There were 3,777 non-Hispanic white and black patients with locoregional pancreatic cancer in the overall cohort. The mean age of patients was 75.8 ± 6.6 years, with the majority of patients being female (59%). 3,425 patients were white (91%) and 352 were black (9%). The Charlson comorbidity score was zero in 62% of patients. In 58% of cases, patients had an identified primary care physician. Pancreatic cancers were located in the head of the pancreas in 74%, the body/tail in 10%, and in 16% the site was not specified. Cross-sectional abdominal imaging via magnetic resonance imaging (MRI) or computed tomography (CT) was performed in 95% of patients.
Based on the SEER extent of disease codes, 79% were candidates for surgical resection (resectable) and 21% were unresectable due to involvement of the superior mesenteric vein, portal vein, or organs outside the typical field of resection. Only 29% of the cohort was resected. When stratified by resectability based on SEER EOD codes, 33% of the resectable patients and 13% of the unresectable patients were resected.
Baseline Characteristics: Black vs. White Patients
shows the differences in demographic factors, tumor characteristics, evaluation, social support, and socioeconomic status between white and black patients. When compared to white patients, black patients were similar in age, but were more likely to be female, unmarried, and have more comorbidities. Black patients were significantly more likely to be in the lowest quartile for both income and education.
Patient and Tumor Characteristics by Racial Status (N=3,777)
Overall, the tumor characteristics were similar between the two groups. Most importantly, the number of patients that were resectable based on SEER EOD codes was similar between the two groups. There were no differences in the location of the cancers within the pancreas or the proportion of black and white patients with unknown tumor site or nodal status. In over 60% of patients the data on nodal status are missing. As this information is largely learned after surgery and pathologic examination, nodal status was not included in the multivariate models evaluating the factors affecting surgical resection.
Black patients were less likely than white patients to have a cross-sectional abdominal imaging study, to be evaluated by an oncologist, or to have a primary care physician. They were equally as likely to be evaluated by a gastroenterologist.
In locoregional pancreatic cancer, as with any other disease process requiring surgical resection, the first step in achieving surgical resection is evaluation by a surgeon. 31% of all black patients with locoregional pancreatic cancer (N=3,777) and 29% of black patients with potentially resectable pancreatic cancers (N=3,002) never receive surgical evaluation, the first step in the process. In an unadjusted model, Black patients with locoregional pancreatic cancer were 32% less likely than white patients to be evaluated by a surgeon (unadjusted OR 0.68, 95% CI, 0.53 – 0.86, unadjusted percentages in ). 3,614 of the 3,777 patients had complete data for the final logistic regression model. After adjusting for demographic factors, tumor characteristics, preoperative evaluation, social support, and socioeconomic status black patients were 43% less likely to be evaluated by a surgeon (OR 0.57, 95% CI, 0.42 – 0.77, ).
Logistic Regression Analysis Modeling the Odds of Evaluation by a Surgeon*
In addition to race, several other factors also influenced surgical evaluation (). Older patients and patients with higher Charlson comorbidity scores were less likely to undergo preoperative surgical evaluation. Patients who were resectable based on SEER EOD codes and had cross-sectional abdominal imaging were more likely to be evaluated by a surgeon. When compared to patients with cancers in the head of the pancreas, those with cancers in the body/tail were equally as likely to be evaluated by a surgeon. Evaluation by a medical oncologist was also predictive of surgical evaluation, while evaluation by a gastroenterologist had the opposite effect. Patients who were identified as having a regular primary care physician were 37% more likely to undergo surgical evaluation. SEER region was also a significant predictor of surgical evaluation, with an overall type 3 P-value of <0.0001 for differences among the SEER regions (Connecticut, Detroit, Greater California, Hawaii, Iowa, Kentucky, Los Angeles, Louisiana, Metropolitan Atlanta, New Jersey, New Mexico, Rural Georgia, San Francisco-Oakland, San Jose-Monterey, and Seattle-Puget Sound. The individual odds ratios for each region depend on the reference group and are not included in the table. However, it is critical to understand there are regional differences and control for this when evaluating the effect of race on surgical evaluation.
Surgical Resection: Overall and Following Surgical Evaluation
For the overall cohort (n=3,77) regardless of whether surgical evaluation occurred, 30% of white patients but only 21% of black patients with locoregional pancreatic cancer were resected, for an unadjusted OR of 0.64 (95% CI, 0.49 – 0.84). In the subset of 3,002 patients with technically resectable tumors based on SEER EOD codes (regardless of surgical evaluation), 34% of white patients were resected versus 24% of black patients (unadjusted OR = 0.61, 95% CI, 0.46 – 0.81).
There were 2,872 black and white patients who underwent surgical evaluation. Data on all covariates were available in all 2, 872 patients. The unadjusted odds ratio of surgical resection after surgical evaluation for black patients was 0.71 (95% CI, 0.53 – 0.94). After adjusting for all the same factors in the previous models, black patients were 33% less likely to be resected even after surgical evaluation (OR=0.67, 95% CI, 0.48 – 0.95, ).
Factors Influencing Surgical Resection After Evaluation by Surgeon*
An analysis was also performed on the 2,267 resectable patients who were evaluated by a surgeon. When considering resectable patients only, black patients were 40% less likely to be resected following surgical evaluation (OR 0.60, 95% CI, 0.41 – 0.87).
Other factors influencing surgical resection after surgical evaluation are shown in . In the adjusted model, with each increasing year of diagnosis, patients were 10% more likely to be resected. Older patients, those with higher Charlson comorbidity score, and patients with an unknown tumor site were less likely to undergo surgical resection. Patients with tumors in the body and tail were more likely to be resected. Patients who were resectable based on SEER EOD codes were 5.7 times more likely to be resected. Gender, abdominal imaging, evaluation by a gastroenterologist, marital status, primary care physician, SEER region, income, education, and population did not influence surgical resection following surgical evaluation.
The Kaplan-Meier overall 2-year survival for black versus white patients is shown in . In this unadjusted survival curve, black patients had worse overall survival with a median survival of 8.6 months and a 2-year survival rate of 14% compared to 9.8 months and 18% in white patients (P=0.01). The same is true when evaluating resectable patients only (N=3,002). The median survival for resectable black patients was 9.1 months, with a 2-year survival rate of 15%. The median survival for white patients was 9.9 months, with a 2-year survival rate of 19% (P=0.02).
Figure 1 Overall Kaplan-Meier actuarial 2-year survival for black and white patients. Black patients (gray line) had a median survival rate of 8.6 months and a 2-year survival rate of 14% compared to white patients (black line) who had a median survival rate of (more ...)
In a Cox proportional hazards model, the unadjusted hazard ratio for black patients was 1.17 (95% CI, 1.04 – 1.32). Surgical resection accounted for a significant proportion of the observed racial difference; when resection alone was added to the model, race was no longer a significant predictor of survival (HR 1.08, 95% CI, 0.96 – 1.22). In the final model adjusting for demographics, resectability, socioeconomic status, social support, and year of diagnosis race remained non-significant (HR 1.00, 95% CI, 0.87 – 1.14). In the final model, younger age (P<0.0001), lower Charlson comorbidity score (P=0.0005), negative lymph nodes, and surgical resection (P<0.0001) were independent predictors of better survival.