Are patients of low socioeconomic status receiving suboptimal management for pancreatic adenocarcinoma?




The objective of this study was to define the effects of socioeconomic status (SES) and other demographic variables on outcomes for patients with pancreatic adenocarcinoma.


Florida cancer registry and inpatient hospital data were queried for pancreatic adenocarcinoma diagnosed from 1998 to 2002.


In total, 16,104 patients were identified. Low SES (LSES) patients were younger at diagnosis (P < .001) but presented with similar disease stage and tumor grade. LSES patients were less likely to receive surgical extirpation (16.5% vs 19.8%; P < .001), chemotherapy (30.7% vs 36.4%; P < .001), or radiotherapy (14.3% vs 16.9%; P = .003). Among surgical patients, 30-day mortality was significantly higher (5.1% vs 3.7%; P < .001) and overall median survival was significantly worse (5.0 months vs 6.2 months; P < .001) in the LSES cohorts. Although surgical patients who were treated at teaching facilities (TF) did significantly better; an increased 30-day surgical mortality (2.2% vs 1.3%; P < .001) and decreased median survival (5 months for poverty level >15% vs 6.2 months for poverty level <5%; P < .001) also were observed for patients of LSES. In a multivariate analysis that corrected for patient comorbidities, significant independent predictors of a poorer prognosis included LSES (hazard ratio [HR], 1.09); treatment at a non-TF (HR, 1.09); and failure to receive surgical extirpation (HR, 1.92), chemotherapy (HR 1.41), or radiation (HR 1.25).


Patients of LSES were less likely to receive surgical extirpation, chemotherapy, or radiation and had significantly higher perioperative and long-term mortality rates. A greater understanding of the barriers to providing optimal care and identifying means for improving successful delivery of therapies to the poor with pancreatic cancer are needed. Cancer 2010. © 2009 American Cancer Society.

Pancreatic adenocarcinoma is the fourth leading cause of cancer death in the United States and will cause the death of more than 35,000 Americans in 2008.1 Although the overall incidence has not changed over the last several decades, pancreatic adenocarcinoma remains a very common malignancy among patients of low socioeconomic status (LSES).2-5 Overall, worse outcomes for LSES patients have been identified for several other malignancies.6-16 Causes for these disparate outcomes likely are multifactorial and are caused in part because of lack of insurance and poor access to healthcare—including preventive services and inadequate monitoring and care of chronic conditions.11, 17 Understanding the relative contribution of demographic factors to such disparities is critical in identifying potential ways to address disparate outcomes.18-20

By using a large, population-based cancer registry with information on patient comorbidities, demographic, and treatment data, we sought to determine treatment and outcome differences for patients with pancreatic cancer based on race, ethnicity, and SES. In identifying the factors associated with altered outcomes, we hoped to identify potential treatment and outcome differences and begin to define possible points of intervention that may help erase potential disparities.


The Florida Cancer Data System (FCDS) is a population-based registry of all cancer cases diagnosed or treated in the state of Florida, which represents about 6% of the total US population. The most current 2007 FCDS dataset was used to identify all incident cases of pancreatic adenocarcinoma diagnosed in the state of Florida from 1998 to 2002. In total, 16,104 primary cases, according to the International Classification of Disease for Oncology, were diagnosed antemortem and extracted for analysis. Individuals who had missing information on any of the key variables were excluded from univariate analysis. In addition, non-Florida residents were not included in the analysis, because follow-up, particularly survival information, may be inaccurate in up to 10% of such patients.

The FCDS dataset was enhanced with data linked from the Florida Agency for Health Care Administration (AHCA), which maintains 2 databases (Hospital Patient Discharge Data and Ambulatory Outpatient Data) on all patient encounters within hospitals and freestanding ambulatory surgical and radiation therapy centers in Florida. All hospitals have been required to report all discharges and outpatient encounters to AHCA since 1987. The AHCA data used in this study contain comorbid conditions, diagnoses, and procedures performed during every hospitalization or outpatient encounter in the state of Florida for the period from 1998 to 2002. Patients in the FCDS and AHCA datasets were linked on the basis of unique identifiers. These matches were confirmed with the patient's date of birth and sex. Then, postal codes listed in the FCDS-AHCA database were used to determine community poverty levels according to the US Census Bureau report.21

Medical facilities were defined as either teaching facilities (TFs) or nonteaching facilities (NTFs) based on recognition as a teaching institution by the Association of American Medical Colleges (AAMC). At the time of the current report, there were 11 AAMC-recognized TFs in the state of Florida. The data from FCDS were tabulated to determine the number of surgical resections for pancreatic adenocarcinoma performed at each institution in the state of Florida during the study period. Medical facilities were grouped into tertiles based on the number of surgeries with curative intent performed during the study period. The upper one-third of institutions were classified as high-volume centers (HVCs), and the lower two-thirds were classified as low-volume centers (LVCs).

The FCDS database records the type of surgical treatment patients undergo for their disease. Surgical resections that were included in the current analysis were (FCDS code for surgery of primary site 30) partial pancreatectomy, ie, distal pancreatectomy; (code 35) local or partial pancreatectomy and duodenectomy (code 36) WITHOUT distal/partial gastrectomy or (code 37) WITH partial gastrectomy; (code 40) total pancreatectomy; (code 60) total pancreatectomy and subtotal gastrectomy or duodenectomy; (code 70) extended pancreatectomy; (code 80) pancreatectomy, not otherwise specified (NOS); and (code 90) surgical resection, NOS. Overall, distal pancreatectomies accounted for 8.8% of the cohort that underwent surgical resection, whereas >50% of the cohort underwent pancreatectomy and duodenectomy with partial gastrectomy. For statistical purposes, we grouped those who underwent surgery for curative intent, as designated by the FCDS, into 1 cohort, because only 18% of the entire study population underwent a curative procedure.

Statistical analysis was performed with the Statistical Package for Social Sciences (version 15; SPSS, Inc., Chicago, Ill). For group comparisons of categorical variables, the chi-square test was used. Because the FCDS collects only the primary cause of death, we analyzed overall survival instead of disease-specific survival. Overall survival was calculated from the time of the initial diagnosis to the date of either death or last contact. Median survival rates were calculated using the Kaplan-Meier method. By using significant variables that were identified from univariate analysis, a multivariate Cox proportional-hazards regression was used to determine the risk of death associated with pancreatic adenocarcinoma. Patient demographics, clinical characteristics, and patient comorbidities were included as categorical variables in the analysis.

The staging criteria used by the FCDS is consistent with the National Cancer Institiute's Surveillance, Epidemiology, and End Results summary staging and is different from the (tumor, lymph node, and metastasis [TNM]) staging guidelines, because TNM staging was not available through this cancer registry. In the current report, local staging represents disease that does not extend beyond the primary organ, whereas positive lymph nodes at the time of resection were considered regional disease. Distant disease represents the detection or identification of metastases during the perioperative period.


Patient Demographics and Clinical Variables

Patient demographics and clinical variables are summarized in Table 1. The cohort was divided evenly among men (50.8%) and women (49.2%). The majority of patients were Caucasian (90.3%), non-Hispanic (89.2%), and aged ≥65 years (68%). Approximately half of the cohort (43.2%) reported either current tobacco use or a history of tobacco use, whereas a larger proportion (89.8%) reported no alcohol consumption. Greater than half of the patients (56.7%) were identified according to the US Census data as living in a community in which ≤10% of the residents were at or below Florida's poverty level. Patients most frequently had private insurance, Medicare, or a combination of Medicare and Medicaid (31.9%, 35.8%, and 22.4%, respectively).

Table 1. Demographic, Social, and Clinical Characteristics
CharacteristicEntire Cohort, n=16,104Subset Analysis
Area PovertyFacility
% of TotalP% of TotalP% of TotalP
No.% of Total<5%5%-10%10%-15%>15%HVCLVCTFNTF
  1. HVC indicates high-volume center; LVC, low-volume center; TF, teaching facility; NTF, nonteaching facility; AA, African American; NOS, not otherwise specified.

Mean age at diagnosis, y70.1 68.570.9<.00165.971<.001
 Women792249.247.249.351.350.7 48.449.6 47.149.6 
Age, y             
 40-65494430.730.927.629.533.5 35.628.4 43.228.3 
 >6510,9576868.271.469.164.5 62.970.4 54.770.6 
 AA13128. 7.18.7 9.48 
 Other1460. 0.90.9 10.9 
 Non-Hispanic14,36289.892.993.388.482.2 84.592.3 8790.4 
 No14,45994.395.694.794.991.2 94.194.4 93.294.5 
 No612346.846.946.248.146.4 45.347.6 43.447.6 
Area poverty             
 5%-10%494232.529.733.8 30.532.9 
 10%-5%268617.716.318.3 17.217.8 
 >15%338122.321.822.5 22.922.1 
 No insurance4963.222.635 3.92.8 4.82.9 
 Medicare577237.140.439.135.531.3 36.137.5 28.938.7 
 Medicaid4102. 2.92.5 4.52.3 
 Medicare/   Medicaid NOS360623.121.223.723.925.2 19.524.9 22.323.3 
 Government15511. 0.81.1 1.30.9 
Tumor stage             
 Regional239339.239.240.941.836.8 44.236.5 53.536.1 
 Distant312451.250.749.849.453.6 4554.7 35.354.7 
Tumor grade             
 Well differentiated76312.312.912.412.611.2.48212.712.3561511.5<.001
 Moderately differentiated247639.841.239.839.238.9 40.739.3 40.739.5 
 Poorly differentiated279544.943.744.545.246.8 43.545.8 4146 
 Undifferentiated1903.12.23.333.1 3.13 3.43 
Tumor size, cm             
 2.5-5362823.12523.123.321.5 28.520.6 28.422 
 >511,05470.368.470.569.772 6373.7 61.672 
Lymph node status             
 Negative6054. 6.44.1 9.24 
 None examined11,10488.787.889.388.489.5 83.991 78.490.7 
 Yes28771819.816.817.916.5 23.915.3 3015.7 
 Yes242015.416.91514.114.3 18.913.7 19.614.6 
 Yes512533.536.432.831.630.7 4030.5 38.532.6 
Volume status             
 LVC10,98268.263.471.371.169.3 2.880.9 
 NTF13,49083.88385.184.583.7 50.499.3 

The majority of patients had tumors that measured >5 cm in greatest dimension (68.6%). Among the reported tumor stage and grade, the most common stage and grade were distant disease (19.4%) and poorly differentiated tumors (17.4%), respectively. The majority of patients did not undergo surgical treatment (81.2%), or receive radiation (82.7%), or receive chemotherapy (63%). Patients most frequently received treatments at LVCs (68.2%) and NTFs (83.8%). Patients who were treated at TFs more frequently presented with regional disease than those who were treated at NTFs (53.5% vs 36.1%; P < .001), whereas NTFs saw a greater percentage of patients with distant disease compared with TFs (54.7% vs 35.3%; P < .001).


The median survival for the entire cohort was 5.7 months (Table 2). Patients aged <40 years had a longer median survival than older patients (8.8 months vs 4.6 months for patients aged >65 years; P < .001). We observed that Caucasian patients had a longer median survival than African-American patients (5.7 months vs 5.1 months; P = .021), and Hispanic patients had an improved median survival compared with non-Hispanic (7.5 months vs 5.5 months; P < .001). The median survival shortened as the community poverty level increased (6.2 months for patients at <5% of the poverty level vs 5 months for patients at >15%; P < .001). Patients who had private insurance had a longer median survival than patients who had Medicare or those without insurance coverage (7 months vs 5.2 months vs 5.2 months, respectively; P < .001).

Table 2. Median Survival
CharacteristicEntire Cohort, n=16,104Subset Analysis
Area PovertyFacility
% of TotalP% of TotalP% of TotalP
No.% of Total<5%5%-10%10%-15%>15%HVCLVCTFNTF
  1. HVC indicates high-volume center; LVC, low-volume center; TF, teaching facility; NTF, nonteaching facility; AA, African American; NOS, not otherwise specified; ND, not determined.

Age, y             
 No5.6 6.35.454.9<.0017.64.8<.0019.25.1<.001
Area poverty             
 5%-10%5.5 7.44.7<.0019.34.9<.001
 10%-5%5.1 74.4<.0018.34.7<.001
 >15%5 6.54.3<.0017.34.5<.001
 No insurance5
 Medicare/Medicaid NOS4.5<.0018.33.9<.001
 Government7.7 9.910.
Tumor stage             
Tumor grade             
 Well differentiated10.2<.001129.310.110.2.76013.88.2<.00113.98.5<.001
 Moderately differentiated7.8 8.686.86.5<.00111.66<.00113.46.5<.001
 Poorly differentiated4.5<.0018.93.7<.001
 Undifferentiated2.9 2.22.734.1.8234.32.2.01752.2<.001
Tumor size, cm             
Lymph nodes             
 Negative11.3 13.910.19.79.6<.001149.9<.00114.210.3<.001
 None examined4.2 4.543.93.7.0045.53.7<.0016.33.8<.001
 Yes14.9 17.61413.512.4<.00118.213.1<.00119.813.6<.001
 Yes11.6 11.611.411.210.4.04012.810.9<.00113.811.1<.001
Volume status             

Patients with local disease (median survival, 10.7 months vs 3.3 months for patients with distant disease; P < .001) and those who had well differentiated tumors (median survival, 10.2 months vs 2.9 months for patients with undifferentiated tumors; P < .001) had improved median survival, as did patients with smaller tumors (11.3 months for tumors <2.5 cm vs 5.1 months for tumors >5 cm; P < .001). Undergoing surgery, receiving radiation, or receiving chemotherapy significantly improved median survival (14.9 months vs 4.4 months, 11.6 months vs 4.6 months, and 9 months vs 3.7 months, respectively; P < .001). Patients who received care at HVCs had better median survival than patients who were treated at LVCs (7.7 months vs 4.8 months; P < .001). Similarly, patients who received care at a TF had an improved median survival compared with patients who received care at an NTF (9.3 months vs 5.1 months; P < .001).

Surgical Outcomes

Among surgical patients, higher mortality was observed in patients from the lowest SES category compared with patients in the highest SES category (5.1% vs 3.7%; P < .001). Significantly improved 30-day mortality rates were observed at TFs (2.8% vs 5.9% at NTFs; P < .001) and at HVCs (2.8% vs 6.7% for LVCs; P < .001). Poor short-term outcomes were observed for the LSES cohort even when they were treated at TFs or HVCs: At TFs, the 30-day mortality rate was 2.2% for patients of LSES versus 1.3% for patients of high SES (P < .001); and, at HVCs, the 30-day mortality rate was 2.6% for patients of LSES versus 1.3% for patients of high SES (P < .001)

Effects of Area Poverty

Patients from less affluent areas were diagnosed at a younger age (69.2 years vs 70.1 years; P < .001), reported more alcohol consumption, and were comprised from a larger proportion of African Americans and Hispanics (Table 1). Patients at the lowest SES level were less likely to undergo surgical treatment (16.5% vs 19.8%; P < .001) and were less likely to receive radiation therapy (14.3% vs 16.9%; P = .003) or chemotherapy (30.7% vs 36.4%; P < .001) compared with patients at the highest SES level. In addition, LSES patients were less likely to be treated at an HVC and had a poorer median survival than high SES patients. A shorter median survival was observed across sex, age, race, and ethnicity as poverty increased (Table 2). In patients who underwent surgery, received radiation, or received chemotherapy, the median survival was shorter for LSES patients. In addition, decreased survival was observed for these patients compared with those in other poverty categories, even at HVCs or TFs (Fig. 1).

Figure 1.

These Kaplan-Meier curves illustrate survival according to patient socioeconomic status category (a) for all patients, (b) for patients who were treated at teaching facilities, and (c) for patients who were treated at high-volume centers.

Effects of Comorbidities

The frequency of patient comorbidities is shown in Table 3. Approximately half of the entire cohort had a diagnosis of hypertension (58.4%). The least frequent comorbidity observed was acquired immunodeficiency syndrome/human immunodeficiency virus (AIDS/HIV) (0.4%). In univariate analysis, cardiac arrhythmias, peripheral vascular disease, paralysis, neurologic disorders, diabetes, renal failure, AIDS/HIV, obesity, and iron deficiency anemia did not have a significant effect on overall survival. The distribution of comorbidities among patient subsets is shown in Table 3. Patients in the LSES cohort had a greater frequency of hypertension, diabetes, AIDS/HIV, and psychoses.

Table 3. Outcomes of Patients Undergoing Surgical Resection
VariableArea PovertyP
  1. HVC indicates high-volume center; TF, teaching facility.

30-Day mortality, %   
 Entire cohort3.77.1<.001
Median survival, mo   
 Entire cohort17.612.4<.001

Multivariate Analysis

The results of a multivariate analysis using the Cox regression method, including patient comorbidities as categorical variables, are summarized in Table 4. Preoperative, independent predictors of worse survival in pancreatic cancer were African-American race (hazard ratio [HR], 1.06; P < .001), non-Hispanic ethnicity, and poverty (HR, 1.09; P < .001). Advanced histologic grade and disease stage, tumor size >2.5cm, and positive lymph node status were among the clinical variables that predicted a worse outcome. All treatment variables—surgical extirpation (HR, 0.52; P < .001), radiation therapy (HR, 0.80; P < .001), and chemotherapy (HR, 0.71; P = .013)—improved outcomes. Receiving therapy at an HVC conferred a survival benefit (HR, 0.88; P < .037). No differences in outcomes were observed for sex, age, alcohol consumption, or tobacco use.

Table 4. Multivariate Regression Analysesa
VariableHR95% CIP
  • HR indicates hazard ratio; CI, confidence interval; NOS, not otherwise specified; LVC, low-volume center; HVC, high-volume center; NTF, nonteaching facility; TF, teaching facility.

  • a

    Patient comorbidities were included as categorical variables.

 MenReference group  
Age, y   
 <40Reference group  
 WhiteReference group  
 African American1.061.057-1.058<.001
 Non-HispanicReference group  
 YesReference group  
 YesReference group  
Area poverty   
 <5%Reference group  
 InsuredReference group  
 No insurance1.190.885-1.594.251
 Medicare/Medicaid NOS0.990.983-1.005.313
Tumor stage   
 LocalReference group  
Tumor grade   
 Well differentiatedReference group  
 Moderately differentiated1.341.212-1.486<.001
 Poorly differentiated1.761.617-1.923<.001
Tumor size, cm   
 <2.5Reference group  
Lymph nodes   
 PositiveReference group  
 NoReference group  
 NoReference group  
 NoReference group  
Volume status   
 LVCReference group  
 NTFReference group  

In a separate analysis that did not include hospital volume in the regression model, facility teaching status was a significant predictor of improved survival (HR, 0.84; P < .001; data not shown). Failure to reach significance when both factors (hospital volume and teaching status) were included in the regression analysis most likely was a result of the collinearity of the 2 variables, because the majority of HVCs also were TFs.


In the last 30 years, despite the recognition of inequalities in the outcomes of cancer care based on race, ethnicity, and SES, these problems remain. In an effort to understand differences in outcome, identify global prognostic factors important for survival, and potentially improve survival for patients with pancreatic cancer, we examined a large population-based cancer registry. To date, this study represents 1 of the largest, most comprehensive analyses, inclusive of patients of all ages, that examines the effects of race, ethnicity, and SES on treatment as well as short-term and long-term outcomes of patients with pancreatic adenocarcinoma. Furthermore, we have included patient comorbidities in our analysis, thus preventing potentially erroneous findings caused by health differences between patient groups. This study also takes into account the effects of clustering, because we included outcomes for patients who were treated at TFs and HVCs.18, 22

Patients of the lowest SES presented with pancreatic adenocarcinoma at younger ages compared with patients of the highest SES. These patients consumed alcohol more frequently and received therapies less often. Regardless of whether or not patients received treatment, median survival was related inversely to SES. Patients in the lowest SES category were treated more frequently at LVCs compared with patients in the highest SES category. Among the lowest SES cohort, those who were treated at HVCs did not experience a significant benefit from regionalized care. LSES patients continued to demonstrate increased perioperative mortality and had a considerably shorter median survival at all treatment facilities regardless of teaching or volume status. Overall, LSES was an independent predictor of worse outcomes among patients who were diagnosed with pancreatic adenocarcinoma.

Our results confirm prior studies demonstrating that LSES is associated independently with poor survival in patients with pancreatic cancer.16, 23 Similar to observations made among African Americans, increased consumption of alcohol and the increased presence of diabetes among the impoverished is associated with an increased risk of developing pancreatic adenocarcinoma.24 However, patients of LSES may be facing different treatment-related factors, such as lack of access to care and refusal or delay of care. Although SES is linked to education, patients of LSES may lack the knowledge to comprehend the complex details of their diagnosis, successfully navigate the healthcare system, and, hence, obtain the optimal multidisciplinary treatment for their pancreatic adenocarcinoma. Furthermore, the social structure and support may be lacking in patients of LSES, and coordinating care among the separate specialists and minimizing loss of income from missed work may prove to be difficult or impossible. Steinberg et al examined the creation and implementation of a lay patient navigator program for the underserved populations in Los Angeles, attempting to equalize access to cancer care.25 The role of patient navigators includes, but is not limited to, identifying patient's barriers to care, coordinating care, providing health education, and finding financial assistance and emotional support systems. Such programs require ample resources and the dedication of an entire community. However, they may be effective in lowering the barriers to care for patients of LSES and should be replicated in other regions of the country.

Patients of LSES may face other barriers that prevent seeking healthcare in a timely fashion, including the financial constraints of traveling and loss of wages. In addition, the impoverished patient may not be sufficiently informed or educated about the improved outcomes at HVCs.26-35 Although the dataset that we examined could not address this point, patients of LSES may not receive the same care at HVCs that privately insured, more educated, and higher SES patients may be receiving. Bilimoria et al observed that as much as 38% of patients who had local disease that was amenable to resection simply were not offered surgery as an option.36 The current study is limited, because we could not identify who was or was not offered treatment or which patients refused treatment. However, our data (from nonoverlapping patient cohorts) support the conclusions that the failure to offer treatment and the refusal of treatment are probable causes of worse outcomes among the impoverished with pancreatic adenocarcinoma.

All cancer cases that were diagnosed or treated in the state of Florida are reported to the FCDS, and Florida represents approximately 6% of the total US population. The quality of the data are enhanced by linkage of the FCDS dataset with the Florida AHCA, which provides additional data modifiers, such as comorbidities, improved treatment information, and improved follow-up. This linked dataset was limited to the 5-year period from 1998 to 2002 and has several limitations. Mortality is reported as all-cause mortality as opposed to disease-free survival, and it is obtained from death certificates reported from the Social Security dataset, which may underestimate patient deaths by up to 5%.37-43 The status of surgical resection margins is not included in the database and could not be included in the current analysis. Furthermore, although data on radiotherapy and chemotherapy were examined, information on whether the patient received palliative versus curative treatment, including the specific regimen and dosage, also were unavailable. In addition, using area poverty as a proxy for SES may result in the misclassification of some patients whose postal code does not accurately reflect the true income level of the individual. This classification scheme, however, has been described elsewhere and is the best available data.38, 44-46 Other variables that are not available in the dataset, such as diet, education, and access to healthcare, also may influence SES.47-53

In conclusion, significant disparities in pancreatic adenocarcinoma outcomes continue to exist between patients of differing SES even after correcting for stage and comorbid conditions. Problems in delivery of healthcare to the indigent appear to contribute to these disparities. The poor appear to be offered surgery and other therapies at a significantly lower rate, and they continue to suffer dramatically worse 30-day mortality at high-volume or low-volume institutions. After resection, the indigent are deriving a dramatically reduced rate of palliation and cure even after surviving the perioperative period. Further efforts to understand the causes leading to these outcome gaps must be made. Our results support the continued care of patients with pancreatic adenocarcinoma at specialty HVCs. Furthermore, we suggest, that until we gain a better understanding of the causes of these disparities, all patients should be referred to HVCs for multidisciplinary evaluation and potential surgical resection. The increased involvement of experienced medical staff who predominantly provide care for private patients with diseases of the pancreas also appears to be indicated to potentially improve survival rates in the underserved.


The authors made no disclosures.