By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Presented in part at the Annual Meeting for the Society of Surgical Oncology; Chicago, Illinois; March 13-16, 2008; and the American Society of Clinical Oncology Gastrointestinal Cancers Symposium; Orlando, Florida; January 25-27, 2008.
Blacks are affected disproportionately by pancreatic adenocarcinoma and have been linked with poor survival. Surgical resection remains the only potential curative option. If surgical disparities exist, then they may provide insight into outcome discrepancies.
Patients with pancreatic adenocarcinoma were identified using the National Cancer Institute's Surveillance, Epidemiology, and End Results data from 1992 to 2002. Univariate analyses were used to compare demographics, tumor characteristics, and surgical data; and logistic regression was used to determine independent predictors for recommendation/performance of surgery. Kaplan-Meier survival was assessed, and a Cox proportional hazards model was used to examine adjusted predictors of survival.
In total, 27,828 patients were identified; 81.4% were white, 11.5% were black, 7.2% were of other race. White patients and black patients presented with similar stage and had surgery recommended at similar rates (34.5% vs 34%, respectively; P = .57). Black patients underwent fewer resections (10.6% vs 12.7%; P < .001). Multivariate analysis confirmed that black patients were less likely to undergo resection (adjusted odds ratio, 0.69; 95% confidence interval [95% CI], 0.57-0.84). Overall, black patients had worse univariate survival. The survival among black patients who underwent resection did not differ statistically from the survival of similar white patients, although the median survival trended lower (11 months vs 13 months; P = .13). In a multivariate Cox model, black race predicted worse survival (hazards ratio, 1.11; 95% CI, 1.07-1.16), and pancreatic resection was protective (hazards ratio, 0.56; 95% CI, 0.53-0.59).
Pancreatic adenocarcinoma disproportionately affects blacks. The average annual age-adjusted incidence rates are higher for blacks (15.2 per 100,000) than for whites (11.4 per 100,000).1 Once patients are diagnosed with pancreatic adenocarcinoma, black race is linked with poorer survival.2, 3 Racial disparities in survival have been reported in colorectal, breast, and lung cancer as well as pancreatic cancer.4-7 At the national level, racial disparities in healthcare have been highlighted as a national priority.8
Cancer disparities may stem from complex patient, social, and institutional factors. It has been proposed that some patient factors, including preferences,9 socioeconomic status (SES),10 and biologic differences,11 contribute to observed disparities. In addition, treatment-related differences, including patient-physician interactions,12 and structural barriers, including insurance coverage and the type and location of the institution providing care,13 have been associated with worse outcomes for minorities. The relative contribution of surgical care to these disparities is not clear.14 Racial differences have been described for rates of surgical procedures,15 for types of surgeries offered,16 and for both surgical and overall mortality.4, 17
Despite advances in neoadjuvant and adjuvant therapy, pancreatic resection remains the only potential curative option in the treatment of pancreatic adenocarcinoma.18 To determine whether disparities exist in whether blacks are recommended for and undergo pancreatic resection and in their subsequent survival, we performed a retrospective observational study within the Surveillance, Epidemiology, and End Results (SEER) registry. Our primary objectives were to: 1) determine predictors of recommendation and receipt of surgery, and 2) compare resection rates and overall survival among blacks and whites.
MATERIALS AND METHODS
Seventeen tumor registries participate in the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program, including Connecticut, Iowa, New Mexico, Utah, Hawaii, Alaska, Kentucky, New Jersey, Louisiana, California, Detroit, Atlanta, Seattle/Puget Sound, Georgia, and Arizona. Combined, these sites provide a representative survey of the United States approximating 26% of the population. This sampling is comparable to the general US population for several characteristics, including those of potentially at-risk groups. It is noteworthy that the percentages of minorities represented in the SEER registry either closely approximate or exceed those that occur nationally.19 Information within the registry includes patient demographics (age, sex, race/ethnicity), disease-related data (cancer site, tumor characteristics, lymph node involvement), and first course of treatment (cancer-directed surgery, radiation). Survival data are available from the SEER database through its linkage with the National Death Index, a centralized index of death records obtained through state vital statistics offices.
A cohort of patients with pancreatic tumors was identified in the SEER registry between the years 1992 and 2002. This cohort was then limited by selecting records that had International Classification of Diseases for Oncology, 3rd edition codes consistent with adenocarcinoma (8140). Patients who were identified through autopsy and death certificate only were deleted from the cohort because of the lack of data. SEER racial information was used to classify patients as white, black, or other. “Other” was a composite group that included American Indians/Alaskan Natives, Asians or Pacific Islanders, and other/nonspecified individuals.
Tumor Characteristics and Surgical Data
Clinical stage information is recorded in SEER and includes tumor size, lymph node involvement, presence of metastases, and tumor location. In addition, SEER codes patients by using “SEER historic stage,” which categorizes patient disease in 1 of 4 ways: localized, regional, distant, or unknown. Localized pancreatic cancer is defined as tumor in situ or tumor confined to the pancreas. Regional disease includes invasion into adjacent structures or locoregional lymph node involvement. Distant disease demonstrates metastases outside the locoregional area.20 SEER historic stage was used in our analyses.
For the SEER Program, cancer-directed treatment is limited to procedures directed toward cancer tissues. If a specific therapy normally affects, controls, changes, removes, or destroys cancer tissue, then it is classified as definitive treatment, and in the case of surgery, is defined as cancer-directed surgery.21 For each patient, SEER records whether cancer-directed surgery was recommended, and this classification was used in our analyses. Using these data, patients were categorized as either “recommended for cancer-directed surgery” (Group R) or “not recommended for cancer-directed surgery” (Group NR). Currently, pancreatic resection remains the only available curative option for patients with pancreatic adenocarcinoma. Therefore, to further describe the patients who were recommended for cancer-directed surgery (Group R), we subcategorized patients according to whether they underwent pancreatic resection (Group R/YES) or not (Group R/NO). These groups are illustrated schematically in Figure 1. For patients in Group R/NO, the reason for no resection was reported and included: patient refusal, other cancer-directed surgery, and unknown reason. Resection was defined by SEER site-specific surgery codes (1992-1997) and SEER site-specific surgery of primary site codes (≥1998) (Table 1). All types of pancreatectomy (total, partial, distal pancreatectomy, and pancreaticoduodenectomy) were pooled into a single composite category of “resection.”
Table 1. Pancreas Surveillance, Epidemiology, and End Results (SEER) Procedure Codes Used
Site-specific Surgery: 1988-1997
Site-specific Surgery of Primary Site: 1998 and Later
NOS indicates not otherwise specified; CDS, cancer-directed surgery.
Local or partial surgical excision of pancreas
Local excision of tumor, NOS
Radical regional (partial) pancreatectomy with lymph node dissection and adjacent soft tissue resection
Partial pancreatectomy, NOS
Total pancreatectomy with/without splenectomy
Total pancreatectomy and subtotal gastrectomy or duodenectomy
Subtotal gastrectomy, duodenectomy with complete or partial pancreatectomy with/without splenectomy
Local or partial pancreatectomy and duodenectomy
Without distal/partial gastrectomy
With partial gastrectomy
Incisional, needle, or aspiration biopsy other than primary site
Incisional, needle, or aspiration biopsy of primary site
Exploratory and incisional, needle, or aspiration biopsy
Bypass surgery and incisional, needle, or aspiration biopsy
Surgery of regional and/or distant sites/lymph nodes only
Patient demographics and tumor characteristics for each race were compared by using chi-square tests for categorical variables and Student t tests for continuous variables. Unadjusted survival was compared by subgroup stratification, including race, recommendation, and resection, using the Kaplan-Meier method.22
To ensure that patients in Group R/NO, who were recommended for cancer-directed surgery but did not undergo pancreatic resection, were not simply patients who had unsuccessful attempts, thereby potentially making intergroup comparisons less valid, a sensitivity analysis was performed. In this sensitivity analysis, we examined Group R (recommended for cancer-directed surgery) on an intent-to-treat basis. Group R patients were classified based on whether any operation was performed. Patients who underwent any type of surgery, including pancreatectomy, biopsies, metastasectomies, palliative bypasses, and feeding tubes, were included. This sensitivity analysis investigated the extreme possibility that all patients who underwent any type of operation represented attempted resections. Patients who underwent any type of surgical procedure were pooled into 1 Group R/ANY SURGERY (recommended for cancer-directed surgery/any surgery) and compared with Group R/NO SURGERY (recommended for cancer-directed surgery/no surgery).
Significant predictors of resection on univariate analyses and those with biologic plausibility were then incorporated in multivariate logistic regression to determine whether race was an independent predictor of resection. A second multivariate logistic regression analysis was constructed to evaluate predictors of recommendation for pancreatic resection. A Cox proportional hazards model was used to determine the effect of pancreatic resection on survival and was adjusted for potential confounders, including race, age, sex, SEER stage, and tumor location. All data analysis was performed using SAS statistical software (version 9.1.3; SAS Institute Inc, Cary, NC). Significance was accepted at the P < .05 level.
Sample Size and Patient Demographics
There were 27, 828 patients identified in the SEER registry with pancreatic adenocarcinoma during 1992 through 2002 after exclusion criteria were applied. The demographic characteristics of the cohort are listed in Table 2. Men comprised 50.6% of the patients. Whites comprised the majority of patients (81.4%), followed by blacks (11.5%), and others (7.2%). The mean age of patients in the cohort was 68.9 years. Patient age, sex, marital status, tumor location, resection status, and year of diagnosis differed significantly between whites, blacks, and others. Black patients tended to be younger (aged 65.9 years; P < .001) and more often were women (52.2%; P < .001) compared with whites and others.
Table 2. Patient Characteristics and Baseline Univariate Analyses by Race for All Patients With Pancreatic Adenocarcinoma, 1992-2002
No. of Patients (%)
SEER indicates the National Cancer Institute Surveillance, Epidemiology, and End Results Program.
There were 1004 patients with unknown radiation status.
No significant difference in SEER historic stage was detected between blacks, whites, or others. The majority of patients had distant disease at first diagnosis (56.2%). Of the patients who had potentially resectable disease, 6.5% had localized disease, and 28.7% had regional disease. Unstaged patients comprised 8.6% of cases. Pancreatic head tumors were most common (53.4%), followed by nonhead pancreatic tumors, (39.3%) and multifocal disease (7.3%). Pancreatic head tumors were identified more commonly in black patients (54.5%; P = .01).
Recommendation and Receipt of Pancreatic Resection
Patients recommended for cancer-directed surgery
Overall, 34% of patients (n = 9461) were included in the group that was recommended for cancer-directed surgery (Group R) (Table 3). On univariate analysis, younger age, early stage (localized and regional), and tumor location in the pancreatic head were associated with both recommendation for surgical management and undergoing resection. In total, 3448 patients underwent pancreatectomy (Group R/YES), whereas 6013 patients did not undergo resection (Group R/NO), as demonstrated in Table 4.
Table 3. Univariate Analysis: Pancreatic Resection and Recommendation for Resection, 1992-2002
R indicates the group that was recommended for cancer-directed surgery (CDS); R/YES, the group that was recommended for CDS/pancreatic resection; R/NO, the group that was recommended for CDS/no pancreatic resection; NR, the group that was not recommended for CDS; SEER, the National Cancer Institute Surveillance, Epidemiology, and End Results Program.
R indicates the group that was recommended for cancer-directed surgery; R/YES, the group that was recommended for cancer-directed surgery/pancreatic resection; R/NO, the group that was recommended for cancer-directed surgery/no pancreatic resection; CDS, cancer-directed surgery; NR, the group that was not recommended for cancer-directed surgery.
P values for comparisons of black patients versus white patients.
P values for global comparisons.
Reason for no resection
Blacks and whites were recommended for surgery at similar rates (34% vs 34.5%, respectively; P = .57). Blacks underwent significantly fewer resections than whites (10.6% [n = 337] vs 12.7% [n = 2885]; P < .001). Blacks were significantly more likely to refuse recommended surgery than whites (1.8% vs 1.3%, respectively; P = .01).
Patients not recommended for cancer-directed surgery
The majority of patients were not recommended for cancer-directed surgery (Group NR; 66%; n = 18,367). Of these patients, 5612 (31%) underwent surgical procedures, including palliative procedures, feeding tubes, and bypasses.
Predictors of Recommendation and Pancreatic Resection
Significant covariates that we identified on univariate analyses and those variables that had biologic plausibility (age, sex, marital status, race, tumor location, resection, and year of diagnosis) were incorporated into a logistic regression model to evaluate the predictors of recommendation for cancer-directed surgery. Localized disease, nonhead tumor location, and more recent year of diagnosis were positive predictors of receiving a recommendation for cancer-directed surgery, whereas other race was a negative predictor, as shown in Table 5. Race was not associated with recommendation for surgery for blacks (referent group, whites; adjusted odds ratio [AOR], 1.00; P = .93). Compared with distant disease, localized and regional disease were predictive of recommendation for surgery (referent group, distant disease; localized disease: AOR, 1.93; 95% confidence interval [95% CI], 1.71-2.16; regional disease: AOR, 1.29; 95% CI, 1.20-1.39). Nonhead tumor location was predictive of recommendation for surgery (referent group, pancreatic head location; nonhead location: AOR, 1.52; 95% CI, 1.53-2.06). With the period from 1992 to 1994 as the referent period, more recent year of diagnosis was predictive of recommendation for surgery for all year groups(1995-1997: AOR, 1.14; 95% CI, 1.03-1.25; 1998-2000: AOR, 2.81; 95% CI, 2.58-3.06; and 2001-2002: AOR, 1.43; 95% CI, 1.31-1.56).
Table 5. Multivariate Analysis: Predictors of Recommendation for Resection and Predictors of Resection in Patients With Pancreatic Adenocarcinoma, 1992-2002
In the logistic regression model that was used to evaluate resection, black race was a negative predictor of undergoing pancreatectomy (referent group, whites; blacks: AOR, 0.69; 95%CI, 0.57-0.84). Similar to recommendation for surgical management, locoregional disease, tumor location, and more recent year of diagnosis were predictive of undergoing pancreatic resection.
Survival for all patients
The median survival for all patients with pancreatic adenocarcinoma was 4 months, and univariate survival is shown in Table 6, including stratified subgroup analyses. Increasing age was correlated with worsening survival (aged >70 years: median survival, 3 months; aged <50 years: median survival, 6 months; P < .001).
Table 6. Univariate Survival in Months for All Patients With Pancreatic Adenocarcinoma, 1992-2002
NA indicates not applicable; R, recommended for cancer-directed surgery; NR, not recommended for cancer-directed surgery; R/YES, recommended for cancer-directed surgery/pancreatic resection; R/NO, recommended for cancer-directed surgery/no pancreatic resection.
The log-rank test was used to assess difference in overall Kaplan-Meier curves.
Advanced tumor stage was associated significantly with worse survival. The median survival for patients with distant disease was 3 months, whereas, for patients with local and regional disease, the median survival was 8 months (P < .001).
Survival by recommendation for surgery and pancreatic resection
Patients who received a recommendation for surgical management had significantly better survival than those who did not received such recommendation (Group R: median survival, 6 months; Group NR: median survival, 4 months; P < .001). Patients who underwent pancreatic resection (Group R/YES) had significantly improved median survival (13 months) compared with patients who did not undergo resection (Group R/NO; median survival, 4 months; P < .001).
To approximate an “intent-to-treat” comparison, a sensitivity analysis was performed as described earlier using the extreme assumption that all patients in Group R/ANY SURGERY (n = 4647) underwent surgery with the intent to resect. That group had a median survival of 9 months, which remained significantly different from the group R/NO SURGERY group (median survival, 3 months; P < .001) (Table 6).
Survival by race
Kaplan-Meier curves demonstrated worse overall survival for blacks (log-rank test; P < .001) (Fig. 2). Recommendation for and undergoing pancreatic resection were associated with significantly improved survival for all races. After resection, black median survival improved from 4 months to 11 months (P < .001), whereas white median survival improved from 4 months to 13 months (P < .001). After resection, black survival no longer differed statistically from white survival, although the median survival trended lower (resected blacks, 11 months; resected whites, 13 months; P = .13) (Fig. 3). To exclude the possibility of confounding by other race, univariate survival analyses were repeated comparing blacks with whites and excluding the other race category, and no changes were observed in overall survival. The results displayed include all race categories.
A multivariate Cox proportional hazards model was used to identify factors that independently influenced survival (Table 7). The covariates that were included in the model were sex, age (as a continuous variable), marital status, race, resection, SEER historic stage, and tumor location. Increasing age was predictive of decreased survival (hazards ratio [HR], 1.02; 95% CI, 1.01-1.02). Patients who had nonhead pancreatic tumors had worse survival compared with patients who had pancreatic head tumors (HR, 1.52, 95% CI, 1.01-1.16).
Table 7. Multivariate Cox Proportional Hazards Model of Survival for Patients With Pancreatic Adenocarcinoma, 1992-2002
HR indicates hazard ratio; 95% CI, 95% confidence interval; SEER, the National Cancer Institute Surveillance, Epidemiology, and End Results Program.
Black race was identified as an independent predictor of worse survival compared with white race (HR, 1.11; 95% CI, 1.07-1.16). Pancreatic resection was associated independently with improved overall survival for patients with pancreatic adenocarcinoma (HR, 0.56; 95% CI, 0.53-0.59).
By using the national SEER registry, we examined the effect of race, and specifically the difference between blacks and whites, on pancreatic resection rates and survival in patients with pancreatic adenocarcinoma. Patient and tumor characteristics were investigated, and we paid particular attention to recommendations for surgery and whether recommended surgery was performed. Survival was evaluated by race, stage, and resection status.
We observed that blacks and whites presented with pancreatic adenocarcinoma with similar stage and were recommended for cancer-directed surgery at similar rates. However, despite being recommended equally, blacks underwent significantly fewer pancreatic resections than whites. Unadjusted survival analyses demonstrated that black patients had worse overall survival compared with white patients. Both recommendation and receipt of pancreatic resection were associated with significantly improved survival for all races on stratified subgroup analysis. After resection, crude survival for blacks and whites did not differ significantly, although the median survival for blacks continued to trend lower. Multivariate analyses demonstrated a survival disadvantage for blacks despite adjusting for resection, whereas undergoing pancreatic resection was protective for all races.
After resection, the median survival for patients in our cohort was comparable to the survival reported in a recent California study (13 months vs 13.3 months, respectively).23 In that state-based cohort, those investigators did not detect a significant difference in survival between whites and blacks after resection, although the median survival continued to trend lower for blacks.
Recent reports have demonstrated a decrease in 1-year and 5-year mortality rates for patients with pancreatic cancer who have locoregional disease,24 whereas population studies indicate that pancreatic resection rates are increasing.25 Multiple studies have reported that undergoing pancreatic resection is predictive of increased survival among patients with pancreatic cancer.23, 26, 27 Consistent with our findings, Bilimoria et al reported that black race was a negative predictor of undergoing pancreatic resection using the National Cancer Database, which represents data from American College of Surgeons-approved tumor registries.28 Conversely, those authors observed that blacks were to be less likely than whites to be offered surgery; whereas, in the current study, we observed no significant difference in the recommendation for surgery between blacks and whites. An earlier SEER-based study (1988-1999), indicated that pancreatic resection was a positive predictor of survival, whereas black race, age, and being a man rather than a woman had a negative influence on survival.27 Our more current study period demonstrated the persistence of black race as a negative predictor of survival after adjusting for covariates, including resection.
Disparities in incidence, surgical mortality, and overall survival for black patients with pancreatic adenocarcinoma have been well documented in the literature.2, 3, 29, 30 It has been proposed that individual patient risk factors, including cigarette smoking, alcohol use, elevated body mass index, and diabetes mellitus, account in part for the increased incidence of pancreatic adenocarcinoma among black patients.29 It also has been demonstrated that blacks have higher surgical mortality while undergoing pancreatic surgery in comparison to whites.30 Once patients are diagnosed with pancreatic adenocarcinoma, black race has been linked with poorer survival compared with whites.2, 3
The factors underlying worse survival for blacks are unclear. Population-based studies have proposed that SES and treatment differences contribute in part to this survival disparity.2, 10 Because of SEER database limitations, the impact of SES on survival could not be assessed in our analyses.
Patient level preferences and biases may play a role in seeking and receiving care. Lack of quality communication between providers and patients may contribute to disparities in health between blacks and whites.12 It has been reported that black patients may receive different advice from their physicians than white patients.31 In addition, it has been demonstrated that refusal of recommended surgical intervention for pancreatic cancer is 3 times higher for blacks than for whites.26 Our findings support this trend, as 1.8% of blacks in our cohort refused surgery compared with 1.3% of whites. However, this accounts for <20% of the observed difference in resection rates.
Currently, surgical resection offers the only potential curative option for patients with pancreatic adenocarcinoma. Although pancreatectomy does not entirely explain the survival discrepancy between blacks and whites, our work suggests that providing recommended care, including pancreatic resection, may improve overall survival for black patients with pancreatic adenocarcinoma.
The SEER cancer registry provides a wealth of information regarding cancer cases at the national level, including information on patient demographics, tumor characteristics, and surgical information. However, there are some limitations to using these types of data. Individual patient level information including SES, medical comorbidities, patient performance, and hospital course information, is not available.
Further limitations include the inability to determine whether operations were performed for cure or palliation, and perisurgical mortality is not included. Neoadjuvant or adjuvant chemotherapy information is not available, and much of the data on radiation are missing. Information on surgical referral, consultation, and individual providers is not available. In addition, multiple revisions of the SEER registry have resulted in different time periods with differing coding schemas.
Despite these limitations, the SEER registry provides a mechanism with which to study population-based cancer in a diverse patient population. Compared with whites, blacks have a higher incidence of pancreatic adenocarcinoma, present with similar disease stages, and are recommended for cancer-directed surgery equally. However, this recommendation is not translated into surgical resection for a significant number of black patients. The current results suggest that the survival disparity may be narrowed by ensuring that black patients receive recommended pancreatic resection. Further work needs to be done to elucidate why blacks are undergoing fewer resections than their white counterparts, including individual patient characteristics, provider differences, and hospital characteristics. Identifying these specific barriers to care may provide a framework in which to focus efforts transitioning patients from a diagnosis of pancreatic adenocarcinoma, to the operating room for definitive therapy, and ideally to increased survival.
Conflict of Interest Disclosures
Supported by the Evans-Allen-Griffin Fellowship (Melissa M. Murphy), by the Pancreatic Cancer Alliance (Jessica P. Simons and Jennifer F. Tseng), and by the American Surgical Association Foundation, a Pancreatic Cancer Action Network-Samuel Stroum American Society of Clinical Oncology Young Investigator Award, and a Howard Hughes Early Career Award (all to Jennifer F. Tseng).