|Home | About | Journals | Submit | Contact Us | Français|
The relationship between obesity and screening mammography adherence has been examined previously, yet few studies have investigated obesity as a potential mediator of timely follow-up of abnormal (Breast Imaging Reporting and Data System [BIRADS-0]) mammography results in minority and medically underserved patients.
We conducted a retrospective cohort study of 35 women who did not return for follow-up >6 months from index abnormal mammography and 41 who returned for follow-up ≤6 months in Nashville, Tennessee. Patients with a BIRADS-0 mammography event in 2003–2004 were identified by chart review. Breast cancer risk factors were collected by telephone interview. Multivariate logistic regression was performed on selected factors with return for diagnostic follow-up.
Obesity and gynecological history were significant predictors of abnormal mammography resolution. A significantly higher frequency of obese women delayed return for mammography resolution compared with nonobese women (64.7% vs. 35.3%). A greater number of hysterectomized women returned for diagnostic follow-up compared with their counterparts without a hysterectomy (77.8% vs. 22.2%). Obese patients were more likely to delay follow-up >6 months (adjusted OR 4.09, p=0.02). Conversely, hysterectomized women were significantly more likely to return for timely mammography follow-up ≤6 months (adjusted OR 7.95, p=0.007).
Study results suggest that weight status and gynecological history influence patients' decisions to participate in mammography follow-up studies. Strategies are necessary to reduce weight-related barriers to mammography follow-up in the healthcare system including provider training related to mammography screening of obese women.
Although mammography has been associated with a 44% reduction in risk of late-stage breast cancer, mammography screening can enhance survival rates only if additional diagnostic testing follows initial abnormal or inconclusive mammograms for definitive determination of breast cancer status.1,2 One factor that diminishes the maximal benefit of mammography as an early detection breast cancer tool is inadequate follow-up for abnormal or inconclusive mammograms.3 For women ultimately destined to have breast cancer, even a short delay in detection could have significant consequences. A meta-analysis demonstrated that women with delays of 3–6 months between index abnormal mammography findings and diagnostic resolution had 7% lower 5-year survival than those with shorter delays (odds ratio [OR] for death 1.24, 95% CI 1.17–1.30).4
The purpose of this study was to evaluate factors of delay of abnormal mammography follow-up in minority/medically underserved women. We hypothesized that delay in timely mammography resolution of an abnormal mammography status by minority/medically underserved women was related to (1) the patient's sociodemographic status, (2) biomedical breast cancer risk factors (e.g., family history, gynecological factors), and (3) lifestyle factors (e.g., weight status, physical activity, smoking status, and alcohol intake).
The design was a retrospective cohort study. Patients were minority/medically underserved women having screening mammography performed at the breast health clinic located in the Nashville General Hospital at Meharry, Davidson County/Nashville, Tennessee. The American College of Radiology (ACR) Breast Imaging Reporting and Data System (BIRADS™) categories used to standardize interpretation of mammography findings were used to identify eligible study participants. The eligibility criteria for the study were (1) an abnormal mammography result requiring diagnostic follow-up prior to the next routine screening of BIRADS-0,5 (2) age of 40–75 years between January 2003 and December 2004, (3) residency in an eight county radius around Davidson County, and (4) ability to provide informed consent. Eligible women were identified using retrospective chart review abstracted from radiology clinic notes in the medical records. The administrative appointment database, Mammography Reporting System (MRS), was abstracted for the date of the abnormal mammography finding, demographic data, and the date when the patient returned for diagnostic follow-up.
Follow-up was defined as return for diagnostic resolution within 6 months from the date of the abnormal screening mammography finding BIRADS-0 (incomplete, needs additional imaging).5 Diagnostic resolution included such procedures as additional mammography views (compression or magnification), biopsy (core or excisional), and ultrasonography. Although no gold standard exists to determine what constitutes timely follow-up,6 we dichotomized this variable based on literature that suggests diagnostic and treatment delays of 3–6 months may negatively impact survival.2 Study participants who received inadequate follow-up were defined as BIRADS-0 patients without diagnostic resolution within 6 months after their index abnormal finding. Study participants who received adequate follow-up were defined as BIRADS-0 patients with diagnostic resolution within 6 months after their index abnormal finding. The disposition of the follow-up of the BIRADS-0 was determined prior to the survey interview, using the reference date of the index mammography finding with the BIRADS-0 result in the medical record and abstracting the records for all mammograms that occurred or did not occur subsequent to that date. Women with a prior history of cancer or who were deceased at the time of the scheduled interview were not eligible for the study.
Study procedures were approved by the Institutional Review Boards at Meharry Medical College and Nashville General Hospital prior to study implementation. Those BIRADS-0 patients identified as eligible from the medical record review were recruited for the survey interview portion of the study. Eligible women were sent a letter from the medical director of the breast health center (A. G.) inviting them to participate in the study interview, explaining the study, and informing them that they would be reimbursed $15 for their time and effort in participating in the study. A toll-free number was provided for patients to call for study refusal. The postal service's ancillary service endorsement (“Forwarding Service Requested”) was written on the envelopes of the introductory letters to provide free forwarding of the letter to the patient's most up-to-date address if it was different from the address listed in the patient's medical record. In many instances, patients had relocated, and the introductory letter was returned as undeliverable. If another address was obtained by the research staff, a second introductory letter was sent to the new address. Alternate addresses were attempted when necessary, including the emergency contact identified from the patient's medical record and the billing system.
All patient contacts and interviews were completed by one full-time trained research assistant, who completed a week of training on psychosocial and epidemiological data collection techniques: obtaining informed consent and Health Information Portability and Accountability Act (HIPAA) waivers over the phone, explaining study confidentiality, probing techniques, how to pose study questions without being leading or biased, data entry, and verbatim editing. The research assistant piloted these techniques through an epidemiological survey instrument with 20 subjects.
After the study invitation letter was sent, the women were allowed 10 days to call the refusal line. A call center unrelated to study staff and location was established to answer the refusal line to avoid coercion of study participants. Tacit consent was obtained if the woman did not call the refusal line within 10 days. After the refusal period, the research assistant attempted to schedule the patient for the survey interview. If the woman was reached, an appointment for completion of the survey was scheduled contingent on the patient's receiving the study materials. Participants were sent a study packet consisting of an informed consent form, an HIPAA waiver, a reimbursement form, and the study instruments. During the scheduled telephone appointment, participants verbally consented, and the survey instrument was read to them over the telephone.
Demographic and clinical variables, such as race, ethnicity, age, marital status, health insurance and social history, mammography interpretation, and radiologist recommendation, were obtained from medical record chart review by a certified tumor registrar. The survey contained 166 items and was adapted from Champion's Community Interventions to Increase Mammography Screening (CIIMS) study.7 Telephone interviews assessed epidemiological and clinical risk factors, including family history of cancer, smoking status, alcohol intake, gynecological factors, hormonal history, and menopausal status. Self-reported weight status and height were used to calculate the body mass index (BMI) expressed in kg/m2. Physical activity was measured by recording the past 10 years of regular exercise activities in metabolic equivalents (MET) of duration (hours/week) values, years of participation, and average energy expenditure during the past 10 years period (MET-hours/day/year) using standard methods.8 Eight psychosocial constructs were examined using multivariate logistic regression for their association with abnormal mammography follow-up: (1) perceived barriers,7 (2) perceived benefits,7 (3) perceived susceptibility,7 (4) self-efficacy,9 (5) cancer fatalism,10–12 (6) health temporal orientation,13 (7) multidimensional health locus of control with two dimensions (internal/external),14 and (8) spiritual health locus of control measuring active and passive spirituality.15,16
A total of 227 women were identified who received a screening mammography between January 2003 and December 2004 with an abnormal BIRADS-0 result. Of these women, 11% (n=18) were not eligible to participate in the study because of age, not residing in an eight county radius around Nashville, a history of cancer, inability to provide consent, or being deceased or too ill to participate, leaving 209 women. In addition, we were not able to contact 68 women because of disconnected phones, changed phone numbers, changed addresses, or no answer to the telephone calls (n=27, 11.9% of the inadequate follow-up patients; n=41, 18.0% of the adequate follow-up patients). An additional 18% refused (n=17, 7.5% of the inadequate follow-up patients; n=24, 10.6% of the adequate follow-up patients), and 10.5% partially completed the interview (n=13, 5.7% of the inadequate follow-up patients; n=11, 4.8% of the adequate follow-up patients). Findings on the 151 women who did not participate in this study after 3 months of effort or 12 attempts to contact them for study participation have been reported elsewhere.17
Overall, there was a marginally significant disparity in age between women who did and did not participate. A higher frequency of women between the ages of 40 and 50 years (56.3%) did not participate vs. women who did participate in this age range (47.7%) (p<0.07). More inadequate follow-up patients were reached than adequate follow-up patients (61% inadequate follow-up vs. 49% adequate follow-up, p=0.03), and African American women with inadequate follow-up status were more likely to respond than African American women with an adequate follow-up status (odds ratio [OR] 4.07, 95% confidence interval [CI] 1.59–10.30, p=0.003). Although 41 women refused to participate, only 29 women self-reported their race. Of the 29 women with race identified who refused to participate, 52% (n=15) were African Americans. Of 24 women who agreed to participate but did not complete the interview or withdrew their consent, 58% (n=14) were African Americans. Minority women who were located from the contact information in the medical record and who were approached for study participation were not opposed to enrolling in abnormal mammography follow-up research. However, a higher percentage of minority women than their Caucasian counterparts failed to complete the detailed phone interview.17
Baseline demographic, clinical, and survey data were compared between inadequate follow-up patients and adequate follow-up patients. Bivariate associations between the covariates and two groups (based on timely follow-up) were assessed with univariate logistic regression; ORs with 95% CIs are reported. Significant predictors of inadequate follow-up identified in the univariate logistic regression model were further adjusted for other potential confounders. We ran two models for adjustment. In model A, each significant predictor variable was analyzed independently adjusting for sociodemographic variables, including age, race, education, income, insurance, and menopausal status. In model B, all significant predictor variables were included in the model along with the sociodemographic variables and menopausal status. The objective of this pilot study was to estimate effect sizes for this population, so that full-scale and appropriately powered studies can be planned. Because of the small sample size, we were unable to look at interaction terms between the predictor variables. All statistical analyses were performed with SAS version 9.1. (SAS, Cary, NC).
A total of 76 interviews was achieved, including 35 inadequate follow-up patients and 41 adequate follow-up patients, for an overall response rate of 54% among inadequate follow-up patients and 46% among adequate follow-up patients. Of these, 59% (n=39) of African American women, 69% (n=29) of Caucasian women, and 50% (n=8) of women self-categorized as race/ethnicity “Other” of Hispanic/Latina or Middle Eastern ethnicity completed the survey. Generally speaking, the psychosocial factors were not associated with the significant covariates found in the current analysis. Results from the psychosocial analyses will be reported in a future companion article.
Table 1 shows the percentages by follow-up status of sociodemographic characteristics and breast cancer risk factors of the participants. There were no statistically more likely to be obese significant differences between inadequate follow-up and adequate follow-up with respect to age, income, health insurance, race, and education. Compared with adequate follow-up patients, inadequate follow-up patients were more likely to be obese (BMI >30kg/m2), and a higher frequency of obese women delayed return for timely mammography resolution compared with nonobese women (64.7% vs. 35.3%). A greater number of women with a hysterectomy returned for diagnostic follow-up than those without a hysterectomy (77.8% vs. 22.2%) (Table 1). Odds ratios for the association of breast cancer risk factors with timely mammography follow-up are shown in Table 2. In multivariate models adjusted for sociodemographic variables and menopausal status, obese women were more likely to experience a 6 month delay in mammography follow-up compared with women who were nonobese (OR 4.09, 95% CI 1.26–13.22, p=0.02). A significant association also was observed between hysterectomy status delay in mammography follow-up. Women with hysterectomies were more likely to return for timely mammography follow-up within 6 months than were women who did not have hysterectomies (OR 7.95, 95% CI 1.78–35.49, p=0.007). In the second multivariate model in Table 2, adjusting for sociodemographic variables, menopausal status, and the three significant variables listed in Table 1, BMI >30kg/m2 (obesity) was no longer a significant predictor of return for abnormal mammography follow-up.
In this study, more women who experienced a delay in timely mammography were obese than were women obtaining timely mammographic follow-up. The reasons obese women are less likely to obtain abnormal mammography follow-up are still poorly understood. We surmise that reasons for this discrepancy may be similar to those women in the obese population who do not receive screening mammography or Pap smear screening for cervical cancer.
Previous research has shown either a positive,18 inverse,18–29 or null19,20,23 association between obesity and adherence to cancer screening tests. Disparities in cancer screening in obese women may be exacerbated by patient and clinician attitudes. Increasing weight has been associated with having a negative self-perception and poor body image26,28,29 and being reluctant to obtain pelvic examinations.20–28 Obese women may experience greater embarrassment about weight and have a perception of increased pain and discomfort from gynecological procedures.30 Additionally, obese women may delay screening mammograms and Pap smears because they encounter negative attitudes, judgments, and bias from health professionals.20,28 They may delay or avoid preventive visits because they do not want to be weighed or receive lectures about their weight.4,31,32 Amy et al.28 found that 41% of women with BMIs ranging from 25 to 122kg/m2 delayed seeking cancer screening tests because of their weight. An association emerged where the percentage of women who delayed seeking cancer screening increased significantly as BMI increased. Unlike our population of medically underserved patients, access to healthcare for these obese women was not a barrier, as >90% had health insurance.28
In terms of disparities, this study has only indirect implications. The minority and medically underserved participants analyzed in this study are socioeconomically disadvantaged and have limited or no health insurance to cover medical costs; they were not compared to a nonminority, adequately served comparison group. Therefore, we are limited in evaluating these effects.
The findings about hysterectomy status as related to mammography follow-up were unexpected. One plausible interpretation of this finding is that hysterectomized women are more likely to return for follow-up diagnostic resolution if they are on estrogens and progestin regimens, which have a higher probability of requiring short interval follow-up for mammography screening.25,33,34 Previous studies35 show that women who take hormone replacement therapy (HRT) are more likely to use various medical and screening services, including mammography, than women who do not use HRT.36,37 Furthermore, the relationship between HRT and hysterectomy may cause increased vigilance and cancer surveillance for hysterectomized women, thereby prompting providers to recommend mammography follow-up.38
Hysterectomy may be a key discriminator of follow-up of abnormal mammography, as this finding suggests that women may be referred or reminded more often by their healthcare providers to obtain preventive health visits or to be more conscientious about their health overall. Healthcare provider recommendations are among the strongest independent predictors of a person's decision to have a cancer screening test.21,35,39 It can be posited that women who have received a hysterectomy are exposed to a greater number of encounters with specialists in women's health, which provides heightened opportunities to refer hysterectomized women to mammography screening.
This study is not population based and has patients from a single health center. Our study results are limited to medically underserved women served by our public hospital who could be reached for interview. Despite this limitation, our statistically significant findings on obesity and hysterectomy status provide factors for additional studies to examine for generalizability.
The subjects in this study were recruited to participate in a telephone questionnaire up to 1 year after their index abnormal mammography finding (BIRADS-0). Recall bias represents a threat to the internal validity and a tendency toward misclassification errors among study subjects. Because our study relies on self-reporting of past behaviors, patients with inadequate follow-up may recall their behaviors in a way different from that of adequate follow-up patients. As returning for mammography follow-up is a socially desirable behavior, a more conservative estimate of inadequate mammography follow-up may have occurred, attenuating the true association and suggesting that some of the null associations in our study may be due to misclassification error in inadequate mammography follow-up assessment.40,41 When possible, the medical records and questionnaire were used in concert to reduce recall bias and verify the information of the last mammogram reported after the index abnormal mammogram. The number of women for whom the outcome, mammography follow-up in ≤6 months, was verified by self-report is 71 vs. 5 women whose outcome was verified by medical record abstraction.
The primary instrument used to obtain mammography follow-up information was tested for reliability and validity in a population of medically underserved and minority mammography screening patients from the CIIMS study.7 In addition, distinct from other studies,18,19,25–27,42 mammography screening history was obtained from medical record abstraction as opposed to patient self-report. Although self-report of mammography history is highly correlated with recorded information,26 self-reported mammography use tends to be overestimated, and the validity of self-reported mammography is relatively low in medically underserved minority women.42
Locating respondents for interview was more difficult than anticipated, contributing to our lower than planned patient study participation rates. The concern with diminished participation in this study is selection bias resulting in differential exposure between refusing and participating inadequate and adequate follow-up patients, respectively.41,43 Medical record data collected from the nonrespondents provided the opportunity to examine possible selection bias. Nonrespondents were more likely to be African American or in the “Other” ethnic/racial category and between the ages of 40 and 50 years of age.17 Adequate participation and retention are critical methodological issues in observational epidemiology. Where there are differences in participation by race, findings may be misinterpreted or irrelevant to the target population of interest.44–47
The lower than anticipated numbers of study participants resulted in a study underpowered to detect interactions in the multivariate logistic regression models. We discovered that the obesity findings were no longer significant when all the variables were adjusted for in the model simultaneously. A larger sample size is required to detect an OR of 4 for BMI >30kg/m2 (obesity) with eight other covariates in the model. We attribute the inability to detect small effects shown as significant in the mammography screening literature (smoking, alcohol intake) to inadequate power.24,48 Moreover, a concerted effort was made to contact, consent, and interview women. On average, it took 82 days and 4.5 (range 1–19) attempts to reach each woman in our sample. The majority of African American participants (58%) did not complete the study or refused to participate once located (52%), contributing to our lower than planned completion and consent rates.17
We surmise that two factors led to poor participation: (1) the length of the questionnaire (166 items) was structured to be administered in one telephone session and may have caused a response burden, and (2) this study entailed one telephone interaction with the participant, but we did not provide informative feedback on how study participation could benefit her.49 In an attempt to maximize retention of minority study participants, the interviewer was an ethnic minority female, although not of African American descent. In a previous epidemiological case-control study of breast cancer, African American women were less likely to complete the interview when they were invited to complete the survey by a non-African American interviewer.47
This is an exploratory retrospective cohort study that provides preliminary evidence that weight status and gynecological factors are constructs that appear to influence patients' decisions to participate in mammography follow-up studies. These factors may be surrogate measures that mediate the observed associations. There are a number of potentially mutable issues that influence the patient's decision to participate in mammography resolution, ranging from biases in the healthcare system to doctor's recommendations that affect the patient's ability to obtain diagnostic follow-up.
These results have valuable clinical and public health implications for this high-risk population, adversely affected by breast cancer mortality. Reducing disparities in breast cancer diagnosis and survival requires timely and efficient follow-up diagnostic care once an abnormality has been detected. Additional research is needed to better understand the mechanisms underlying associations observed in this report. Further refinement of a prospective, longitudinal design with an initial assessment of baseline anthropometry, gynecological history, and external factors in the healthcare environment at the time of the index screening mammography abnormal finding is currently underway. Obese and nonhysterectomized women in this sample of minority, underserved patients exhibit delayed diagnostic resolution of abnormal breast cancer screening mammography and would benefit from further intervention to affect disparities in breast cancer survival.
We thank Drs. Ken Wallston and Ted Speroff for their helpful comments. Research was funded by the National Cancer Institute U54 CA915408-04. A.M.F. was funded by ACS grant MRSGT-07-008-01-CPHPS, the Clinical Research Center of Meharry Medical College, grant P20RR011792 from the National Institutes of Health and RCMI Clinical Research Infrastructure Initiative, and National Cancer Institute grant U54 CA915408-06.
The authors have no conflicts of interest to report.