African-American (AA) patients with colorectal carcinoma have a worse prognosis compared with Caucasians. To analyze the causes of this disparity in survival, a retrospective study of patients with colorectal carcinoma was undertaken. The impact of treatments received and the role of socioeconomic factors such as income, education, and poverty levels were studied.
A retrospective analysis of patients with colorectal carcinoma at a single institution was conducted. The overall survival of AA and Caucasians, stage at presentation, treatment received, and socioeconomic factors were analyzed using the institutional tumor registry and 1990 census data.
The overall survival of AA patients was worse compared with Caucasians, both due to all causes (P < 0.001) and cancer-related deaths (P < 0.001). The relative risk of death due to all causes was 1.4 (95% confidence interval [CI] 1.2–1.8) for AA, 4.3 for patients with Stage IV disease (95% CI 3.2–5.7), and 2.3 for patients not undergoing surgery (95% CI 1.7–3.1). After multivariate adjustment for gender, site, socioeconomic factors, and therapeutic modalities, the relative risks for death were 1.5 (95% CI 1.2) for AA, 1.4 (95% CI 1.1–1.7) for patients 60 years of age or older, and 4.2 (95% CI 3.4–5.2) for Stage IV disease. The survival difference between AA and Caucasians was not influenced by income, poverty level, and education. African Americans were treated less frequently with chemotherapy and radiation therapy compared with their Caucasian counterparts.
African-American (AA) patients with malignant disorders have a worse prognosis compared with Caucasians.1–4 This may be due to socioeconomic factors, which influence access to health care, histologic variations, nutritional and health status, differences in the aggressiveness of the therapeutic approach, compliance with treatment, and coping strategies.2 Biologic factors also may play a role in the differential outcome of patients with various malignancies.5, 6 The importance of socioeconomic factors is emphasized by some and disputed by others.7–15 Equal access to health care facilities has eliminated the survival difference among veterans with colorectal carcinoma, suggesting that access to health care plays a role in the survival of these patients.16, 17 Patients with lower socioeconomic status (SES) present with more advanced-stage colorectal carcinoma.18 To better understand the factors influencing the survival difference between AA and Caucasians with colorectal carcinoma, we conducted a retrospective analysis of patients with colorectal carcinoma treated at a single institution. We evaluated disease stage, treatment received, and socioeconomic factors.
MATERIALS AND METHODS
A retrospective analysis of 617 patients with a diagnosis of colorectal carcinoma treated between 1984 and 1997 at a state-funded university hospital located in Arkansas was performed. Patients at this institution are treated regardless of their ability to pay. The patient population consists mainly of Caucasians and AA, with few patients belonging to other ethnic groups. Caucasians account for approximately 70–80% of patients and the majority of patients live in rural communities. During the study period, there was only one AA radiation oncologist at this institution. Patients aged 18 years and older with colorectal carcinoma were identified through the institutional tumor registry. The data in the tumor registry are constantly updated regarding the treatment received and patients are followed untill the time of their death. Patients were stratified according to their race and gender. The stages at diagnosis and treatment characteristics (surgery, chemotherapy, and radiation therapy) were analyzed. All the data were obtained from the computerized database available at the tumor registry using appropriate tumor registry codes. The age, gender, stage, treatment received, and survival data were obtained from the database using appropriate codes for each of these parameters. The cause of death is documented as cancer related or noncancer related, based on information on the death certificate.
The SES data were derived from the 1990 Census of Population and Housing for the state of Arkansas.18 Tape file 3B represented the patient population. The zip code of the patient's residence was used to assess the median education level, income, and poverty level for the place of residence. These data provide summary education and economic data for each zip code in the state. The following data were used to analyze SES: areas with low high school graduation rates, areas with low income, and areas with high poverty rates. The median high school graduation rate for Arkansas was 65.4%. Zip codes with high school graduation rates less than 65.4% were defined as areas with low high school graduation rates. Similarly, zip codes with median incomes less than $20,018, the median income for Arkansas in 1990, were deemed to be low-income areas.
Using Arkansas zip codes, the median of the percentage of households below the poverty threshold was 19%. Zip codes with rates above this level were defined to be high poverty areas. To relate this census information to the patients, we used the zip code of the their residence at the time they were diagnosed with their disease.
Patient demographic and treatment characteristics were summarized using descriptive statistics, including percentages, medians, and ranges. Associations between race and the remaining factors were evaluated using Pearson's chi-square tests for categoric data and Wilcoxon rank-sum tests for continuous data. Survival distributions were estimated using the method of Kaplan and Meier and were compared using log rank tests. Cox proportional hazards models were used to model the collective effect of demographic and treatment factors on survival. These results were summarized using risk ratios and 95% confidence intervals (CI). For all analyses, P values less than or equal to 0.05 were statistically significant. All analyses were performed using SAS/STAT software, Version 7 (SAS Institute, Cary, NC).
Table 1 lists the demographic characteristics of the 617 patients with colorectal carcinoma. Age and gender were comparable between Caucasian and AA patients (P = 0.41 and 0.08, respectively). Compared with Caucasians, AA patients had a lower high school graduation rate, as well as lower income and higher poverty rates (Table 1). The stages at presentation were comparable for both Caucasian and AA patients (P = 0.324), although 26 Caucasian and 14 AA patients did not have their stages recorded in our database. These 40 patients were excluded from the multivariate analysis. African-American patients were treated less frequently with chemotherapy and radiation therapy compared with Caucasian patients, although there was no difference in the surgical treatment received by both groups (Table 2).
Table 1. Distribution of Race, Gender, Age, and Stage of Disease
Table 2. Treatment Received by Patients Stratified According to Site of Disease
AA: African American.
Stages II and III
African-American patients with colorectal carcinoma had poorer survival rates due to all causes compared with Caucasians (P < 0.001; Fig. 1a). The median follow-up periods of Caucasian and AA patients were 56 months and 49 months, respectively. The difference persisted when cancer-related deaths were analyzed, excluding the causes of death due to other causes (Fig. 1b). Univariate analysis of demographic characteristics and treatment factors revealed that race, stage, and surgery played a significant role in the outcome of these patients (Fig. 2a). The relative risk of death for AA patients was 1.4 (95 % CI 1.2–1.8) compared with Caucasians, 4.3 (95% CI 3.3–5.7) for patients with Stage IV disease compared with patients with Stage I disease, 1.3 (95% CI 1.03–1.5) for patients aged 60 years of age or older compared with patients younger than 60 years of age, and 2.3 (95% CI 1.7–3.1) for patients who did not undergo surgery compared with patients who did. The relative risk of death for males compared with females, site of the disease (colon vs. rectum), and socioeconomic factors was not significant. The relative risk of death was 1.08 (95% CI 0.8–1.3) for patients living in zip codes with a low high school graduation rate, 1.05 (95% CI 0.8–1.3) for patients living in low income areas, and 1.1 (95% CI 0.9–1.3) for patients living in high poverty areas. In a multivariate analysis, race, age, and stage of disease had a significant effect on the outcome of these patients (Fig. 2b). African-American patients had a worse outcome compared with Caucasian patients, the risk of death being 1.5 (95% CI 1.2–1.9). Patients 60 years of age or older had a worse prognosis, their risk of death being 1.4 (95% CI 1.1–1.70). Patients with Stage IV disease fared worse than patients with other stages, their relative risk of death being 4.2 (95% CI 3.4–5.2). The difference in survival between AA and Caucasians with colorectal carcinoma persisted over and above the differences that might be expected between the groups using socioeconomic factors and education level (Fig. 2a–c).
In our retrospective analysis, data were collected from a computerized tumor registry that is updated continually. In addition, patients are followed until their death and the cause of death is ascertained and entered into the database. Caucasian and AA patients comprised our study population. Patients belonging to other minorities were not seen. We observed that AA patients had a worse overall survival and were treated less frequently with radiation therapy and chemotherapy lcompared with Caucasians.
Socioeconomic factors, which are linked to access to health care, were blamed in the past for the disparity in survival rates between Caucasian and AA patients. Cancer patients treated at public hospitals do not fare as well as patients treated at private hospitals.7 Patients with lower SES who are treated at the same facility as patients with a better SES have worse survival rates.8–13 In addition, lower SES patients have more advanced-stage disease at presentation compared with those from higher SES.19 African-American patients with early-stage lung carcinoma fare worse than Caucasian patients due to lesser operative intervention.20 Because access to health care is not influenced by socioeconomic factors in the veterans health care system, patients with colorectal carcinoma presenting to such institutions have been evaluated. There is no difference in the outcome of patients in these institutions.16, 17 In these studies, both AA and Caucasian veterans received similar treatments such as chemotherapy, radiation therapy, and surgery, supporting the hypothesis that equal access to health care was an important factor for receiving similar therapy. Contrary to this observation, AA patients in our population were treated less frequently with radiation and chemotherapy compared with Caucasians, which may be one of the main reasons for their poorer survival rates.
An important question that needs to be answered is: Do socioeconomic factors such as education level, income, and poverty level play a role in the outcome of patients with colorectal carcinoma? Previous studies suggest that nonveteran AA patients with malignant disorders treated in community hospitals have a poorer outcome compared with Caucasians.8, 12–14, 21 A study of Medicare recipients aged 65 years and older with colorectal carcinoma showed that Caucasians had superior survival compared with AA.22 Although Medicare recipients have similar medical benefits, the analysis was incomplete due to the lack of access to medical records and the absence of treatment details. An analysis of socioeconomic factors influencing the outcome in these studies was not performed. This was partly due to the lack of a standardized method for the collection of socioeconomic data for a given patient population.
Krieger23 described a census-based methodology for the analysis of socioeconomic factors. The SES of our study patients was derived using the 1990 Census of Population and Housing for the state of Arkansas.18 These data provide summary education and economic data for each zip code in the state. For the purpose of our analyses, we used this database to define three summary measures of economic and education status: areas with low high school graduation rates, areas with low income, and areas with high poverty rates. Based on these data, the median high school graduation rate for Arkansas was 65.4%. Zip codes with graduation rates less than 65.4% were defined as areas of low high school graduation. Similarly, zip codes with median incomes less than $20,018, the median income for Arkansas in 1990, were deemed to be low-income areas.
The Census of Population and Housing uses household income, family size, and number of dependent children to define poverty income thresholds and reports the percentage of households below this threshold within each zip code. Using Arkansas zip codes, the median of the percentage of households below the poverty threshold was 19%. Zip codes with rates above this level were defined as high poverty areas. To relate this census information to the patients, we used the zip code of their residence at the time they were diagnosed with their disease.
We collected education, poverty, and income data using the zip code of the area of residence. Using similar data for patients with lung carcinoma, another study found that AA patients lived in geographic areas with lower median income compared with Caucasians.20 However, we did not find a correlation among socioeconomic factors, treatment received, and survival (Fig. 2). We analyzed only patients who sought medical attention as opposed to the other studies where the access to medical care is compromised for these subsets.
In general, patients who do not have access to health care have more advanced stage of disease at presentation and, in turn, a worse prognosis. Socioeconomic status has an impact on access to health care and on the stage of disease at presentation and prognosis.19 We found no difference in the stages of presentation between AA and Caucasian patients.
The outcome of patients with various malignancies is related to the treatment received. This has been shown in patients with lung carcinoma. African-American and Caucasian veterans receiving equal care have similar survival rates.17 We observed that AA patients were treated less frequently than Caucasian patients with chemotherapy and radiation therapy, resulting in their worse overall survival rates. This disparity may be related to poor SES, poor compliance, or lack of motivation.
Our results should be interpreted with caution because our study was performed retrospectively at a single institution. Prospective studies are needed to address this issue. However, prospective studies may be difficult to conduct due to the smaller number of AA patients enrolled in clinical trials.24
In an analysis of five prospective clinical trials conducted by the National Surgical Adjuvant Breast and Bowel Project, there was an overall survival benefit for Caucasian compared with AA patients.25 However, the disease-free survival was the same for both races, which raises the question: Do associated morbidities and other factors contribute to their death? We found that there was a statistically significant difference in survival between Caucasian and AA patients even after excluding the noncancer-related deaths.
The census-based evaluation of SES has its own limitations. This method of imputing data is fraught with many difficulties, which should be considered when trying to interpret these analyses. First, by assigning the same value to all patients living within a certain geographic region, the natural variation within that region is ignored. This artificially deflates estimates of standard errors, thus producing unreliable test statistics. Second, it is difficult to determine the effect of bias introduced by this method. The census information represents the socioeconomic “snapshot” of Arkansas between 1989 and 1990. Our data were collected between 1984 and 1997, several years before and after the census timeframe. We are forced to make the strong assumption that the 1990 census data are representative of the whole time range during which the data were collected. Finally, the assignment of zip codes is not a static process. Over the years, existing zip codes are split to form new ones. Therefore, the census does not provide information for patients diagnosed after 1990 and residing in a “new” zip code area.
In conclusion, AA patients with colorectal carcinoma have a poorer survival compared with Caucasians. We found that this was due to cancer-related deaths. The poorer survival of AA patients with colorectal carcinoma is probably due to decreased utilization of available treatment modalities, such as chemotherapy and radiation therapy. The disparity cannot be attributed entirely to socioeconomic factors and lack of access to health care. The causes of decreased utilization of available health sources may be due to the cultural differences, differences in beliefs, and lack of motivation of the medical community to address their special needs. These factors should be identified and dealt with to improve the health care of AA patients with various malignant disorders.