276
Influence of Family History and Preventive Health Behaviors on Colorectal Cancer Screening in African Americans Kathleen A. Griffith, PhD, CRNP1 Deborah B. McGuire, PhD, RN2 Renee Royak-Schaler, PhD, Med3 Keith O. Plowden, PhD, RN4 Eileen K. Steinberger, MD, MS5
BACKGROUND. African Americans (AAs) have low rates of colorectal cancer (CRC) screening. To the authors’ knowledge, factors that influence their participation, especially individuals with a family history of CRC (‘‘family history’’), are not well understood. METHODS. A secondary analysis of the 2002 Maryland Cancer Survey data examined predictors of risk-appropriate, timely CRC screening (‘‘screening’’) in
The Johns Hopkins University School of Nursing, Baltimore, Maryland.
AAs with a family history and in individuals without a family history. Predic-
2
prostate-specific antigen (PSA) screening, body mass index, activity, fruit/vege-
1
Department of Organizational Systems and Adult Health, University of Maryland at Baltimore, School of Nursing, Baltimore, Maryland.
tors that were evaluated included age, sex, family history, mammogram or table consumption, alcohol, smoking, perceived risk of cancer, education, employment, insurance, access to a healthcare provider, and healthcare pro-
3
vider recommendation of fecal occult blood test (FOBT) and/or sigmoidoscopy/colonoscopy.
4
tion for FOBT (odds ratio [OR] of 11.90; 95% confidence interval [95% CI], 6.84–
Department of Epidemiology and Preventive Medicine, University of Maryland at Baltimore, School of Nursing, Baltimore, Maryland. Department of Family and Community Health, University of Maryland at Baltimore, School of Nursing, Baltimore, Maryland. 5
Department of Epidemiology and Preventive Medicine, University of Maryland, Baltimore, School of Medicine, Baltimore, Maryland.
RESULTS. In individuals without a family history of CRC (N 5 492), recommenda20.71) and sigmoidoscopy/colonscopy (OR of 7.06; 95% CI, 4.11–12.14), moderate/vigorous activity (OR of 1.74; 95% CI, 1.06–2.28), and PSA screening history (OR of 2.68; 95% CI, 1.01–7.81) were found to be predictive of screening. In individuals with a family history (N 5 88), recommendation for sigmoidoscopy/ colonscopy (OR of 24.3; 95%, CI 5.30–111.34) and vigorous activity (OR of 5.21; 95% CI, 1.09–24.88) were found to be predictive of screening. However, family history did not predict screening when the analysis was controlled for age, education, and insurance. AAs who had a family history were less likely to screen compared with their white counterparts (N 5 293) and compared with AAs who
The Maryland Cancer Survey was supported by the Maryland Cigarette Restitution Fund. We thank Karen Soeken, PhD, for her methodological and statistical expertise; Louise Jenkins, PhD, RN, for her conceptual and methodological guidance; and Carmella S. Groves, MS, and Dianne M. Dwyer, MD, at the Center for Cancer Surveillance and Control, Maryland Department of Health and Mental Hygiene, for their support with the 2002 Maryland Cancer Survey data. Address for reprints: Kathleen A. Griffith, PhD, CRNP, the Johns Hopkins University School of Nursing, 525 North Wolfe Street, Suite 306, Baltimore, MD 21205; Fax: (410) 614-1446; E-mail:
[email protected] Received November 27, 2007; revision received February 22, 2008; accepted March 14, 2008.
ª 2008 American Cancer Society
were at average risk for CRC (P < .05).
CONCLUSIONS. Regardless of family history, healthcare provider recommendation and activity level were important predictors of screening. Lower screening rates were observed in AAs who had a family history compared with individuals who did not. The authors believe that, for AAs who have a family history, further examination of barriers and facilitators to CRC screening within the cultural context is warranted. Cancer 2008;113:276–85. 2008 American Cancer Society.
KEYWORDS: mass screening, African Americans, colorectal neoplasms, disparity.
C
olorectal cancer (CRC) is the third most commonly detected cancer and the second leading cause of cancer-related death in the U.S.1 Despite the approximately 5% annual reduction in CRCrelated mortality between 1993 and 2001, which mainly was attributed to improvements in treatment,2 African Americans have a 20% higher incidence of CRC and a 40% higher disease-related mortality
DOI 10.1002/cncr.23550 Published online 9 June 2008 in Wiley InterScience (www.interscience.wiley.com).
CRC Screening in African Americans/Griffith et al.
compared with whites.3 This difference is attributable at least in part to differences in screening practices.3–7 Adherence to national screening guidelines for CRC could result in as much as an 80% reduction in mortality of the disease8 and, with approximately 153,760 new cases diagnosed in the U.S. in 2007,9 the potential impact of risk-appropriate screening is large. CRC screening rates have increased slightly during the last decade,2 but they remain low in all racial groups,10,11 and especially in African Americans.5,12 The literature suggests that several factors influence CRC screening. A family history of CRC often is associated with increased rates of screening in whites compared with the rates among those who do not have a family history of the disease.13–16 The impact of a family history of CRC on screening practices has been reported in a small sample of African-American individuals ages 40 years to 49 years (n 5 31), revealing that their completion of endoscopy (9.3%) was significantly lower than that in their white counterparts (27.9%; P < .03).2 A racial difference was not noted between those same groups for fecal occult blood testing (FOBT). That study suggested that a racial disparity may affect screening practices not only in those at average risk of the disease but also in individuals at increased risk by virtue of a positive family history of CRC. Provider recommendation of screening tests is associated strongly with CRC screening adherence both in African Americans and whites.17–20 Additional factors that reportedly are influential in CRC screening among African Americans include being a woman,21 being older,12,22 having an increased perceived risk of cancer,17,23 having been screened for other cancers,18,24 having a high school degree or higher education,18 exercising,25 and having health insurance.12,26 The American Cancer Society recommends that individuals at average risk begin screening for CRC at age 50 years and at age 40 years if they are at increased risk for the disease because of a family history.27 Few studies have examined completion rates of risk-appropriate CRC screening in a manner consistent with these recommendations.18 By not separating individuals who have received riskappropriate CRC screening within the recommended timeframe from those who have not, conclusions regarding adherence to screening may be of limited value. The current study was undertaken to address the gap in the literature related to riskappropriate, timely screening completion in African Americans and the role of a family history of CRC and other predictors of CRC screening suggested by the literature.
277
MATERIALS AND METHODS Data Collection and Measures Data from the 2002 Maryland Cancer Survey (MCS),28 a cross-sectional survey of 5040 Maryland residents, were abstracted into a data file for this analysis using SPSS-PC version 12.0. The MCS questionnaire was developed with validated questions from national and state surveys, such as the Behavioral Risk Factor and Surveillance Survey, the National Health Interview Survey, the National Health and Nutrition Examination Survey, and the Maryland Department of Health and Mental Hygiene Oral Health Survey. In addition, several new questions were developed for the 2002 MCS. The 2002 MCS was conducted as a population-based, randomdigit dial, computer-assisted telephone interview (CATI) survey using list-assisted, disproportionate stratified sampling. Respondents for the MCS were limited to individuals ages 40 years who resided in a private household in the state of Maryland. Individuals who were not eligible for the survey included those who did not speak English or were unable to communicate because of a physical or mental impairment and those living in group homes or institutions. Maryland was divided into 2 geographic strata: urban (consisting of Baltimore City and the 7 counties in the Metropolitan Baltimore-Washington, DC area) and rural (consisting of the remaining counties in western and southern Maryland, and the eastern shore of Maryland). A survey and research firm (REDA International, Inc.) conducted the interviews for the survey using CATI technology. To reach a final disposition, REDA staff made up to 9 calling attempts. If it was determined that there was only 1 individual aged 40 years living in the household, then that adult was interviewed. If 2 individuals aged 40 years lived at that residence, then 1 was chosen at random for the interview. By using an anonymous questionnaire, interviewers asked questions about demographics, cancer screening behaviors, health risk factors, and access to healthcare.28 Each interview took 20 minutes on average. The secondary analysis study reported here was approved by the Institutional Review Boards at the University of Maryland.
Statistical Analysis African-American representation within the total sample of the MCS was 981 (20%). The inclusion criteria for this study were 1) age 50 years for individuals without a family history of CRC, or 2) age 40 years for individuals with a family history of CRC, and 3) complete data on survey items that were
278
CANCER
July 15, 2008 / Volume 113 / Number 2
required to compute the dependent variable, which was risk-appropriate, timely completion of CRC screening. These criteria were applied to both whites and African Americans. After the application of inclusion criteria, 40l individuals were ineligible for the current analysis, because they were too young (N 5 355), or they had missing data on variables that were required to compute the dependent variable (N 5 43), or they did not complete the item on family history of CRC (N 5 3). Of the 580 individuals who were eligible, 88 reported a family history (1 first degree relative with CRC) and were designated as being at increased risk for the disease. Variables from the MCS dataset were selected based on published reports of cancer screening predictors in the general population and specifically in African Americans. These variables included age, sex, body mass index, risk perception, fruit/vegetable intake, activity level, alcohol intake, smoking, insurance status, and self-reported history of screening for either breast cancer or prostate cancer, education, employment, healthcare access, and provider recommendation for screening (either FOBT within the last year or ever having a recommendation for endoscopy). Items from the MCS that had to be completed to compute the dependent variable, risk-appropriate, timely CRC screening, included name(s) of screening test(s), timeframe for completion of test(s), and reason(s) for undergoing test(s). Allowable reasons for undergoing these tests included part of a routine physical examination/screening and because of a family history of CRC. For individuals without a family history of cancer, risk-appropriate, timely CRC screening was defined as completion of either an FOBT within 1 year, sigmoidoscopy within the last 5 years, or colonoscopy within the last 10 years. In individuals with a family history of CRC, risk-appropriate, timely CRC screening was defined as having completed a colonoscopy within the last 10 years. These definitions were established in accordance with national recommendations that were updated in 2002 for individuals at average risk and increased risk for the disease.29 Fecal immunochemical tests were not included in the survey response options, because they were not part of average-risk screening recommendations until 2003.30 Identical inclusion criteria were applied to whites with a family history of CRC. Descriptive statistics were compiled to profile the African-American individuals; then, comparisons were made between individuals with and without a family history of CRC for each variable. Completion of risk-appropriate screening also was compared between African Americans with and without a family history of CRC and between African Americans
and whites in the MCS who had positive a family history of CRC. Qualification of individual variables for multivariate models to examine predictors of risk-appropriate, timely screening was determined using P < .20 as a cutoff value in bivariate logistic regression analysis. Three multivariate models—the first consisting of individuals with a family history of CRC, a second consisting of individuals without a family history of CRC, and a third consisting of the entire sample—were constructed to evaluate the independent influence of predictors on risk-appropriate, timely CRC screening. After specification of the main model for the group without a family history of CRC, interaction terms were added to this model based on correlations between the variables (F > .30) and the plausibility of each correlation. The intent of this procedure was to identify any moderating effect of one variable on the other. Then, a model containing interaction terms was compared with the main model. Interaction terms were not constructed for the group with a family history because the sample size was small, and the power was not sufficient to allow for additional variables beyond those that qualified in the bivariate analysis. The third model was constructed primarily to evaluate the role of a family history of CRC in the completion of riskappropriate, timely CRC screening and to identify any interactions between the family history variable and selected demographics. The demographic control variables that were included in this model—age, education, and health insurance status—were chosen because of their common use in multivariate modeling. Mediation, that is, the extent to which a variable accounts for the relation between the predictor and the outcome of risk-appropriate, timely screening, was considered in all multivariate models using the 3 standard criteria: 1) significant relation between the suspected mediator variable and the outcome, 2) significant relation between the suspected mediator and predictor variable, and 3) loss or reduction in significant relation between the predictor variable and the outcome variable when the suspected mediator was added to the model.31 The significance value for multivariate models was set at P < .05.
RESULTS Characteristics of the Study Sample In the sample that was eligible for the analysis of African Americans (N 5 580), individuals predominantly were women (65%) ages 50 years to 65 years. Nearly half had some college education or higher (48%) and were either employed (46.7%) or retired
CRC Screening in African Americans/Griffith et al.
(41.3%). The majority had health insurance (88.8%; data not shown). When individuals with and without a family history were compared, there was a significant difference between the groups (P < .05) on several variables, including age and cancer risk perception (see Table 1). The minimum age allowed for inclusion in the study differed between groups— age 40 years for those in the family history group and age 50 years for the group without a family history of CRC—and most likely accounted for the significant difference on this variable. African Americans and whites with a family history were compared on demographic variables, including age, sex, education, employment, and health insurance status. Significant differences included health insurance (94.2% of whites vs 87.5% of African Americans reported having complete health insurance during the previous 12 months; P 5 .044) and age (53% of whites vs 37.5% of African Americans were aged >65; P 5 .001). No other differences on demographics were noted (data not shown). Increased health insurance coverage may have been related to older age among whites. Reported CRC screening rates that were not based on risk between African Americans with and without a family history of CRC did not differ significantly for either endoscopy procedures (58.0% and 49.8%, respectively; P 5 .07) or FOBT (48.9% and 42.7%, respectively; P 5 .18; data not shown). The completion rate for risk-appropriate, timely CRC screening was 39.8% for African Americans with a family history of CRC and 53.4% for African Americans without a family history of CRC (P < .05). Among white respondents in the 2002 MCS with a family history of CRC (n 5 293), 65.5% completed risk-appropriate, timely CRC screening; but only 39.8% (n 5 35) of African Americans with a family history of CRC reported having received such screening (P < .05) (Table 2).
Individuals Without a Family History of CRC (n 5 492) Age, sex, activity level, daily servings of fruits and vegetables, history of prostate-specific antigen (PSA) screening in men, cigarette use, employment, insurance, having a primary healthcare provider, and recommendation of either endoscopy or FOBT were evaluated in the multivariate model after qualification in univariate analysis (see Table 3). In multivariate analysis, however, only PSA screening history, activity level, and provider recommendation were predictive of risk-appropriate, timely screening. Men with a history of PSA screening were 2.68 times more likely (95% confidence interval [95% CI], 1.01–7.81 times more likely) to have had risk-appropriate,
279
TABLE 1 Sample Characteristics by Family History No family history of CRC (N 5 492) Characteristic
No.
Sex Men 177 Age, y* 40–49 50–64 329 65 163 Mammogram history (women) Yes 308 PSA history (men) Yes 124 BMI, kg/m2 24.9 121 25–34.9 172 35 168 Activity level Minimally active 196 Moderately active 166 Very active 103 Daily fruits and vegetables 2 131 3–4 189 5 158 Cigarette use Current smoker 91 Alcohol use Moderate to heavy drinker 85 Concern about cancer* Very concerned 229 Perceived amount of cancer in family* High 43 Perceived risk of cancer* High 76 Highest level of education
