Influence of Socioeconomic and Cultural Factors on Racial Differences in Late-Stage Presentation of Breast Cancer FREE

ALTHOUGH breast cancer incidence is somewhat lower among African American women than among white women in the United States, breast cancer mortality is consistently higher among African Americans.^1- 5 A large part of the explanation for this is that African Americans present with more advanced-stage disease.^6- 12 The reasons for this racial difference in stage, however, are not clear. Socioeconomic factors have been shown to strongly influence the stage of disease presentation^13- 16 and may partially account for the difference. African Americans have less access to medical care and are less likely to undergo breast cancer screening.^17- 18 Most studies, however, show that socioeconomic variables alone do not account for all of the observed effect.^11,19- 23 Even when universal access to medical care is ensured, there are still racial disparities in breast cancer diagnosis and outcome.^24- 25

Cultural factors such as beliefs, attitudes, and knowledge about cancer are also known to vary dramatically by race,^26- 32 and the importance of these cultural factors is increasingly recognized. No study to date, however, has shown a direct relationship between these cultural factors and breast cancer stage or mortality, nor have the interactions among socioeconomic factors, cultural factors, and race been explored.

In 1988 a comprehensive, long-term study to evaluate factors that influence breast cancer diagnosis, treatment, and outcome was started at the Leo W. Jenkins Cancer Center of East Carolina University, Greenville, NC. Emphasis was placed on psychosocial factors as well as medical factors. This report presents study results showing how race, socioeconomic factors, and cultural factors interact to influence breast cancer stage at diagnosis.

The University Medical Center of Eastern Carolina (consisting of East Carolina University School of Medicine, Pitt County Memorial Hospital, and the local private-practice community) is the only inpatient or surgical health care facility in Pitt County or Greene County and is the regional tertiary care referral center for the 29 surrounding counties in eastern North Carolina. Numerous marketing surveys have shown that over 95% of all Pitt County and Greene County residents and a significant percentage of the population of the surrounding counties who require inpatient care or surgery receive this at the facility. The area is predominately rural, and the largest town has a population of fewer than 60000. The population of the entire referral area is about 1.5 million, of whom one third are African American. About 25% of the population lives below the federal poverty level, compared with 14.4% nationally.³³

Starting in 1988 all patients diagnosed as having breast cancer between January 1, 1985, and December 31, 1992, at the University Medical Center of Eastern Carolina were evaluated for interview. Of the 743 white or African American patients diagnosed (1 person each of Asian and Hispanic ethnicity was excluded), 37 (5%) of the total died before interviews could be completed, and another 35 (5%) either moved out of eastern North Carolina prior to being interviewed or were determined to be either physically or mentally incapable of being interviewed. Of the 671 patients still resident in eastern North Carolina and well enough to be interviewed, 53 (8%) refused to participate in the project, and 1 physician at the medical center refused to allow his 78 patients (12%) to participate. Therefore, interviews were completed with 540 of the 671 eligible patients for a response rate of 80%. Interviews were conducted as soon as possible after diagnosis; 58% were completed within 3 months, 70% within 1 year, and 90% within 3 years of the date of diagnosis.

About half of the interviewed patients were residents of Pitt County or Greene County, and the other half were from 20 surrounding counties. Table 1 shows demographics for the patients interviewed, patients treated at the medical center but not interviewed, and patients in the North Carolina State Tumor Registry for the 20 surrounding counties who were not interviewed. The study participants were slightly younger than patients who did not participate, but Table 1 suggests they were otherwise representative of all breast cancer patients in eastern North Carolina.

The study also included a control group of community-dwelling women without breast cancer from eastern North Carolina matched to each breast cancer patient by race, age, and residence type (ie, rural nonfarm, rural farm, small town, large town). These control women were interviewed in their homes with the same instrument administered to the patients.

A multistage, stratified, random procedure developed and used previously by members of the research team³⁴ was used to identify a pool of adult women older than 25 years living in Pitt County or 4 adjacent counties in eastern North Carolina. These counties each demonstrated demographic and residential variability similar to that found in the 20 eastern North Carolina counties from which the patient population was derived. Sampling began with the random selection of 15 census enumeration districts (equivalent to track blocks) from within the 5 counties with a minimum of 2 enumeration districts from each county. With maps to define enumeration district boundaries, a door-to-door canvass identified 1350 households containing eligible adult women. Verbal consent to participate in the study was obtained during the initial contact, and women who had breast cancer were eliminated from the potential respondent pool. About 90% of those contacted initially agreed to participate in the project. Households were then stratified by residential type (households on rural farms, small towns, and so forth) and by race and age categories of the women residents. As each breast cancer patient was entered into the study, a control subject was chosen randomly from all women in the respondent pool who matched the patient by age, race, and residence categories.

A total of 414 matched controls were interviewed over the 5-year study period. There were 135 more patients interviewed than controls. This difference occurred because control group sampling and interviews lagged behind patient interviews. As the end of the funding cycle approached, a preliminary analysis of the data indicated that a sufficient number of controls existed to conduct statistical comparisons with the patient population, and a decision was made to end control group recruitment. The data in Table 1 verify that there are no significant differences in demographic and socioeconomic characteristics between the patient and control groups.

Each patient and matched control was interviewed in person by a sex- and race-matched interviewer. Thirty older adult women residents of the study counties, who had never had breast cancer, were recruited as interviewers. Interviewers were paid to attend 2 intensive day-long training sessions that taught the procedures for obtaining informed consent, the mechanics for administering and recording responses on the structured interview guide, and ways to minimize bias in the interview encounter and to probe neutrally for more feedback. The project team observed 10% of the interviews conducted by each interviewer, and callbacks were made to each of the first 10 respondents visited by an interviewer to make sure she had indeed been in the home and that the interview had proceeded smoothly. The project team edited every interview for completeness and checked for any inconsistencies in response patterns signaling the need to modify the interviewer's techniques.

Interviewers were matched to patients and controls by race, but within racial categories, each interviewer received assignments randomly. The interviewers knew only that the patients had been diagnosed as having breast cancer, and they were given no additional information. Since none of the interviewers received any specific training in breast cancer, it is doubtful that they were aware at all of stage differences or attentive to any clues in physical appearance that might have signaled such differences. They knew only that we wanted to learn more about the reasons why women did or did not seek medical care for breast cancer.

The structured interview was developed based on preliminary in-depth interviews by a cultural anthropologist with a random sample of patients diagnosed previously at the institution. The instrument was subsequently pretested by project staff on both clinic and control populations and revised prior to the beginning of the study. The final structured interview included 173 questions in a variety of closed-ended answer formats on the topics of demographics, health history, support systems, socioeconomic status, and beliefs and attitudes regarding breast cancer, cancer treatment, and health care in general. Many of the questions related health care to cultural factors such as folk beliefs, religious beliefs, and attitudes regarding family and relationships with men.

Questions were read to the respondent by the interviewer since many of the respondents had low literacy levels. The structured interview took a mean time of 50 minutes to complete. At the end of the structured component, each patient was asked 10 open-ended questions about her particular illness experience, and these were analyzed separately from the answers to the structured interview.

Data regarding tumor stage was collected from each patient's medical record. Tumors were staged using the American Joint Committee on Cancer's Manual for Staging of Cancer. When the project first began, the second edition of the manual (1983) was used and tumors that were TNM stage III or IV were considered "advanced stage," whereas all others were considered "early stage." In subsequent editions of the manual, T3, N0 tumors (ie, tumors >5 cm) are now considered stage IIB instead of stage III. For consistency, we defined any tumor larger than 5 cm as "advanced stage," including all stage III and IV cancers and some stage IIB tumors.

Data were analyzed for statistical significance using the logistic regression procedures of SPSS software version 7.5. Tumor stage was the dependent variable dichotomized as either early or late. Various demographic, socioeconomic, or cultural factors were the independent variables. Per capita income was calculated by dividing household income by the number of people supported by that income and categorized as more than $10000, between $5000 and $10000, or less than $5000. Age and education were also split into 3 categories: younger than 50 years, aged 50 to 65 years, or older than 65 years; and less than high school education, high school graduate, or more than high school education. The cultural variables were all dichotomized into true and false or agree and disagree answers.

Univariate analyses were performed to describe the relationship between each of the independent variables and tumor stage. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for each comparison. The prevalence of each of the cultural beliefs among race, income, age, education, and tumor status (ie, either patient or control) subgroups was also analyzed using univariate techniques. In these analyses, holding a particular belief was considered the dependent variable, and each of the demographic characteristics was the independent variable. The effect of race was studied both alone and adjusted for income, age, and education.

Multivariate logistic regression models were constructed to evaluate the influence of race, socioeconomic factors, and cultural factors on tumor stage. The potential interaction of race with the socioeconomic and cultural factors was assessed with a hierarchical backward elimination approach described by Kleinbaum.³⁵ Determination of whether the socioeconomic and cultural factors were potential confounders of race involved examination of the ORs for different models containing the socioeconomic and cultural factors.

Of the 540 patients interviewed, 94 (17.4%) presented with advanced-stage disease. The method of presentation of the breast cancer patients is shown in Table 2. Whereas a significant percentage of early-stage cancers were found on screening mammography or screening clinical breast examination, almost all patients with late-stage cancers presented with breast symptoms or symptoms of distant metastases. In many cases the tumor replaced a large portion of the breast. About 25% of the late-stage patients had frank ulceration, which led to bleeding or tissue infection.

The association of a variety of demographic and socioeconomic factors with advanced disease stage is shown in Table 3. Being African American, having a per capita income of $10000 or less, and lacking private health insurance all had a major impact on disease stage. In addition, as might be expected, patients who reported that they had put off seeing a physician because of money or transportation problems were also significantly more likely to present with late-stage disease. Never having been married was also significantly associated with stage, perhaps because these women lacked both the financial and social support provided by a spouse. Age did not have a significant association with stage at presentation and, surprisingly, the relationship between education and stage was only marginally significant.

Table 4 shows a variety of cultural and psychosocial factors that all had a significant influence on tumor stage when tested independently. These cultural factors seem to fall into the following categories: health care utilization, folk beliefs, fundamentalist religious beliefs, relationships with men, perceived risk or fatalism, belief in various treatments for breast cancer, and breast cancer knowledge. Not having a regular physician or seeing the physician once a year or less was highly associated with advanced disease presentation. Women who held any of the culturally derived folk beliefs, fundamentalist religious beliefs, or beliefs about relationships with men or fatalism were all significantly more likely to present with late-stage disease. In contrast, women who scored correctly on 2 breast cancer knowledge items were significantly less likely to present with late-stage disease. Belief that herbs, medicines, or chiropractic are effective treatments for breast cancer also correlated positively with late-stage disease, whereas belief in the effectiveness of surgery was associated with early presentation.

There were strong, consistent patterns in the prevalence of the psychosocial and cultural beliefs among demographic subpopulations. Most of the folk beliefs, fundamentalist religious beliefs, and beliefs about fatalism or relationships with men were significantly more common among African Americans and among those with lower income or less education. In contrast, knowledge about breast cancer was significantly more common in whites, upper-income persons, and better educated women. Although the psychosocial and cultural variables were associated with income, age, and education, the majority of them correlated most closely with race. Table 5 shows that most of the variables were significantly associated with race even when adjusted for age, income, and education. Interestingly, however, health care utilization, as measured by having a regular physician and number of yearly visits, did not seem to differ by race.

Table 5 also shows the prevalence of these cultural beliefs in women from the control population compared with breast cancer patients. Many of the beliefs showed very similar distributions in both groups. Not surprisingly, breast cancer patients were less likely than controls to believe that "people who take good care of themselves usually don't get cancer" (ie, OR for patients compared with controls, <1), and more likely than controls to believe that "health professionals control their health" (ie, OR for patients compared with controls, >1). In addition, the patients had much stronger belief in surgery as an effective treatment for breast cancer. Most importantly, however, several of the folk beliefs and fundamentalist religious beliefs that were associated with late-stage cancer were actually more prevalent in the control group than in the patient group. Clearly these beliefs did not result from experiences related to having breast cancer or interacting with the health care team but rather were preexistent in the general population.

To assess the confounding of socioeconomic and cultural factors on race and to evaluate the relative importance of these variables in predicting breast cancer stage, various multivariate models were constructed. Before analyzing the models, the assessment of race as a possible effect modifier was performed by analyzing the interaction of race with each socioeconomic and cultural factor using a backward elimination procedure.³⁵ There were no statistically significant interactions. Next, the effect of race on tumor stage was assessed both alone and when the socioeconomic or cultural variables were added as a group to the model as potential confounders. Table 6 shows that the OR for advanced disease among African Americans compared with whites decreased from 3.0 to 1.8 when either the socioeconomic variables or the cultural variables were added to the model. When both the socioeconomic and cultural variables were added to the model, the OR diminished further to 1.2. Thus, it appears that both the socioeconomic and cultural variables are strong confounders of race and together can largely account for the effect of race on late-stage disease presentation.

The most important reason for the lower survival rate from breast cancer among African American women compared with white American women^1- 5 is that African American women present with more advanced-stage disease.^6- 12 The reasons for this disparity in stage, however, are not completely understood. While several investigators have emphasized the relative importance of socioeconomic status in influencing access to physician care and screening services,^13- 16 others have found that socioeconomic effects alone cannot account for all of the difference,^11,19- 23 and yet others have emphasized the possible role of biological differences in tumor characteristics.^7,36- 37 Race is used by many investigators to refer mainly to overt phenotypic differences, usually of skin color, and is rarely, if ever, operationalized in terms of specific gene frequencies. The American Anthropological Association argues in its Statement on Race³⁸ that modern humans (Homo sapiens) are a fairly recent and homogeneous species, and genetic data indicate that there is as much genetic variability between 2 people from the same "racial" group as there is between 2 people from any 2 different "racial" groups. Thus, there are few biological differences between these large population groups that are of explanatory relevance.³⁹

To the extent that the individuals labeled as belonging to a particular race or ethnic group share certain beliefs, attitudes, and behaviors (culture), they may come to engage in similar health practices and may develop similar illness profiles. Recently, many studies have shown that there is a large difference among races in breast cancer knowledge, beliefs, and attitudes.^{26- 32,40- 42} This has been shown to influence cancer screening and prevention behaviors,^{28,30,43- 47} but until now has not been shown to directly influence stage at diagnosis. Our results demonstrate clearly that psychosocial and cultural variables directly influence stage and, in combination with socioeconomic variables, are sufficient to explain the difference in stage between African Americans and whites.

An important question is how the cultural beliefs that we have identified led to diagnosis of breast cancer at a more advanced stage. In our study, 30% of the cancers in whites and 11% in African Americans were discovered by routine screening mammography. Obviously cancers discovered by screening at an early stage do not become advanced, and therefore, part of the explanation may be that the cultural beliefs were associated with differential use of screening mammography. However, O'Malley et al,⁴⁸ in a study involving women in the same community at the same time, found that women's knowledge and beliefs had very little influence on use of screening mammography, and the most important factor was whether it was recommended by a physician. Furthermore, in our study the majority of early-stage cancers as well as late-stage cancers were found by the patient. Therefore, we believe that the most important effect of the cultural beliefs is that they lead to delayed presentation once a woman has developed a palpable breast abnormality. This would also be consistent with the finding of Jones et al⁴⁹ that mammography alone does not explain all of the racial differences in stage of diagnosis.

Our previous analysis⁵⁰ of the open-ended portion of the interviews with African American women with late-stage disease allows some understanding of the reasoning process involved and demonstrates that cultural conceptions of breast lumps and cancer directly influence women's recognition and evaluation of breast symptoms. These women interpreted their symptoms with reference to a traditional or "folk" model of the blood. In terms of this model, breast lumps are not necessarily bad. They are a normal part of a woman's system as evinced by the fact that if left alone, they tend to come and go. Consequently, patients reported that "lumps that aren't bothering you are best left alone." However, lumps often change, and once a lump begins to "take on a life of its own," it may become bothersome. This transition was reported by patients to happen for several reasons—either a blow or bruise to the breast triggers the growth, or an ever-increasing number of impurities in the blood begins to "move up" or "grow out," or a lump stays in 1 place too long and begins to "take root."

Recognizing that a lump is growing and becoming bothersome, however, does not necessarily imply that a woman will seek medical treatment for it. The majority of patients with late-stage presentation believed that "letting air get to a cancer" or "cutting on a cancer" was to be avoided because it would cause the cancer to spread. Instead, many advocated the use of alternative treatments including herbal remedies, over-the-counter medications, chiropractic regimens, and, perhaps most importantly, prayer and a reliance on God to heal the disorder. Finally, a significant number of the African American women presenting with late-stage disease were reluctant to seek medical treatment for cancer because of their fears about how the disease would impact on their relationships with the men in their lives. These women were afraid that their husbands or male partners would leave them if a cancer diagnosis became known, both because they would no longer be physically attractive but also because they might become burdens to men financially and emotionally.

A very important question is whether the results presented in this article can be generalized to a larger population. It seems likely that the attitudes and beliefs identified are fairly widespread throughout the rural southeastern United States. Whether they will be found to apply to inner-city African Americans in the northern or western areas of the United States, however, will require further study. If the cultural beliefs are found to be regionally specific, at least the principle that cultural beliefs and socioeconomic factors act together to explain the racial effect on tumor stage at diagnosis may be broadly applicable.

The findings reported here have important implications for physicians. Physicians must be aware of the cultural and psychosocial biases of their patients and address them in culturally sensitive ways. If physicians understand their patients' fears and misconceptions about breast cancer, they will be more effective in promoting behavior that may allow early detection and treatment. For example, physicians can explain that air and surgery do not make breast cancer spread, but instead early removal offers the best chance for cure. To counteract the fatalistic notion that the breast cancer is part of God's plan, it is helpful to promote the concept that God works through physicians to cure breast cancer. Finally, physicians must be aware of a patient's concern about the influence a diagnosis may have on her husband or partner and deal with this in a sensitive manner.

These findings also have positive implications for cancer education efforts and public health interventions. If the racial difference in breast cancer stage at diagnosis were purely socioeconomic, this would be very discouraging as economic issues are difficult to change. However, by recognizing that at least part of the problem is cultural, it may be possible to modify key beliefs in a way that would lead to earlier breast cancer diagnosis and still be consistent with the underlying cultural attitudes of the target population. Promotion of increased breast examination and mammography in the absence of research on the compatibility of these practices with predominant cultural beliefs will fail to have the desired impact. This may explain why recent community interventions in Detroit, Mich,⁵¹ and North Carolina¹⁷ increased breast cancer screening significantly more among white women than among African American women. However, interventions that address the specific cultural beliefs and psychosocial attitudes identified in this research may be effective at decreasing the excess stage III and IV cancer among African Americans. Although these cancers account for only 15% to 30% of the total, they have a high mortality and contribute significantly to the difference in survival rates by race.