Association of Treatment Inequity and Ancestry With Pancreatic Ductal Adenocarcinoma Survival

Key Points

Question

To what extent are reported poor outcomes for black patients with pancreatic adenocarcinoma associated with treatment inequity?

Findings

In a contemporary cohort study using the National Cancer Database, black patients with pancreatic adenocarcinoma were found to have inferior survival and to present with more advanced disease compared with white patients. Stage-specific gaps and differences in the provision of treatment were identified, suggesting potential factors underlying the disparity in survival; however, when controlled for treatment modality and disease stage, survival was comparable.

Meaning

Inconsistencies in the care of black patients with pancreatic adenocarcinoma are associated with outcome differences observed when compared with their white counterparts.

Abstract

Importance

Pancreatic ductal adenocarcinoma (PDAC) has a higher incidence and worse outcomes among black patients than white patients, potentially owing to a combination of socioeconomic, biological, and treatment differences. The role that these differences play remains unknown.

Objectives

To determine the level of survival disparity between black and white patients in a modern PDAC cohort and whether treatment inequity is associated with such a disparity.

Design, Setting, and Participants

This cohort study used data on 278 936 patients with PDAC with database-defined race from the National Cancer Database from January 1, 2004, to December 31, 2015. The median follow-up for censored patients was 24 months. The National Cancer Database, comprising academic and community facilities, includes about 70% of new cancer diagnoses in the United States. Race-stratified receipt of therapy was the primary variable of interest. Multivariable analyses included additional demographic and clinical parameters. Data analysis was initially completed on November 30, 2018, and revised data analysis was completed on June 27, 2019.

Main Outcomes and Measures

Overall survival was the primary outcome, analyzed with Kaplan-Meier and multivariable Cox proportional hazards regression modeling.

Results

The cohort included 278 936 patients (137 121 women and 141 815 men; mean [SD] age, 68.72 [11.57] years); after excluding patients from other racial categories, 243 820 of the 278 936 patients (87.4%) were white and 35 116 of the 278 936 patients (12.6%) were black. Unadjusted median overall survival was longer for white patients than for black patients (6.6 vs 6.0 months; P < .001). Black patients presented at younger ages than white patients (15 819 of 35 116 [45.0%] vs 83 846 of 243 820 [34.4%] younger than 65 years; P < .001) and with more advanced disease (20 853 of 31 600 [66.0%] vs 135 317 of 220 224 [61.4%] with stage III or IV disease; P < .001). Black patients received fewer surgical procedures than white patients for potentially resectable stage II disease (4226 of 8097 [52.2%] vs 39 214 of 65 124 [60.2%]; P < .001) and slightly less chemotherapy for advanced disease (2756 of 4067 [67.8%] vs 17 296 of 25 227 [68.6%] for stage III disease [P = .001]; 8208 of 16 104 [51.0%] vs 58 603 of 105 616 [55.5%] for stage IV disease [P < .001]). Decreased survival for black patients persisted in multivariable modeling controlled for sociodemographic parameters (hazard ratio, 1.04 [95% CI, 1.02-1.05]). Conversely, modeling that controlled specifically for clinical parameters such as disease stage and treatment revealed a modest survival advantage (hazard ratio, 0.94 [95% CI, 0.93-0.96]) among black patients. Resection was the factor most strongly associated with overall survival (hazard ratio, 0.39 [95% CI, 0.38-0.39]).

Conclusions and Relevance

Black patients with PDAC present at younger ages and with more advanced disease than white patients, suggesting that differences in tumor biology may exist. Black patients receive less treatment stage for stage and fewer surgical procedures for resectable cancers than white patients; these findings may be only partly associated with socioeconomic differences. When disease stage and treatment were controlled for, black patients had no decrease in survival.

This cohort study uses the National Cancer Database to examine the level of survival disparity between black and white patients in a modern cohort with pancreatic ductal adenocarcinoma and whether treatment inequity is associated with such a disparity.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is the third most lethal malignant neoplasm in the United States and is expected to become the second-leading cause of cancer death by 2030.^1,2 The cause of these poor outcomes is multifactorial; a major cause is that more than 50% of patients with PDAC in the United States present with metastatic disease.¹ Although overall 5-year survival of PDAC is less than 10%, patients with a diagnosis of disease amenable to surgical resection fare the best; nearly 40% of those with early-stage disease not involving lymph nodes survive more than 5 years.^1,3,4 Conversely, certain cohorts have been found to have disparately poor outcomes. Multiple studies have shown that black patients experience 40% to 90% more incident disease than white patients.^5,6 Similarly, survival among black patients with PDAC is estimated to be 10% to 20% worse than their white counterparts.^7,8,9

The cause of these differences observed in black individuals has been hypothesized to be multifactorial, with environmental factors, socioeconomic factors, and the provision of treatment all postulated as contributing to these differences.^{8,10,11,12,13} It is not fully clear to what extent race per se is associated with these disparities. It is also unclear whether inherent biological or genetic differences in patients of predominantly African ancestry compared with non-African ancestry are associated with alterations in disease initiation and progression.^14,15,16 Furthermore, the study population in several reports that initially described this phenomenon predate the introduction of modern systemic therapies for PDAC,^17,18 and prior studies conducted at the population level have minimal detail regarding the course of treatment.^7,8,9

The primary aim of this study was to better characterize the reported difference in outcomes between black and white patients with PDAC and to assess whether treatment and sociodemographic inequities are associated with any observed difference in outcomes. A secondary aim was to identify whether potential biological factors unique to distinct ancestries were present as targets for future study. In pursuit of these goals, a large national cohort of individuals with PDAC from the National Cancer Database (NCDB) comprising patients treated in a variety of academic and community settings across multiple regions of the United States was attained and evaluated. The NCDB captures more than 70% of new cancer diagnoses in the United States, but database-defined racial categories are heterogeneous by definition.¹⁹ Given our focus on ancestry, we sought to minimize variability for comparative purposes by comparing primarily patients categorized as white or black by the NCDB. The primary outcome of interest was overall survival, with provision of cancer-directed treatments and other variables of interest such, as care setting and demographic factors, as secondary outcomes.

Methods

Cohort Selection

The NCDB participant use files were obtained for patients with pancreatic cancer from January 1, 2004, to December 31, 2015. The NCDB is a combined effort of the Commission on Cancer of the American College of Surgeons and the American Cancer Society. Patients with a final pathologic finding consisting of invasive adenocarcinoma with ductal histologic characteristics, along with major variants, were included (eTable 1 in the Supplement). All disease stages, excluding carcinoma in situ, were considered for inclusion. Patients were subdivided into white and black cohorts using their database-recorded race for descriptive and survival analysis. Additional derived variables included multiple demographic, socioeconomic, and clinical factors. This study was deemed exempt from review and the need for informed consent waived by Yale University’s Institutional Review Board.

Race and ethnicity, which are social constructs ascribed to oneself or by society at large, are heterogeneous whether self-assigned or reporter assigned on database entry. Conversely, ancestry has a geographic component and is thus more genetically perceptible. We considered patients listed as black as likely to have a predominantly African ancestry and patients listed as white as likely to have a predominantly European ancestry. The complex interplay of race and ancestry is difficult to reconcile at the population level. Nonetheless, while recognizing these confines, an analysis of ancestry must account for the parallel factors of race and ethnicity. Thus, for the purposes of study, we used database-assigned race as a surrogate for ancestry and limited the populations in the study to minimize the unavoidable heterogeneity implied by race as a variable. Specifically, nonwhite and nonblack patients were excluded from analysis for the reasons given but also, in part, owing to the smaller cohort size for nonwhite and nonblack patients and to allow more direct comparison with prior literature examining disparities among black patients with PDAC.

Descriptive and Unadjusted Survival Analysis

Median overall survival (OS) stratified by race was assessed using Kaplan-Meier methods, with statistical differences detected by the log-rank test using GraphPad Prism (GraphPad Software). Univariable analyses stratified by race were performed, with comparisons between categorical variables performed by χ² analysis. The NCDB records T, N, and M stages as well as a summary stage according to the American Joint Committee on Cancer system in place at the time the case was diagnosed; this cohort includes time periods during which both the sixth (2004-2010) and seventh edition (2010-2015) systems were in common practice, and clinical staging was generally derived via imaging studies. Age was assessed separately as a continuous variable and dichotomized for categorical analysis at a break point of 65 years. Similarly, Charlson-Deyo Comorbidity Index and cancer stage were treated as categorical variables. Patients missing relevant data were excluded on a per-analysis basis.

Multivariable Modeling and Survival Analysis

Associations between patient characteristics and OS were assessed using a Cox proportional hazards regression approach. Two separate multivariable models were constructed to scrutinize socioeconomic and demographic variables independently of clinical variables; the race variable was input in both models to investigate its association with outcomes when adjusting for those other factors. The first Cox proportional hazards regression model, hereafter referred to as the demographic model, incorporated race and socioeconomic factors including age, sex, region, urbanicity, median income, educational attainment, treating facility, and insurance coverage. Urbanicity, income, and education variables in the NCDB are derived from census data of the zip code of the patient’s home address. The second Cox proportional hazards regression model, hereafter referred to as the clinical model, incorporated race and clinical parameters, including anatomical location (pancreatic head, body, or tail), Charlson-Deyo Comorbidity Index score, T and N stages, and receipt of treatment including surgery, chemotherapy, and radiotherapy. A third pooled Cox proportional hazards regression model examined all demographic and clinical variables together, including race, to assess their relative strength of association with OS. All modeling was performed using SPSS, version 25 (IBM Corp). All P values were from 2-sided tests and results were deemed statistically significant at P < .05.

Results

Cohort Description and Survival Analysis

There were 340 780 cases of pancreatic neoplasms recorded during the 11-year study period (the case to patient ratio was 1:1). Only those representing invasive ductal carcinoma (n = 292 604) were included (eTable 1 in the Supplement). Of these cases, 243 820 (83.3%) were diagnosed in white patients and 35 116 (12.6%) in black patients. The remaining 13 668 patients (4.7%) fell into other racial categories and were excluded from analysis. Thus, 278 936 patients with invasive PDAC comprised the study cohort. Descriptive statistics for the cohort are shown in eTable 2 in the Supplement. Unadjusted analysis of median OS across all stages of disease found that black patients had a small but significant decreased survival compared with white patients (6.0 vs 6.6 months; P < .001) (Figure 1).

Figure 1. Unadjusted Overall Survival From Diagnosis of Pancreatic Ductal Adenocarcinoma .

Black patients (n = 35 116) had worse overall survival with pancreatic ductal adenocarcinoma compared with white patients (n = 243 820) (6.6 vs 6.0 months; P < .001).

The mean (SD) age for presentation was significantly younger for black patients than for white patients (66.2 [11.7] vs 69.1 [11.5] years; P < .001) (eFigure in the Supplement). As such, 15 819 of 35 116 black patients (45.0%) presented at ages younger than 65 years compared with 83 846 of 243 820 white patients (34.4%) (P < .001) (Figure 2A). Also, black patients presented with more advanced disease: 20 853 of 31 600 black patients (66.0%) had American Joint Committee on Cancer stage III or IV disease at diagnosis compared with 135 317 of 220 224 white patients (61.4%) (P < .001) (Figure 2B). Black patients also had a higher burden than white patients of comorbid disease (13 728 of 35 116 [39.1%] vs 80 267 of 243 820 [32.9%] with a Charlson-Deyo Comorbidity Index score of ≥1; P < .001) (Figure 2C). Finally, black patients were much more likely than their white counterparts to be treated at academic or tertiary care centers (16 993 of 34 785 [48.9%] vs 101 541 of 242 241 [41.9%]; P < .001) (Figure 2D).

Figure 2. Age and Stage at Presentation, Burden of Comorbidity, and Treating Facility.

A, Age at diagnosis. Black patients presented at a younger age than white patients, with 15 819 of 35 116 (45.0%) vs 83 846 of 243 820 (34.4%) presenting younger than 65 years of age (P < .001). B, American Joint Committee on Cancer (AJCC), eighth edition, stage. Black patients presented with a more advanced disease stage than white patients, with 20 853 of 31 600 (66.0%) vs 135 317 of 220 224 (61.4%) presenting with stage III or IV disease (P < .001). C, Charlson-Deyo Comorbidity Index. Black patients presented with a higher burden of comorbidities than did white patients, with 13 728 of 35 116 (39.1%) vs 80 267 of 243 820 (32.9%) presenting with a Charlson-Deyo Comorbidity Index score of ≥1 (P < .001). D, Treating facility. Black patients were treated more often at academic centers than were white patients, with 16 993 of 34 785 (48.9%) vs 101 541 of 242 241 (41.9%) being treated at academic centers (P < .001).

Analysis of Treatment Modality

Examining the receipt of treatment by race revealed significant differences in the provision of therapy. When diagnosed with stage I or II disease, patients with PDAC are usually considered candidates for surgical resection. Among black patients with stage I disease, surgery was performed at a statistically similar rate to their white counterparts (874 of 2650 [33.0%] vs 6941 of 19 783 [35.1%]; P = .057). However, among patients with stage II disease, black patients underwent surgery less frequently than white patients (4226 of 8097 [52.2%] vs 39 214 of 65 124 [60.2%]; P < .001) (Figure 3). Furthermore, among patients with advanced disease, who are predominantly treated with systemic therapy, black patients received chemotherapy less often than white patients. This difference was most pronounced among those with stage IV disease (2756 of 4067 [67.8%] vs 17 296 of 25 227 [68.6%] for stage III disease; P = .001; 8208 of 16 104 [51.0%] vs 58 603 of 105 616 [55.5%] for stage IV disease; P < .001) (Figure 3). The rate of expected treatment was low across the entire treatment cohort, with only 51 255 of 95 654 patients with stage I or II disease (53.6%) undergoing surgery and only 86 863 of 151 014 patients with stage III or IV disease (57.5%) receiving chemotherapy.

Figure 3. Treatments Administered at Each Disease Stage.

For early stages of disease, black patients underwent surgery less frequently than white patients (874 of 2650 [33.0%] vs 6941 of 19 783 [35.1%] for stage I disease; P = .057; 4226 of 8097 [52.2%] vs 39 214 of 65 124 [60.2%] for stage II disease; P < .001). In addition, black patients received chemotherapy slightly less often than white patients for stage II disease (4895 of 7679 black patients [63.7%] vs 40 552 of 61 988 white patients [65.4%]; P < .01). For more advanced disease stages, black patients received chemotherapy less frequently (2756 of 4067 [67.8%] vs 17 296 of 25 227 [68.6%] for stage III disease; P = .001; and 8208 of 16 104 [51.0%] vs 58 603 of 105 616 [55.5%] for stage IV disease; P < .001). Although few patients underwent resection for stage III disease, black patients were also less likely than white patients to undergo surgery at this stage (240 of 4213 black patients [5.6%] vs 2123 of 26 131 white patients [8.2%]; P < .001).

Multivariable Survival Analysis

Two models were designed for Cox proportional hazard regression analysis using different subsets of the available variables. The demographic model, shown in Table 1, demonstrated that, when adjusting for socioeconomic and health care access factors, black race had a modest but significant association with reduced survival (hazard ratio [HR], 1.04 [95% CI, 1.02-1.05]). Treatment at an academic cancer center was associated with survival (HR, 0.67 [95% CI, 0.66-0.68]) relative to treatment in a community facility. Similar but less pronounced associations were seen for insurance type (private insurance being most favorable), living in an area with higher educational attainment, and living in an area with higher median income (Table 1).

Table 1. Multivariable Cox Proportional Hazards Regression Model of Overall Survival: Demographic and Socioeconomic Factors.

Variable	HR (95% CI)	P Value
Race
White	1 [Reference]
Black	1.04 (1.02-1.05)	<.001
Age, y
<65	1 [Reference]
≥65	1.20 (1.18-1.21)	<.001
Sex
Male	1 [Reference]
Female	0.96 (0.95-0.97)	<.001
Diagnosis era
2004-2009	1 [Reference]
2010-2015	0.84 (0.83-0.85)	<.001
Region
Northeast	1 [Reference]
South	1.00 (0.99-1.02)	.58
Midwest	1.05 (1.03-1.06)	<.001
West	1.01 (1.00-1.03)	.06
Urbanicity
Metropolitan	1 [Reference]
Urban or rural	0.95 (0.94-0.96)	<.001
Income, median, $
<38 000	1 [Reference]
38 000-47 999	0.94 (0.93-0.96)	<.001
48 000-62 999	0.92 (0.91-0.94)	<.001
≥63 000	0.88 (0.86-0.89)	<.001
No high school degree, %
≥21.0	1 [Reference]
13.0-20.9	1.01 (0.99-1.02)	.31
7.0-12.9	0.98 (0.97-1.00)	.003
<7.0	0.95 (0.94-0.97)	<.001
Facility
Community	1 [Reference]
Comprehensive community	0.90 (0.89-0.92)	<.001
Integrated network	0.80 (0.78-0.82)	<.001
Academic	0.67 (0.66-0.68)	<.001
Insurance
None	1 [Reference]
Medicaid	0.97 (0.94-1.00)	.09
Other government-provided insurance	0.86 (0.82-0.90)	<.001
Medicare	0.97 (0.94-0.99)	.01
Private	0.78 (0.76-0.80)	<.001

Abbreviation: HR, hazard ratio.

A second clinical model focusing on tumor-specific and treatment-specific variables is shown in Table 2. This analysis demonstrated that, when adjusting for clinical factors alone, black race was no longer independently associated with mortality and was associated with survival when controlling for disease stage at presentation and other clinical factors such as treatment (HR, 0.94 [95% CI, 0.93-0.96]). Greater burden of comorbidity, tumor located in the pancreatic tail, and higher T and N stages were all significantly associated with mortality. Surgical resection had the strongest association with survival among the clinical parameters (HR, 0.39 [95% CI, 0.38-0.39]), followed by receipt of chemotherapy (HR, 0.45 [95% CI, 0.44-0.45]).

Table 2. Multivariable Cox Proportional Hazards Regression Model of Overall Survival: Clinical Factors.

Variable	HR (95% CI)	P Value
Race
White	1 [Reference]
Black	0.94 (0.93-0.96)	<.001
Charlson-Deyo Comorbidity Index score
0	1 [Reference]
1	1.15 (1.14-1.16)	<.001
2	1.34 (1.31-1.36)	<.001
3	1.60 (1.56-1.65)	<.001
Pancreas site
Head	1 [Reference]
Body	0.97 (0.96-0.99)	<.001
Tail	1.09 (1.08-1.11)	<.001
Other	1.04 (1.03-1.06)	<.001
T stage–clinical
1	1 [Reference]
2	1.22 (1.19-1.25)	<.001
3	1.15 (1.12-1.18)	<.001
4	1.11 (1.08-1.14)	<.001
N stage–clinical
0	1 [Reference]
1	1.08 (1.07-1.10)	<.001
Surgery
No	1 [Reference]
Yes	0.39 (0.38-0.39)	<.001
Chemotherapy
No	1 [Reference]
Yes	0.45 (0.44-0.45)	<.001
Radiotherapy
No	1 [Reference]
Yes	0.95 (0.94-0.97)	<.001

Abbreviation: HR, hazard ratio.

Finally, an all-inclusive model using variables from both the demographic and clinical models was constructed to assess their relative degree of association with survival simultaneously (eTable 3 in the Supplement). Findings from this analysis largely mirrored the separate models, with black race again having a modest association with survival, as in the clinical model (HR, 0.95 [95% CI, 0.93-0.96]).

Discussion

The primary aim of this study was to assess the magnitude and association of sociodemographic and treatment inequities with survival among patients with PDAC stratified by race. In this national cohort derived from the NCDB during an 11-year study period, patients identified as black were found to have decreased median OS compared with their white counterparts. The absolute survival reduction was modest but significant among a cohort of nearly 279 000 patients. Univariable and multivariable analyses of demographic, disease, and treatment variables supported a multifactorial rationale for this finding. There were significant differences between the groups in presentation, with black patients presenting at younger ages, with more comorbidities, and at more advanced disease stages. Overall, black patients were less likely to receive cancer-directed treatment at all stages of disease except for stage I. Black patients were also significantly more likely to be treated at academic facilities.

Treatment deficiencies and poorer survival among black patients compared with their nonblack counterparts have been previously demonstrated across an array of malignant neoplasms.^11,13 Studies examining PDAC specifically at the institutional, regional, and national levels suggest that survival is approximately 10% to 20% worse for blacks but may be gradually decreasing over time.^{7,8,9,14,15,20} The absolute difference in median OS of 10.4% observed in our analysis (0.6 months) (Figure 1) is consistent with these data.

This study adds several new findings for consideration. A detailed description of stage-specific treatment of PDAC is beyond the scope of this article, but, in general, early-stage (I and II) disease is treated with surgical resection, which is often rendered in combination with chemotherapy. Stage III disease in our study population is chiefly locally advanced disease (defined by American Joint Committee on Cancer 6th and 7th editions) potentially amenable to operative therapy if downstaged. However, most cases of stage III PDAC and nearly all cases of stage IV PDAC are treated exclusively with chemotherapy.²¹ In our analysis, black patients with both stage II and III disease were treated surgically far less often than white patients. Furthermore, chemotherapy was less frequently administered to black patients with stages II to IV PDAC (Figure 3).

Most previous reports have found similarly decreased rates of operative therapy among black patients with PDAC.^7,13,20 Singal and colleagues⁸ found the opposite when analyzing the Surveillance, Epidemiology, and End Results (SEER) database for a 20-year period starting in 1988. These findings may reflect the older study period during which time surgical practice for pancreatic cancer was evolving. A more modern cohort using the SEER database from 2004 to 2011 examined by Shapiro et al²² mirrored the lower rate of surgery among black patients observed in our study, although their analysis was limited to early-stage disease. Another recent study by Moaven et al¹⁶ also limited to resectable tumors and using the NCDB found a similar decrease in receipt of surgery among black patients. Our study aligns with these findings but also questions the association between race, health care access, and survival; black patients in our analysis were treated more frequently at academic hospitals yet, at most stages, received less cancer-directed treatment.

Our findings were particularly surprising in light of recent studies by Lufti et al¹⁴ and Swords et al,²³ which demonstrated that treatment at low-volume or nonacademic hospitals was an independent risk for not undergoing surgery for potentially resectable PDAC. In this context, the current data are striking, in that our cohort included all disease stages and treatment modalities, including patients managed with or without surgery. This is a notable strength relative to other recent studies, which have tended to exclusively focus on resectable tumors. Conversely, such a large cohort is necessarily heterogeneous and thereby limits determination of definitive conclusions in subgroups of interest. A related strength is that we considered all Committee on Cancer–approved sites (academic and community) to capture risk factors that extended beyond high-volume academic centers. The NCDB data set cannot be used to determine individual motivations behind the observed decrease in surgical rates, but these data suggest that demographic and socioeconomic factors may not fully explain treatment choice and receipt. A previous NCDB-based study by Shah et al²⁴ demonstrated that black patients were less likely to accede to surgery when it was offered. Clinician or health system bias could also possibly be associated with these findings, but such individual sources of bias cannot be assessed at the national level. Although these themes cannot be conclusively demonstrated with the data available here, there is certainly a growing appreciation of the importance of cultural context and sensitivity as a barrier to providing appropriate surgical care.²⁵ Greater effort in this area may be crucial to overcoming the disparately poor rate of resection observed among black patients with PDAC.

In the demographic model controlling for socioeconomic and care access factors, black race remained an independent risk factor for poor survival. The increased incidence of PDAC in patients with predominantly African ancestry would suggest a biological susceptibility to PDAC in black patients. However, analysis of risk factors and environmental exposures suggests that at least some of the increased incidence may be associated with higher rates of smoking, diabetes, obesity, and other health behaviors, as well as family history of PDAC.^5,10,12,26 Silverman et al⁵ suggest that nearly all the increased incidence among black patients with PDAC can be associated with these risks, whereas Arnold and colleagues¹² reached the opposite conclusion. The present study furthers this debate. Although the later stage at which black patients receive a diagnosis of PDAC might be explained by differences in health care access, the considerably lower age at which they present compared with their white counterparts suggests that biological susceptibility may partially contribute to the risk of PDAC in patients of African ancestry. Disease-specific survival stratified by ancestry would be intriguing to examine, but it is unavailable in the NCDB data set. However, with median survival less than 1 year for patients with PDAC and the rapid progression of disease when it is metastatic or recurrent, OS is likely strongly correlated with disease-specific survival. Future studies examining this phenomenon should correlate findings with genetic haplotype to objectively characterize ancestry.

An important and encouraging finding from the clinical model was that black race was associated with a better survival outcome when controlling for stage at presentation and receipt of treatment. Similarly, in the combined model including all available factors in a unified multivariable analysis, black race was not associated with worse outcome. In other words, black patients might experience significant survival benefit if provision of stage-appropriate treatment were improved. Increased attention to the disparities in treatment highlighted in this and other studies may lead to improved survival for black patients.

Limitations

Even with statistical adjustment for suspected confounding variables, unmeasured factors may influence treatment decision-making in our study, and the registry database does not provide the ability to study the clinical reasoning for each case. However, despite identified effect sizes that are modest in magnitude, highly statistically significant analyses can be derived given the large sample size available. Investigating nonwhite and nonblack minorities may further shed light on disparities in modern PDAC care; to our knowledge, these populations have been sparingly studied. In any case, in a modern cohort of patients with pancreatic cancer, there are significant disparities in outcomes when stratified by race. Further examination of the equitable availability of novel combinatorial, targeted, and neoadjuvant treatment strategies as they continue to evolve are merited. Finally, efforts examining the biological role of ancestry may improve these disparities. Perhaps most important, ensuring equitable provision of care in both early-stage and late-stage disease should be a major focus in the future.

Conclusions

This retrospective study of the NCDB supports prior studies that have shown outcomes disparities for black patients with PDAC. The disparity may be associated in part with genetic differences in tumor behavior, but a significant portion of the difference in outcomes appears to be associated with treatment inequities. Future work will need to better characterize these inequities to durably improve disparately poor outcomes that persist for black patients with pancreatic malignancy.

Supplement.

eTable 1. Included Pancreatic Ductal Adenocarcinoma Subtypes

eTable 2. Descriptive Variables of All Assessed Cases Stratified by Race

eTable 3. Multivariable Cox Proportional Hazards Regression Pooled Model of Inclusive of All Available Variables

eFigure. Age of Presentation With Pancreatic Adenocarcinoma