Race, Socioeconomic Status, Treatment, and Survival Time among Pancreatic Cancer Cases in California

Pancreatic cancer was the fourth leading cause of cancer deaths in U.S. men and women during 1999 to 2003 (1). Overall survival (OS) for pancreatic cancer is poor, particularly in the setting of advanced or unresectable disease, and only 10% to 20% of pancreatic cancer patients have potentially resectable tumors (2, 3). The estimated 5-year OS for advanced pancreatic cancer with current systemic therapy is <3% (4). Even with optimal therapy median OS for advanced pancreatic cancer is estimated at just 6.4 months (5). However, clinical trials and surgical series overestimate survival for this deadly malignancy compared with population-based studies, which offer more accurate survival estimates. Predictors of survival among pancreatic cancer patients have been identified but there are some remaining gaps.

Among pancreatic cancer cases, survival differences by race have been observed. African-American race has been associated with poor survival among pancreatic cancer cases in population-based studies (6, 7). Additionally, low socioeconomic status (SES) has been found to be an independent risk factor for mortality in a population-based study of pancreatic adenocarcinoma cases (7). However, this study did not investigate SES differences by race, or the modifying effect of SES on survival for each race. Thus, it is currently not known whether the observed poor survival for African-American pancreatic adenocarcinoma cases is due to differences in SES. The interaction of SES with race on survival has not been determined, in part because SES variables are not readily available in population-based databases, such as the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program.

Adjuvant chemoradiation is a predictor of improved survival among nonmetastatic pancreatic adenocarcinomas (7), but the independent contributions of radiation therapy and chemotherapy (also not available in SEER) have not been assessed for advanced pancreatic cancer cases (i.e., the majority of pancreatic cancer cases). Thus, it is not known whether treatment disparities related to SES are responsible for the observed poor survival of African-American pancreatic cancer patients.

The present study was designed to quantify the relevant prognostic factors for survival among pancreatic adenocarcinoma cases, including analysis by race, the contribution of SES on survival for each race, and the effect of treatment with radiation and chemotherapy on survival for each major ethnic group using data from the California Cancer Registry (CCR).

We conducted a retrospective, case-only analysis of pancreatic cancer cases in the CCR database. The CCR is the largest contiguous area population-based cancer registry in the world, collecting >130,000 new cancer cases yearly in California as described elsewhere (8). The state of California legally mandated cancer reporting in 1988; standardized data collection procedures and quality control procedures have been in place ever since (9-12). The CCR, containing cancer data from the entire state of California, is also part of the National Cancer Institute's SEER program. Case reporting is estimated at >98% for the entire state of California (13), and due to data completeness, accuracy, and timeliness, CCR has received the highest level of certification from the North American Association of Central Cancer Registries (14). Data were abstracted from medical and laboratory records by trained tumor registrars according to Cancer Reporting in California: Vol. 1, Abstracting and Coding Procedures for Hospitals (11). Tumor site and histology were coded according to criteria specified by the WHO in International Classification of Diseases for Oncology (15). In situ and invasive tumors were identified using pancreatic cancer SEER primary site code (21100) and International Classification of Diseases-O-3 histology codes as follows: adenocarcinoma (8140, 8141, 8143, 8144, 8190, 8211, 8261, 8262, 8263, 8290, 8440, 8503, 8560, and 8570), mucinous tumors (8470, 8471, 8472, 8473, 8480, and 8481), and endocrine (8150-8155 and 8246), as described previously (6), in addition to carcinoma-not otherwise specified (carcinoma-NOS; 8010). Data were obtained on 24,735 incident pancreatic cancer cases during the period 1989 to 2003. Recorded data included demographic information (age, gender, and ethnicity), stage at presentation, histology, CCR region, SES, and vital status. SES is denoted as a single index variable in CCR using statewide measures of education, income, and occupation from census data as described previously (16). The SES variable used is a composite index based on principle component analysis of census block level CCR data linked to census data assessing: education level, median household income, proportion <200% poverty level, median house value, median rent, percentage employed, and percentage with blue-collar employment. Cases diagnosed before 1996 were linked to 1990 census data, and cases diagnosed after 1996 were linked to 2000 census data as described previously (8). Geocoding to a particular street address in CCR is not available for 5.5% of cases, and in these instances, cases were randomly allocated to census block groups within their county of origin as described previously (8). The SES variable in CCR and its relationship to race/ethnicity have been analyzed in previous epidemiologic studies of the breast, colon, prostate, ovary, nasopharynx, cutaneous melanoma, and pancreas (8, 16-21). Quintiles for the SES score were used for analysis, with SES-1 and SES-5 denoting the lowest and highest SES quintiles, respectively.

Treatment during the first course of therapy was ascertained using available data from CCR to determine whether cases underwent surgical resection (including type of resection), radiation therapy, or chemotherapy. Cause of death was recorded according to the International Classification of Diseases criteria in effect at the time of death (22). Hospital registrars contacted cases annually, and CCR staff annually reviewed state death certificates to identify deceased registry cases. The last date of follow-up was either the date of death or the last date the case was contacted. Smoking history was obtained by abstracting the text fields from our regional database (CCR regions 7 and 10, covering the geographic areas of Orange, Imperial, and San Diego counties), via a customized data mining program using SAS 9.1 statistical software (SAS Institute, Inc., Cary, NC) as done previously (23). Cases with any documented history of tobacco smoking at the time of diagnosis were classified as “ever smokers.” Cases with documentation of no smoking history were classified as “never smokers.” Cases lacking documented information on smoking history were excluded from the relevant analyses.

The clinical characteristics, including age, gender, race, SES quintile, treatment status, smoking status, and SEER summary stage were analyzed with Pearson's χ² test or Fisher's exact test for categorical and dichotomous variables and ANOVA with Tukey's post hoc test for comparison of continuous variables. Life tables and Kaplan-Meier curves were generated for cases with each histologic subtype of pancreatic cancer. OS comparisons between groups were analyzed with the log-rank test. Multivariate survival analysis and analysis for effect modification were done using Cox proportional hazards ratios (HR). All statistical analyses were conducted using SAS 9.1 statistical software. Statistical significance was assumed for a two-tailed P < 0.05.

This research study involved analysis of existing data from the CCR database without subject intervention. No identifiers were linked to subjects. Therefore, the study was approved by the University of California Irvine Institutional Review Board under the category “exempt” status (Institutional Review Board #2006-5217).

There were 24,735 incident cases of pancreatic cancer analyzed, including adenocarcinomas (n = 17,326), mucinous tumors (n = 1,692), carcinoma-NOS tumors (n = 4,926), and endocrine tumors (n = 791). Table 1 displays demographic characteristics for the major pancreatic cancer histologic subtypes. There was an even distribution of adenocarcinomas and mucinous tumors among men and women but a male predominance in endocrine tumors and a female predominance in carcinoma-NOS. A smaller proportion of endocrine cases were African-Americans. Most cases did not undergo surgery. Interestingly, 97.7% of carcinoma-NOS cases did not have surgery, although 11.7% had localized disease. The majority of cases did not receive radiation therapy or chemotherapy. Endocrine cases had the greatest proportion of cases from the highest SES quintile.

Overall, 1.8% of cases refused surgery, 1.8% refused radiation treatment, and 4.1% refused chemotherapy. The proportion of cases refusing surgery (P = 0.11), radiation therapy (P = 0.64), and chemotherapy (P = 0.18) was similar across all ethnic groups analyzed in this study. Decreasing SES quintile was associated with an increased likelihood of refusing treatment with surgery (P_trend = 0.002), radiation therapy (P_trend < 0.0001), or chemotherapy (P_trend = 0.016).

Among adenocarcinoma cases, the majority of African-Americans and Hispanics were in the lowest SES quintiles compared with Caucasians, who were predominantly in the higher SES quintiles (Table 2). For example, 83.6% of those in the highest SES quintile were Caucasians compared with only 2.6% African-Americans and 5.7% Hispanics. Asians were more evenly distributed among the different SES quintiles among adenocarcinoma cases. Although statistical differences were noted for various stage distributions across the five SES quintiles, no clinically important relationship between stage at presentation and SES was apparent. However, increasing SES quintile was associated with surgery, radiation therapy, and chemotherapy for adenocarcinoma cases (Table 2). The demographic characteristics by each SES quintile for pancreatic mucinous, endocrine, and carcinoma-NOS histologies are presented elsewhere (Supplementary Table S1).

OS was significantly greater for pancreatic endocrine cases compared with pancreatic adenocarcinoma, mucinous tumors, and carcinoma-NOS (Fig. 1). One-year and 5-year survival rates were as follows: endocrine tumors (63% and 33%), mucinous tumors (22% and 6%), adenocarcinoma (16% and 2%), and carcinoma-NOS cases (10% and 2%; P < 0.0001). Median OS for each of the pancreatic cancer histologic subtypes is as follows: endocrine tumors [OS, 27 months; 95% confidence interval (95% CI), 23-33], mucinous tumors (OS, 4 months; 95% CI, 4-5), adenocarcinomas (OS, 4 months; 95% CI, NR), and carcinoma-NOS [OS, 2 months; 95% CI, not reached (NR)]. Compared with adenocarcinoma cases, endocrine tumors (HR, 0.28; 95% CI, 0.26-0.30; P < 0.0001) and mucinous tumors (HR, 0.83; 95% CI, 0.79-0.88; P < 0.0001) had improved OS in a multivariate survival analysis after adjustment for age, year of diagnosis, gender, stage, and race. Carcinoma-NOS cases had an increased risk of death (HR, 1.21; 95% CI, 1.17-1.25; P < 0.0001) compared with adenocarcinoma cases in the above multivariate survival analysis.

To determine the relevant effectors of risk (including treatment variables and SES) by ethnicity in this model, separate multivariate survival analyses were conducted among adenocarcinoma cases, as displayed in Table 3. After adjustment for age, year of diagnosis, and gender, African-Americans (HR, 1.14; 95% CI, 1.08-1.21) and Hispanics (HR, 1.06; 95% CI, 1.01-1.11) had an increased risk of death compared with Caucasians (Table 3). After further adjustment for stage at presentation, the increased risk of death persisted for African-Americans (HR, 1.13; 95% CI, 1.07-1.20) and Hispanics (HR, 1.08; 95% CI, 1.03-1.13) compared with Caucasians (Table 3). After further adjustment for surgery, radiation therapy, chemotherapy, and SES, the risk of death for African-Americans (HR, 1.00; 95% CI, 0.94-1.06) and Hispanics (HR, 0.97; 95% CI, 0.93-1.02) was not statistically different than for Caucasians (Table 3). Adjustment for CCR registry (i.e., the 10 regions that comprise CCR) had minimal or no effects on the HR for race in each of the above multivariate models (data not shown). Non-Chinese Asians were found to have improved adjusted OS compared with Caucasians after adjustment for each of the aforementioned clinical factors (HR, 0.85; 95% CI, 0.79-0.91). SES, treatment with chemotherapy, and radiation therapy were significantly associated with improved OS among adenocarcinoma cases. On univariate survival analysis, higher SES was associated with improved OS (median OS of 3, 4, 4, 4, and 5 months for SES-1, SES-2, SES-3, SES-4, and SES-5, respectively; P < 0.0001). Complete multivariate survival tables for each major pancreatic histologic subtype are presented elsewhere (Supplementary Table S2).

Overall, 23,405 of the 24,735 (94.6%) cases in this study died. Cause of death analysis revealed that 19,979 (85.4%) deaths were due to pancreatic cancer. Infection, unknown cause, and chronic obstructive pulmonary disease comprised the majority of the nonpancreatic cancer-related deaths.

Tobacco history (i.e., ever smoker status versus never smoker status) was obtained on a subset of cases with available text files from our regional registry (i.e., the Cancer Surveillance Programs of Orange, San Diego, and Imperial counties). Overall, 4,107 incident cases were available for analysis, including 2,975 adenocarcinomas, 264 mucinous tumors, 122 endocrine tumors, and 746 carcinoma-NOS cases. Smoking status could be determined for 66% of these cases. Due to the limited sample size, only adenocarcinoma cases were considered to have sufficient numbers for detailed analysis. Among the 2,043 adenocarcinoma cases with available smoking history, 1,357 (66.4%) were ever smokers and 686 (33.6%) were never smokers. Significant differences in the proportion of ever smokers were noted across the major ethnicities (P = 0.0001): African-Americans (44 of 54 cases, 81.5%), Caucasians (1,074 of 1,575 cases, 68.2%), non-Chinese Asians (75 of 127 cases, 59.1%), Hispanics (153 of 262 cases, 58.4%), Chinese (10 of 20 cases, 50%), and Others (1 of 5 cases, 20%). A greater proportion of male pancreatic adenocarcinoma cases were ever smokers (804 of 1,095 cases, 73.4%) compared with females (553 of 948, 46.4%; P < 0.0001). Smoking status was not associated with SES among pancreatic adenocarcinoma cases (P = 0.25).

Multivariate survival analysis was done to analyze the effect of smoking on survival among adenocarcinoma cases, after adjustment for age at diagnosis, year of diagnosis, gender, race, stage, surgery, radiation therapy, chemotherapy, and SES (Table 4). Smoking was found to be a significant adverse prognostic factor for survival (HR, 1.22; 95% CI, 1.11-1.35; P < 0.0001), as was advanced stage at presentation (P < 0.0001) and remote year of diagnosis (P = 0.0001). Race and gender were not significant prognostic factors for survival. A significant decrease in the risk of mortality was associated with surgery (HR, 0.48; 95% CI, 0.41-0.57; P < 0.0001), radiation therapy (HR, 0.86; 95% CI, 0.77-0.96; P < 0.006), and chemotherapy (HR, 0.56; 95% CI, 0.51-0.61; P < 0.0001). There was a trend toward decreased mortality with increasing SES quintile (HR, 0.97; 95% CI, 0.94-1.00; P = 0.07). After removing smoking status from the above model, higher SES quintile was significantly associated with improved survival (HR, 0.97; 95% CI, 0.94-0.99; P = 0.03).

This population-based analysis of pancreatic cancer cases in California confirms the findings of previous studies and reveals poor survival for African-Americans and Hispanics compared with Caucasians among adenocarcinoma cases. However, our study is the first to show that, among pancreatic adenocarcinoma cases, these survival differences between racial groups are partially explained by treatment differences and differences in SES. In particular, the increased risk of death noted for African-American and Hispanic cases in multivariate analyses unadjusted for treatment variables and SES (Table 3) is predominantly explained by treatment disparities. Improved SES was shown to be associated with improved OS and also with treatment involving surgery, radiation therapy, and chemotherapy among pancreatic adenocarcinoma cases. Our results indicate that the major survival benefit associated with SES likely stems from treatment differences.

Our findings correlate with other reports showing that SES affects survival in pancreatic cancer patients. One retrospective cohort study of 396 nonmetastatic pancreatic adenocarcinoma cases using SEER-Medicare–linked data analyzed prognostic factors for poor survival (7). Treatment in a nonteaching hospital, lack of adjuvant chemoradiation therapy, lymph node metastasis, poor histologic grade, tumor size >2 cm diameter, and low SES were independently associated with poor survival. African-American race was associated with poor survival on univariate analysis but not after adjustment for SES, treatment with adjuvant chemoradiation, and the above clinical variables. Interpretation of these data are limited, as the Medicare database is comprised predominantly of cases >65 years of age, and metastatic cases were not included in the analysis. In a population-based analysis of pancreatic cancer in all stages by various histologic subtypes, Fesinmeyer et al. (6) noted poor survival for African-Americans among adenocarcinoma and mucinous histologies after adjustment for age, year of diagnosis, gender, stage, and surgery status. Interestingly, the adjusted risk of death in that study (HR, 1.11) was identical to the observed HR in our study after adjustment for the same factors (i.e., including adjustment for surgery, stage at presentation, but no adjustment for radiation therapy, chemotherapy, or SES; Table 4). The SEER database does not include data on chemotherapy treatment or SES. Thus, the above investigators could not account for the effects of these variables on survival. Additionally, the above study did not examine survival for Hispanics.

SES was strongly associated with ethnicity, stage, treatment, and survival among pancreatic adenocarcinoma cases. Stage- and treatment-related differences across different SES strata indicate disparities in cancer care among pancreatic cancer cases. Recently, a separate population-based analysis of pancreatic cancer cases in Alabama revealed that African-Americans have similar survival characteristics to Caucasians after adjustment for surgery, medical treatment, tumor location, and age at diagnosis (24). Although the effects of SES were not investigated in that report, treatment-related factors were addressed. For all stages of pancreatic cancer cases, African-American cases were less likely to receive surgery or chemotherapy, treatment variables that were shown to be associated with improved survival rates. Interestingly, a greater proportion of African-American pancreatic cancer cases refused treatment with surgery, radiation therapy, or chemotherapy. The authors suggest that such findings may indicate different attitudes and different patient-physician interactions among minority cancer patients. Treatment refusal was less common in our study compared with the one above (i.e., only 1.8%, 1.8%, and 4.1% overall for surgery, radiation therapy, and chemotherapy, respectively). No statistically significant differences in the proportion of cases refusing treatment were noted across the six races investigated here. However, a small but statistically significant decrease in the proportion of cases refusing treatment by SES quintile was noted for surgery, radiation therapy, and chemotherapy. In a smaller study of CCR data from 1994 to 2000, other investigators have reported that SES (as a dichotomized variable) was not significantly associated with survival among pancreatic adenocarcinoma cases on adjusted analysis (18). However, cases with low SES were noted to be less likely to receive surgical treatment. Effect modification to address the interaction of SES and treatment variables on survival of African-American cases was not done in that study, and the investigators comment in their discussion that further research is needed to explore apparent treatment disparities (18).

In a subset of adenocarcinoma cases from our regional CCR database (i.e., covering the following regions: Orange, San Diego, and Imperial counties), we have shown that a history of tobacco use is independently associated with poor survival. The effect of smoking on survival has been examined by several investigators, with differing results. For example, it has been reported that smoking increases the chance of detection at an earlier age but has no effect on survival (25). Conversely, in separate, large cohort studies, the risk of death from pancreatic cancer was associated with high levels of tobacco use (26, 27). The association of smoking with poor survival in these studies may actually reflect the risk of developing pancreatic cancer, since the mortality for pancreatic cancer (once diagnosis is established), is so high. Our results displaying the independent, adverse effect of smoking on survival in this case-only analysis of pancreatic cancer cases support the conclusions drawn from investigators in the aforementioned cohort studies. Importantly, the effects of smoking were independent of SES in our subset analysis of pancreatic adenocarcinoma cases.

This epidemiologic study shares limitations of other population-based analyses, including the lack of independent histologic review of specimens, variable sources of reporting and diagnostic methods, and broad staging criteria used (i.e., SEER summary staging instead of American Joint Committee on Cancer staging methods). Analysis of all possible confounding factors related to survival among pancreatic cancer cases cannot be done, thus important aspects affecting survival have not been accounted for including access to health care, insurance coverage, and comorbid conditions. For example, higher pancreatectomy case loads per hospital (2, 28) and receiving treatment at a teaching hospital versus a nonteaching hospital (7) have been associated with improved survival for pancreatic cancer patients. We are not able to account for these factors using CCR data. The data on tobacco use are somewhat limited because only 66% of cases could be classified, in a relatively small subset of adenocarcinoma cases. Furthermore, the tobacco use variable was restricted to “ever smokers” or “never smokers,” thus the quantity of tobacco use (i.e., pack-years) could not be calculated using this type of analysis. Nonetheless, we believe it is important to do such analyses to estimate the effects of smoking on survival among pancreatic adenocarcinoma cases.

Our findings show that African-Americans and Hispanics with pancreatic adenocarcinoma fare similarly to Caucasians, but only after accounting for treatment differences and differences in SES. Such data indicate a need for exploration of the causes for these treatment disparities among African-Americans and Hispanics and appropriate targeted intervention strategies. Tobacco smoking was found to be associated with poor outcome among pancreatic adenocarcinoma cases, suggesting the potential for further research and intervention efforts at smoking cessation among pancreatic cancer patients.