the cohort of patients who underwent curative-intent surgical resection, survival data were available for 15,492 patients, 11,601 of whom were deceased at last follow-up. Notably, in this cohort, rural residence was not associated with decreased survival; (HR, 0.97; 95 CI 0.85– 1.11) Factors independently associated with decreased survival among those who underwent surgical resection included older age, female gender, non-private insurance, lower income quartiles, higher Charlson comorbidity score, and higher AJCC stage (Table 4).

Discussion

Rural residence predicts lower rates of receipt of optimal pancreatic cancer care. Our analysis demonstrates that rural residence is associated with decreased receipt of curative-intent surgical resection and multimodal therapy. However, in the patients who receive curative-intent surgical resection, there is no difference in post-operative mortality. Residence in a rural location was predictive of decreased overall survival in patients diagnosed with PDAC. However, when the cohort of patients who received curative-intent surgical resection was assessed, rural residence was not associated with decreased overall survival. This suggests that access to guideline-concordant curative-intent surgical resection can be the limiting factor leading to survival disparities. Identifying oncologic treatment disparities are necessary to address regional health inequities and implement policy changes. To the best of our knowledge, this study is the first to evaluate disparities in access to guideline-concordant multimodal therapy (surgical resection and chemotherapy) based on rural residence in patients diagnosed with non-metastatic PDAC.

Surgical resection is a potentially curative therapeutic intervention for eligible patients with PDAC 24. Our study found that significantly fewer rural residing patients underwent curative-intent surgical resection compared to their metro and urban counterparts, even when controlling for other factors. Rural locations face a multitude of challenges that may explain the decreased frequency of surgery. For instance, rural patient populations are associated with poorer health literacy, lower incomes, increased medical mistrust, increased comorbidities, lower educational levels, and refusal of cancer treatment 25, 26. Rural hospitals also have lower patient volumes that may preclude them from obtaining appropriate performance measures relating to surgical treatment, and may limit surgeon experience with complex pancreatic operations 27. Similar to other studies, we found that treatment at HVFs was independently associated with increased curative-intent surgical resection. Rural patients typically have to travel long distances to access care at HVFs 17, 28. In this study patients residing in rural locations had a significantly greater Crowfly distance of 75.8 miles compared to their urban and metro counterparts (11 miles vs. 46.6 miles; p<0.001). While the Crowfly distance provides a straight line distance which does not accurately reflect the actual distance traveled, this does provide information about the increased distance that rural patients need to travel to receive care. Rural residents in our study had the lowest rates of private insurance. This study, as well as others have shown that private insurance is associated with increased receipt of curative-intent surgical resection, 29–33 suggesting that decreased receipt of treatment may be multifactorial and related to a number of barriers in access to care rather than to rural residence alone. In this study, uninsured patients were significantly less likely to receive surgical resection, however, subset analysis of uninsured patients did not show an association between residence in a rural location and surgical resection. While several health systems have charity care programs that allow for delivery of care regardless of insurance status, this also requires a degree of access which may be challenging for patients living in rural areas. Therefore, while patients with who are able to reach a health system that offers such a program may be able to access cancer care regardless of area of residence, the greater travel distance involved may pose an additional barrier for patients residing in rural locations.

For the purposes of this study, we looked at receipt of multimodal therapy, which constitutes curative-intent surgical resection and chemotherapy, given in the neoadjuvant or adjuvant fashion. A significant proportion of rural residing patients (48.8%) in our study failed to receive multimodal therapy. The reasons for failure to receive multimodal therapy are complex but can perhaps be attributed to fragmented care at various facilities 34. Given that rural residence has been associated with longer travel distances and time to cancer care centers, 17, 28 it would be expected that rural patients are more likely to receive care at various cancer centers. Indeed, the ability to receive a portion of their therapy closer to home, may facilitate completion of multimodal therapy. Our analysis revealed that rural patients were more likely to receive fragmented cancer care than their metro and urban counterparts. However, in patients who received curative-intent resection, there was no significant difference in fragmentation of care. This finding is congruent with Bertens et al., who found that urban and rural patients were equally as likely to receive adjuvant chemotherapy following resection 12. Combined, these findings suggest that perhaps rural patients who traveled farther distances to receive curative-intent surgery were also more likely to receive multimodal therapy if they opted to receive the entirety of their care at a single HVF. While fragmentation of care has been reported to be associated with negative outcomes, this is nuanced, particularly in the case of complex oncological care. In the case of PDAC, this may be reflective of multiple reasons including the trend towards increased centralization of pancreatectomies to HVF, with the receipt of chemotherapy closer to home. In a recent publication, Khan et al. showed that fragmented care may represent a measure of access that is available to relatively advantaged patient groups 34. In their study, younger patients, White patients, and patients with higher incomes and private insurance were more likely to have fragmented care, and more likely to receive care at academic facilities and HVF. Additionally, fragmented care was noted to be independently associated with improved outcomes, including post-operative mortality and overall survival 34. Also of note, our analysis of those receiving multimodal therapy showed that these patients were significantly more likely to have private insurance. This suggests that wealthier patients in rural areas are able to access and receive multimodal therapy, leaving a large majority of socially disadvantaged patients without the same resources.

Our analysis showed that rural residence independently predicts decreased overall survival among all patients diagnosed with non-metastatic PDAC. However, among the cohort of patients who underwent surgery, we found no significant variations in overall survival across the urban-rural continuum. Our study analysis supports other published literature on various cancers, including the study by Chu et al. on PDAC, which showed that overall survival was decreased among rural residing patients, but this disparity is mitigated when adjusting for access to treatment factors 9, 10, 35. These findings together suggest that residence in a rural location is not, in and of itself, causal to poor outcomes in PDAC. Instead, location-based disparities in PDAC outcomes may be explained by disparities in access to care. In the case of rural patients, the increased likelihood of older age, greater Charlson comorbidity scores, poorer income status, non-private insurance status, and greater distance from treating facilities may also contribute to decreased survival.

Our data differs from studies published from countries outside of the United States in several ways. In their study of patients with advanced pancreatic cancer (Stage III and IV) in six cancer centers in British Columbia, Canale et al. found that rural and urban status was not associated with differences in baseline clinical characteristics, treatment, variations in treatment patterns, or overall survival 17. The authors attribute this to the strategic geographic allocation of cancer care delivery across the province of British Columbia. Kirkegård et al. evaluated the role of location on the treatment and survival of patients with PDAC using the Danish Cancer Registry, and found that urban residing patients were more likely to receive surgery and chemotherapy, and had slightly improved survival, even in a universal, tax-financed healthcare system 16. It is important to note that these studies evaluated different populations, and that neither of these studies were limited to patients with resectable PDAC; the study by Canale et al. was limited to patients with Stage III and IV pancreatic cancer, and Kirkegård et al. included all disease stages 16, 17. Notably, even in countries with universal health coverage such as Denmark, when viewed broadly, location-based disparities exist in treatment and outcomes, highlighting the role of health literacy and other social determinants of health, even in theoretically equal-access systems. While countries such as Denmark with free and equal access to healthcare, have regionalized pancreatic cancer care with improved results, these are generally smaller countries, with limited travel distances, and with systems that allow for patient receive reimbursement for health-related travel expenses.

In addition to insurance related barriers in the U.S system, additional barriers that rural patients may face in accessing high-quality healthcare include lack of specialized healthcare in more remote, rural locations and travel barriers including geographic remoteness, lack of transportation, and the financial cost of transportation. However, some studies have shown that improved insurance access through Medicaid expansion has mitigated some treatment disparities in both non-metastatic and metastatic PDAC 33, 36. Furthermore, digital innovations in the form of telemedicine initiatives and virtual tumor boards may improve health access for rural cancer patients 28, 37. Whether or not these advances can definitively mitigate treatment disparities remains to be seen, but expansion of such initiatives is nonetheless necessary to improve access to care for patients residing in rural locations.

While our study establishes cancer care disparities among rural patients with PDAC, it has several limitations. This study is limited by the typical constraints of the retrospective nature of the database, which includes selection bias and the potential for missing or confounding data. Surgical decision-making for PDAC is complex, with several factors impacting resectability including anatomic considerations, underlying tumor biology and patient comorbidities. We restricted our patient population to those without metastatic disease and with a Charlson-Deyo comorbidity score < 3. While surgical resectability decreases from Stage I through Stage III, a certain percentage of patients with borderline-resectable disease and locally advanced disease will become surgical candidates. Conversely, patients with anatomically resectable disease may not be surgical candidates due to biologically aggressive disease. The NCDB is a public health database that lacks much of the individual-level information necessary to evaluate granular, patient-specific surgical decision making. Given that the distribution of stages was not significantly different based on location, it is unlikely that this impacts the overall analyses. In order to limit the analysis regarding curative-intent resection to those who may be surgical candidates, we excluded patients with a Charlson-Deyo comorbidity score 3 or more however this does not account for patient frailty and other more nuanced factors that are important for surgical risk-assessment. While CoC-accredited facilities are required to report elements of cancer treatment that took place at non-CoC facilities, this requires cooperation from non-CoC institutions and therefore, certain elements of treatment may be under-reported. Additionally, if a patient receives care at two CoC institutions, the more complete record is selected for. Missing data may have a significant but undefined effect on outcomes analysis. While the NCDB database represents 70% of cancer patients, it only represents 30% of US hospitals with only 3.8% of these hospitals being rural referral centers 38. Rural hospitals are less likely to be accredited by the CoC, and therefore this study fails to capture much of the rural population that may have received treatment at a local level. While rural hospitals are less likely to be reported in the CoC, it should be noted that overall CoC-approved programs generally offer a higher level of cancer care and more frequently offer oncology-related services, including screening programs, chemotherapy, and radiation therapy services. Therefore, there may be a larger treatment disparity between rural residents and their metro and urban counterparts than can be elucidated from the NCDB database.

Future research should explore the extent of treatment disparities for patients in rural areas unable to receive treatment at CoC-accredited facilities. Identifying risk factors and barriers specific to rural communities can guide oncologic providers in understanding disparities associated with pancreatic cancer treatment and addressing the underlying disparities.

Conclusion

Our study analyzed the impact of rural residence on receipt of treatment and outcomes in PDAC. We found that rural residing patients diagnosed with non-metastatic PDAC are more likely to receive suboptimal treatment in the form of decreased rates of curative-intent surgical resection and multimodal therapy. While rural residence was associated with decreased survival overall, it was not associated with decreased overall survival among patients who underwent curative-intent surgical resection, likely due to this subset of patients having the ability to seek high quality, surgical care outside their area of residence. Therefore, rural location-based disparities in treatment and outcomes may be explained by substantial disparities in access to high-quality treatment. These disparities in treatment are likely due to multiple, unmeasured barriers along the continuum of cancer care, many of which are potentially modifiable. It is critical that future research identify these barriers and seek to understand how they may be mitigated. Most existing studies evaluating disparities in access to oncologic care, including this one, use retrospective data and administrative databases. While these studies serve as an appropriate starting point to study the existence of disparities, in order to examine the underlying causes and maintainers of extant disparities, it is essential that we integrate quantitative research with high-quality, patient-focused mixed methods research to understand individual- and systems-level barriers that be potential targets for future intervention.

Supplementary Material

Supplementary Tables

Footnotes

Conflicts of Interest: All authors declare no conflicts of interest

References

  • 1.Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin. 2022;72(1):7–33. doi: 10.3322/caac.21708. [DOI] [PubMed] [Google Scholar]
  • 2.Strobel O, Neoptolemos J, Jäger D, Büchler MW. Optimizing the outcomes of pancreatic cancer surgery. Nat Rev Clin Oncol. 2019;16(1):11–26. doi: 10.1038/s41571-018-0112-1. [DOI] [PubMed] [Google Scholar]
  • 3.Ansari D, Tingstedt B, Andersson B, Holmquist F, Sturesson C, Williamsson C et al. Pancreatic cancer: yesterday, today and tomorrow. Future Oncol. 2016;12(16):1929–46. doi: 10.2217/fon-2016-0010. [DOI] [PubMed] [Google Scholar]
  • 4.Ahola R, Sand J, Laukkarinen J. Centralization of Pancreatic Surgery Improves Results: Review. Scand J Surg. 2020;109(1):4–10. doi: 10.1177/1457496919900411. [DOI] [PubMed] [Google Scholar]
  • 5.David JM, Kim S, Placencio-Hickok VR, Torosian A, Hendifar A, Tuli R. Treatment strategies and clinical outcomes of locally advanced pancreatic cancer patients treated at high-volume facilities and academic centers. Adv Radiat Oncol. 2019;4(2):302–13. doi: 10.1016/j.adro.2018.10.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Gluth A, Werner J, Hartwig W. Surgical resection strategies for locally advanced pancreatic cancer. Langenbecks Arch Surg. 2015;400(7):757–65. doi: 10.1007/s00423-015-1318-7. [DOI] [PubMed] [Google Scholar]
  • 7.Gooiker GA, van Gijn W, Wouters MW, Post PN, van de Velde CJ, Tollenaar RA. Systematic review and meta-analysis of the volume-outcome relationship in pancreatic surgery. Br J Surg. 2011;98(4):485–94. doi: 10.1002/bjs.7413. [DOI] [PubMed] [Google Scholar]
  • 8.Hartwig W, Werner J, Jäger D, Debus J, Büchler MW. Improvement of surgical results for pancreatic cancer. Lancet Oncol. 2013;14(11):e476–e85. doi: 10.1016/s1470-2045(13)70172-4. [DOI] [PubMed] [Google Scholar]
  • 9.Hue JJ, Sugumar K, Markt SC, Hardacre JM, Ammori JB, Rothermel LD et al. Facility volume-survival relationship in patients with early-stage pancreatic adenocarcinoma treated with neoadjuvant chemotherapy followed by pancreatoduodenectomy. Surgery. 2021;170(1):207–14. doi: 10.1016/j.surg.2020.12.003. [DOI] [PubMed] [Google Scholar]
  • 10.Chu QD, Hsieh MC, Gibbs JF, Wu XC. Social determinants of health associated with poor outcome for rural patients following resected pancreatic cancer. J Gastrointest Oncol. 2021;12(6):2567–78. doi: 10.21037/jgo-20-583. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Meilleur A, Subramanian SV, Plascak JJ, Fisher JL, Paskett ED, Lamont EB. Rural residence and cancer outcomes in the United States: issues and challenges. Cancer Epidemiol Biomarkers Prev. 2013;22(10):1657–67. doi: 10.1158/1055-9965.epi-13-0404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Bertens KA, Massman JD 3rd, Helton S, Garbus S, Mandelson MM, Lin B et al. Initiation of adjuvant therapy following surgical resection of pancreatic ductal adenocarcinoma (PDAC): Are patients from rural, remote areas disadvantaged? J Surg Oncol. 2018;117(8):1655–63. doi: 10.1002/jso.25060. [DOI] [PubMed] [Google Scholar]
  • 13.Markossian TW, Hines RB. Disparities in late stage diagnosis, treatment, and breast cancer-related death by race, age, and rural residence among women in Georgia. Women Health. 2012;52(4):317–35. doi: 10.1080/03630242.2012.674091. [DOI] [PubMed] [Google Scholar]
  • 14.Walker EJ, Ko AH. Beyond first-line chemotherapy for advanced pancreatic cancer: an expanding array of therapeutic options? World J Gastroenterol. 2014;20(9):2224–36. doi: 10.3748/wjg.v20.i9.2224. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Alvarez MA, Anderson K, Deneve JL, Dickson PV, Yakoub D, Fleming MD et al. Traveling for Pancreatic Cancer Care Is Worth the Trip. Am Surg. 2021;87(4):549–56. doi: 10.1177/0003134820951484. [DOI] [PubMed] [Google Scholar]
  • 16.Kirkegård J, Ladekarl M, Fristrup CW, Hansen CP, Sall M, Mortensen FV. Urban versus rural residency and pancreatic cancer survival: A Danish nationwide population-based cohort study. PLoS One. 2018;13(8):e0202486. doi: 10.1371/journal.pone.0202486. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Canale TD, Cho H, Cheung WY. A population-based analysis of urban-rural disparities in advanced pancreatic cancer management and outcomes. Med Oncol. 2018;35(8):116. doi: 10.1007/s12032-018-1173-9. [DOI] [PubMed] [Google Scholar]
  • 18.About the National Cancer Database. ACS. https://www.facs.org/quality-programs/cancer-programs/national-cancer-database/about/#:~:text=Some%2072%20percent%20of%20all,registries%20across%20the%20United%20States. Accessed December 22, 2022. [Google Scholar]
  • 19.Rural-Urban Continuum Codes [database on the Internet]. Available from: https://www.ers.usda.gov/data-products/rural-urban-continuum-codes.aspx. Accessed: December 23, 2022
  • 20.Complex Adult and Pediatric Surgery. Hospital and Surgery Center Ratings | Leapfrog Group. https://ratings.leapfroggroup.org/measure/hospital/complex-adult-and-pediatric-surgery. [Google Scholar]
  • 21.Meguid RA, Ahuja N, Chang DC. What constitutes a “high-volume” hospital for pancreatic resection? J Am Coll Surg. 2008;206(4):622.e1–9. doi: 10.1016/j.jamcollsurg.2007.11.011. [DOI] [PubMed] [Google Scholar]
  • 22.Chang DC, Zhang Y, Mukherjee D, Wolfgang CL, Schulick RD, Cameron JL et al. Variations in referral patterns to high-volume centers for pancreatic cancer. J Am Coll Surg. 2009;209(6):720–6. doi: 10.1016/j.jamcollsurg.2009.09.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Participant User Files. ACS. https://www.facs.org/quality-programs/cancer-programs/national-cancer-database/puf/. [Google Scholar]
  • 24.Coffman A, Torgeson A, Lloyd S. Correlates of Refusal of Surgery in the Treatment of Non-metastatic Pancreatic Adenocarcinoma. Ann Surg Oncol. 2019;26(1):98–108. doi: 10.1245/s10434-018-6708-y. [DOI] [PubMed] [Google Scholar]
  • 25.Cheung MC, Yang R, Byrne MM, Solorzano CC, Nakeeb A, Koniaris LG. Are patients of low socioeconomic status receiving suboptimal management for pancreatic adenocarcinoma? Cancer. 2010;116(3):723–33. doi: 10.1002/cncr.24758. [DOI] [PubMed] [Google Scholar]
  • 26.Thobie A, Mulliri A, Bouvier V, Launoy G, Alves A, Dejardin O. Same Chance of Accessing Resection? Impact of Socioeconomic Status on Resection Rates Among Patients with Pancreatic Adenocarcinoma-A Systematic Review. Health Equity. 2021;5(1):143–50. doi: 10.1089/heq.2019.0099. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Schmidt CM, Turrini O, Parikh P, House MG, Zyromski NJ, Nakeeb A et al. Effect of hospital volume, surgeon experience, and surgeon volume on patient outcomes after pancreaticoduodenectomy: a single-institution experience. Arch Surg. 2010;145(7):634–40. doi: 10.1001/archsurg.2010.118. [DOI] [PubMed] [Google Scholar]
  • 28.Charlton M, Schlichting J, Chioreso C, Ward M, Vikas P. Challenges of Rural Cancer Care in the United States. Oncology (Williston Park) 2015;29(9):633–40. [PubMed] [Google Scholar]
  • 29.Shapiro M, Chen Q, Huang Q, Boosalis VA, Yoon CH, Saund MS et al. Associations of Socioeconomic Variables With Resection, Stage, and Survival in Patients With Early-Stage Pancreatic Cancer. JAMA Surg. 2016;151(4):338–45. doi: 10.1001/jamasurg.2015.4239. [DOI] [PubMed] [Google Scholar]
  • 30.Moaven O, Richman JS, Reddy S, Wang T, Heslin MJ, Contreras CM. Healthcare disparities in outcomes of patients with resectable pancreatic cancer. Am J Surg. 2019;217(4):725–31. doi: 10.1016/j.amjsurg.2018.12.007. [DOI] [PubMed] [Google Scholar]
  • 31.Makar M, Worple E, Dove J, Hunsinger M, Arora T, Oxenberg J et al. Disparities in Care: Impact of Socioeconomic Factors on Pancreatic Surgery: Exploring the National Cancer Database. Am Surg. 2019;85(4):327–34. [PubMed] [Google Scholar]
  • 32.Tohme S, Kaltenmeier C, Bou-Samra P, Varley PR, Tsung A. Race and Health Disparities in Patient Refusal of Surgery for Early-Stage Pancreatic Cancer: An NCDB Cohort Study. Ann Surg Oncol. 2018;25(12):3427–35. doi: 10.1245/s10434-018-6680-6. [DOI] [PubMed] [Google Scholar]
  • 33.Abraham A, Al-Refaie WB, Parsons HM, Dudeja V, Vickers SM, Habermann EB. Disparities in pancreas cancer care. Ann Surg Oncol. 2013;20(6):2078–87. doi: 10.1245/s10434-012-2843-z. [DOI] [PubMed] [Google Scholar]
  • 34.Khan H, Heslin MJ, Crook ED, Mehari K, Johnston FM, Fonseca AL. Fragmentation of Care in Pancreatic Cancer: Effects on Receipt of Care and Survival. J Gastrointest Surg. 2022;26(12):2522–33. doi: 10.1007/s11605-022-05478-8. [DOI] [PubMed] [Google Scholar]
  • 35.Raman V, Adam MA, Turner MC, Moore HG, Mantyh CR, Migaly J. Disparity of Colon Cancer Outcomes in Rural America: Making the Case to Travel the Extra Mile. J Gastrointest Surg. 2019;23(11):2285–93. doi: 10.1007/s11605-019-04270-5. [DOI] [PubMed] [Google Scholar]
  • 36.Fonseca AL, Cherla D, Kothari AN, Tzeng CD, Heslin MJ, Mehari KR et al. Association of Medicaid Expansion with Pancreatic Cancer Treatment and Outcomes: Evidence from the National Cancer Database. Ann Surg Oncol. 2022;29(1):342–51. doi: 10.1245/s10434-021-10709-4. [DOI] [PubMed] [Google Scholar]
  • 37.Morris BB, Rossi B, Fuemmeler B. The role of digital health technology in rural cancer care delivery: A systematic review. J Rural Health. 2022;38(3):493–511. doi: 10.1111/jrh.12619. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Boffa DJ, Rosen JE, Mallin K, Loomis A, Gay G, Palis B et al. Using the National Cancer Database for Outcomes Research: A Review. JAMA Oncol. 2017;3(12):1722–8. doi: 10.1001/jamaoncol.2016.6905. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Tables
NIHMS1946455-supplement-Supplementary_Tables.docx (27.8KB, docx)

RESOURCES