|Home | About | Journals | Submit | Contact Us | Français|
In March 1997, the American Cancer Society (ACS) updated its recommended mammography screening interval for women ages 40–49 years from once every 1 to 2 years to once every year. At the same time, the National Cancer Institute (NCI), which had previously not recommended routine screening of women in their 40s, began recommending screening at 1 to 2-year intervals. These events occurred during the data collection phase of a prospective study of mammography screening and, thereby, provided an unexpected opportunity to examine the potential influences of changing guidelines on women’s beliefs about how frequently they should obtain screening exams.
This analysis included 1451 African American and white women ages 40–79 years, who obtained an “index” screening exam between October 1996 and January 1998. In baseline and 2-year follow-up telephone interviews, respondents provided information on demographic, socioeconomic, health history, medical care, behavioral and psychosocial factors, and on how frequently they believed women of their age should obtain screening mammograms.
After the ACS and NCI announcements of new screening guidelines for women in their 40s, a significant increase in endorsement of annual screening among women ages 40-49 years was observed, consistent with the ACS recommendation for annual screening in that age group. No increase in endorsement of annual screening among women ages 50 years and older was evident during the same time period.
Women’s beliefs about how frequently they should obtain mammography screenings appear to change in response to changes in recommendations of high-profile health organizations, particularly when those recommendations call for an increase in screening.
The longstanding controversy over the efficacy of mammography screening in women ages 40 to 49 years has been reflected in disparate screening guidelines recommended by major organizations. Between 1993 and 1997, for example, the American Cancer Society (ACS) recommended that women in their 40s obtain mammography screenings every 1 to 2 years, whereas the National Cancer Institute (NCI) and the United States Preventive Services Task Force (USPSTF) did not recommend routine screening of women in that age group.1,2 Then in January 1997, the National Institutes of Health (NIH) convened a consensus development conference to review new evidence and reconsider mammography screening recommendations for women ages 40–49 years. The panel’s majority report, which did not support screening women in their 40s,3 sparked public outcry and vigorous debate.4 Soon after, in March 1997, the ACS conducted its own review and concluded, in contrast to the NIH panel, that recent data provided strong evidence of the benefits of mammography screening among women ages 40–49 years.5 And, on March 23, 1997, the ACS announced that it was changing the recommended mammography screening interval for women beginning at age 40 from 1 to 2 years to annual screening, arguing that more frequent exams were needed to achieve maximal benefit from this screening tool.5 On March 27, 1997, the NCI reversed its earlier guideline and joined ACS in recommending regular mammography screening of women ages 40 years and older; however, NCI recommended a screening interval of 1 to 2 years.6 The USPSTF adhered (at that time) to its 1996 screening guideline that did not recommend mammography in women younger than age 50 years.1
The events described above involved 3 highly publicized announcements of divergent screening recommendations for women in their 40s in less than 3 months. In a limited number of studies, women’s awareness of and reactions to this 1997 mammography screening debate were examined.7-9 Two studies found that the majority of women were aware of the controversy over the efficacy of screening mammograms,7,9 and 1 study found that more women in their 40s than women in their 50s were confused about how frequently they should obtain screenings.8 The latter study also found that women who were confused were less likely to receive screenings according to recommended guidelines.8 However, these retrospective studies surveyed women after the change in guidelines and could not compare women’s beliefs before and after the guidelines changed. A prospective analysis was needed to assess the impact of changes in mammography screening guidelines on women’s intentions to screen, which are highly correlated with actual screening behavior.10
We based our analysis on data from the cohort study, Race Differences in the Screening Mammography Process, that enrolled women between October 1996 and January 1998 and offered a unique and fortuitous opportunity to prospectively examine the impact of the March 1997 changes in mammography screening guidelines on women’s beliefs about how often they should obtaining screenings. Because the study included women ages 40–79 years, we were able to compare the beliefs of women in their 40s (for whom new guidelines were recommended) with those of older women (for whom screening guidelines did not change). We hypothesized that the guideline changes would influence beliefs about screening frequency for women ages 40–49 years but not for women ages 50 years and older. Because the study included a wide range of variables, we were able to 1) examine the relation between guideline changes and women’s beliefs about how frequently they should obtain screening exams independent of many potential confounding variables and 2) examine potential mediators of the relation between guideline changes and women’s beliefs. We hypothesized that this relationship may be mediated by psychosocial factors, including perceived susceptibility to breast cancer and perceived usefulness of mammography, and by a physician’s recommendation to obtain a mammogram. In addition, because the study included relatively equal proportions of African American and white women, we were able to examine whether guideline changes had a differential impact on these racial/ethnic groups.
All eligible African American and randomly selected white women, matched on mammography facility and date of mammogram ages 40–79 years, presenting for an “index” screening mammogram between October 1996 and January 1998 at 1 of 5 hospital-based mammography facilities in Connecticut were invited to participate in this study. Because the cohort was followed to prospectively assess screening behavior, only asymptomatic women were eligible; women with a history of breast disease and those who had received diagnostic mammograms to evaluate symptoms of disease were excluded. In accordance with general population screening recommendations, women younger than age 40 years were excluded as were those older than age 79 years for whom there are no clear guidelines for continued screening. Of 1982 women who were eligible and invited to participate, 1451 (73%) completed a baseline interview (44% African American and 56% white). Of those, 1249 (86%) women completed a 2-year follow-up interview.
Approval from the institutional review boards of Yale University School of Medicine and each participating hospital were maintained throughout the study period. Informed consent was obtained from each participant. Baseline telephone interviews were conducted by trained interviewers 1.5 months, on average, after the index screening mammogram (standard deviation from the mean [SD] ± 0.85 month). The interviews gathered information on demographics, socioeconomic status, access to care, and behavioral and psychosocial factors, including information on women’s beliefs about how frequently they should obtain screening mammograms. The baseline interviews were conducted between October 1996 and March 1998. (The guideline changes occurred during this period, ie, late March 1997.) Follow-up telephone interviews were conducted 29 months, on average, (SD ± 1.42 months) after the baseline interview. The follow-up interviews were all conducted after guideline changes were announced.
Items included in the baseline and follow-up telephone interviews were developed from 1) consultation with radiologists who provided mammograms to ethnically diverse groups, 2) issues identified in an African American focus group, and 3) published literature.14-21
The primary predictor was whether the baseline interview took place before or after April 1, 1997 (ie, before or after the ACS and NCI announced new mammography screening guidelines for women in their 40s in late March 1997). The outcome, assessed at baseline and follow-up, was how frequently women believed they should obtain screenings, assessed by the question: “How often do you think a woman your age should have a screening mammogram?” This was an open-ended question, with responses later categorized for analysis. Initially, we examined 3 categories, 1) endorsement of annual screening, which corresponded to the new ACS recommendation, 2) endorsement of screening every 1 to 2 years, which corresponded to the new NCI recommendation, and 3) other (endorsement of another screening interval, eg, every 2 to 3 years). Because we found substantial changes in endorsement of annual, but not biennial, screening preguideline and postguideline changes (see Results section below), we combined “every 1 to 2 years” with the “other” category and used 2 categories in subsequent analyses, 1) whether women believed they should be screened annually, and 2) other.
In the analysis of baseline data, we examined 2 potential modifiers of the effect of interview date (before versus after guidelines changed) on beliefs about screening frequency. These modifiers were age (40–49 years versus 50–79 years) and self-identified race/ethnicity (African American versus white). We also examined several covariates potentially associated with age or with mammography screening behavior or intention to screen14,19,22-41 for inclusion in multivariate analyses as follows: 1) demographic and socioeconomic factors: marital status (married/living as married, other), education (<12 years, 12 years, >12 years), and family income (<$15,000, $15,000–$49,999, ≥$50,000); 2) breast cancer risk factors: first-degree relative with breast cancer (yes, no), age at menarche (<12 years, 12 to 14 years, >14 years), age at first birth (nulliparous, first birth at <30 years, first birth at ≥30 years), menopausal status (completed, not completed, hysterectomy-associated menopause), and use of hormone replacement therapy (HRT) (never, <1 year, 1 to 5 years, >5 years); 3) health status and behaviors: body mass index (BMI) (17.5–24.9 [underweight/normal], 25.0–29.9 [overweight], ≥30 [obese]), pack-years of cigarette smoking (0–1, 1–10, >10), physical exercise at least once a week (yes, no), self-rated health (excellent, good, fair, poor); and 4) features of the index screening and of the respondent’s medical care: has mammography insurance (yes, no), abnormal result on the index screening (yes, no), pain experienced during the index screening (none, a little, a fair amount, a lot), embarrassment experienced during the index screening (none, a little, somewhat, very), history of nonadherence to pre-1997 ACS mammography screening guidelines (ie, 1 screening every 1 to 2 years for women ages 40–49 years or annual screening of women ages 50–79 years), and continuity of care (same provider over the past year, different provider).
In addition, we examined several potential mediators of the relation between date of interview and beliefs about screening frequency. These included receipt of a health professional’s recommendation to obtain a mammogram (yes, no) and several psychosocial factors drawn from the health-belief model,42,43 and the health locus-of-control construct44-46 that have been associated with mammography screening behavior and intentions to screen1,11,47-49 and were hypothesized to be associated with the change in screening guidelines. The psychosocial variables included perceived susceptibility to developing breast cancer (don’t know, not likely, a little likely, somewhat likely, very likely), perceived usefulness of mammography (very useful, don’t know, somewhat useful, a little useful, not useful), anxiety experienced at the time of the index mammogram (none, a little, a fair amount, a lot), perceived control over cancer development (none, a little, some, a lot, don’t know), and perceived control over cancer recovery if diagnosed (none, a little, some, a lot, don’t know).
All analyses were conducted with SAS software, version 8.2 (SAS System for Windows; SAS Institute, Cary, NC) and used 2-sided P-values. Analyses began with descriptive statistics to examine frequency of selected respondent characteristics and with chisquare tests to examine whether these characteristics differed by age.
Then, using baseline data, unconditional logistic regression analysis that provided maximal likelihood estimates of odds ratios (ORs) and 95% confidence intervals (CIs) was undertaken to examine how frequently women believed they should obtain screening mammograms in relation to whether they were interviewed before or after the guidelines changed. We included in multivariate models those potentially confounding variables, detailed above, that were associated at the bivariate level with age or with the outcome (P < .05). We assessed the potential mediating effect of psychosocial factors and of receipt of a physician’s recommendation to obtain a screening by adding each of these variables to the adjusted model in the total sample and among women ages 40–49 years. A variable was considered a mediator if it altered the estimated OR by ≥10% and was significantly associated with the predictor and the outcome. Potential modifying effects of age and race on the relationship between beliefs about screening frequency and interview date were examined by adding interaction terms of age, race, and interview date to the adjusted model. For significant interactions, ORs and 95% CIs for each subgroup were obtained by summing the relevant estimates and calculating the respective variances of these sums.
With McNemar’s test for matched pairs, we then examined within-person change, from baseline to follow-up interview, in beliefs about how frequently mammograms should be obtained. These analyses were stratified by age group (40–49 years and 50-79 years) and by whether the baseline interview was administered before or after changes in screening guidelines.
Women differed by age on several characteristics (Table 1). Women ages 40–49 years had more education and higher family incomes than women ages 50–79 years. Younger women were also more likely to have adhered to pre-1997 ACS mammography screening guidelines for their age group. The majority of women in both age groups had health insurance coverage for screening mammograms; older women were more likely to have had a consistent healthcare provider.
Among women ages 40–49 years, in keeping with new ACS guidelines, the percentage who endorsed annual screening increased from 49% to 64% after the guidelines changed, whereas the percentage who endorsed screening every 1 to 2 years and the percentage who endorsed “other” screening intervals decreased (P < .0046; Table 2). In contrast, among older women (for whom there was no change in guidelines), the percentage who endorsed annual screening increased only marginally and not significantly after the guidelines changed.
In a logistic regression analysis across age groups, women interviewed after the guidelines changed were more likely to endorse annual screening than women interviewed before the guidelines changed (OR, 1.43; 95% CI, 1.12–1.82). The addition of multiple covariables (age, income, education, mammography insurance, continuity of care, result of the index screening, smoking, exercise, HRT use, age at first birth, menopausal status, embarrassment experienced during the index screening, history of nonadherence to mammography screening guidelines) had virtually no effect on this association (adjusted OR, 1.44; 95% CI, 1.09–1.92). Furthermore, the interactions of race and interview date and of race, age, and interview date were not significant. However, the interaction of age and date of interview was significant (P = .0336) in the adjusted model. The calculated ORs and 95% CIs for each age group indicated increased endorsement of annual screening after (versus before) the guidelines changed only among the younger women to whom the changes applied. The age groups (with their ORs and 95% CIs) were 40-49 years (adjusted OR, 1.96; 95% CI, 1.31–2.92) and 50–79 years (adjusted OR, 1.06; 95% CI, 0.70–1.59). None of the candidate variables that we examined mediated the effect of date of interview on women’s endorsement of annual screening across age groups or specifically among women ages 40–49 years.
When beliefs about screening frequency were compared between baseline and follow-up interviews (Table 3), all women ages 50–79 years, and women ages 40–49 years whose baseline and follow-up interviews both occurred after the guidelines changed, showed little change in endorsement of annual screening from baseline to follow-up interview. However, there was a significant change in beliefs about screening frequency from baseline to follow-up among women ages 40–49 years whose baseline interview occurred before the guidelines changed and whose follow-up interview occurred after those changes. A substantial proportion (31.6%) of these women switched from not endorsing annual screening of women in their 40s at baseline (before the change in guidelines) to endorsing annual screening at follow-up (after change in guidelines) (P <.0001).
Although some defended the January 1997 NIH consensus development panel majority report that did not recommend mammography screening of women ages 40–49 years,50-52 by March 1997, 2 influential organizations, NCI and ACS, had publicly recommended screening women in their 40s. However, inconsistency between these organizations remained with respect to the optimal screening interval for women in that age group. In our study, consistent with new ACS guidelines, there was a decrease in endorsement of biennial screening and an increase in endorsement of annual screening among women in their 40s after public announcements of new screening recommendations. This finding did not differ by race. In addition, there was no significant increase in endorsement of annual screening among women ages ≥50 years. This suggests that the increased endorsement of annual screening in the younger group may be attributed to the guideline changes that were specific to that age group. We were unable, however, to identify factors that might have mediated the relation between changes in screening guidelines and increased endorsement of annual screening among younger women. Additional study is needed to identify such factors and to develop interventions to further impact beliefs about screening frequency.
Nonetheless, the endorsement of more frequent, annual screening mammograms that we observed is consistent with evidence of considerable enthusiasm for cancer screening among persons in the United States.53 For example, following an earlier round of the mammography screening controversy in 1993, when NCI announced that it would no longer support mammography screening of women in their 40s, the vast majority of women in that age group planned to continue with their screening regimens.7,54 In addition, many physicians planned to continue recommending screening to women in their 40s.55,56 This public penchant for screening was reflected in and buttressed by news media, where coverage between 1990 and 1997 largely supported mammography screening.4 Even when reporting on the 1997 NIH consensus development panel’s finding of inadequate evidence for screening younger women, the majority of news stories actually favored screening.57 These findings, which are consistent with our results, suggest that new screening guidelines are more likely to be supported when they call for more frequent screening but not when they recommend less.
Our study has several strengths, including its prospective design, large sample size, broad age range, and ethnic diversity among respondents. In addition, by controlling for multiple potentially confounding variables, we were able to isolate the impact of guideline changes on beliefs about screening frequency and to demonstrate the broad impact of these recommendations across race, socioeconomic status, and other factors. A potential limitation is that our analysis of baseline data was an aggregate-level analysis that compared the group of women interviewed before the guidelines changed with the group interviewed after the guidelines changed with respect to endorsement of annual screening. In such a group-level analysis, it is not possible to determine whether guideline changes influenced beliefs held by individual women. It is, thus, reassuring that results of our individual-level analysis comparing endorsement of annual screening in women at baseline and follow-up were consistent with results of the baseline analysis.
In addition, we could not directly examine whether the increased support of annual screening among women ages 40–49 years that we observed after the guidelines changed was accompanied by an increase in annual screening behavior among women in that age group. All study participants were due for an annual screening after March 1997, thus precluding comparison of screening behavior before and after the guidelines changed. However, other studies have found a strong relation between behavioral intentions and health behaviors.10 Consistent with those findings, our data showed a significant association between endorsement of annual screening (at baseline) and actual annual screening behavior (at follow-up, postguideline change; P <.0001). This suggests that the increased endorsement of annual screening that we observed among younger women after the 1997 guideline changes would lead to an increase in annual screening behavior. Nevertheless, post-1997, women in their 40s remained less likely to obtain annual screenings than women in their 50s and 60s.58-60 This is of concern because of mounting evidence that mammography screening has contributed to the decline in breast cancer mortality observed in the United States61,62 and that annual screening, in particular, may detect smaller and more treatable tumors.63,64 A recent study by White and colleagues65 found that a shorter screening interval was associated with less late-stage disease in younger but not older women with breast cancer, so annual screening may be especially important in women ages 40–49 years.
Since the 1997 mammography screening controversy, the public debate about the efficacy of mammography screening and appropriate screening guidelines has continued. In 2001-2002, for example, a contentious debate ensued after publication of a Cochrane review that found that mammography screening did not decrease mortality in either younger or older women.66,67 More recently, in April 2007, reminiscent of the 1997 NIH consensus development panel’s finding of inadequate evidence to recommend screening women in their 40s, the American College of Physicians withdrew its recommendation for regular mammography screening among women in that age group and instead recommended that physicians tailor recommendations based on individual risk of breast cancer, risks and benefits of mammography, and women’s preferences.68,69 Such actions may fuel future controversies over appropriate target populations, screening intervals, and risks and benefits of available screening modalities. Our findings suggest that in the face of new information, patients are likely to endorse recommendations calling for increased mammography screening. However, further study is needed to examine the impact of new recommendations on screening behavior. In addition, acceptance of new technology, such as magnetic resonance imaging (MRI) for breast screening that was recently recommended by ACS,70 has yet to be evaluated. To better inform health policy and respond to individual patients, ongoing assessment of the influence on beliefs and behavior of controversies over mammography screening and of changes in screening guidelines is needed.
This work was supported by grant RO1 CA70731 from the National Cancer Institute to Dr. Jones, grant RO3 HS11603 from the Agency for Healthcare Research and Quality to Dr. Calvocoressi, and grant RO1 CA95560 from the National Cancer Institute to Dr. Claus.
We thank the following hospitals in Connecticut that allowed access to their patients and medical records, Bridgeport Hospital, Lawrence and Memorial Hospital, St. Francis Hospital and Medical Center, Waterbury Hospital, and Yale-New Haven Hospital.
We also thank Lisa Schlenk, project coordinator, for her invaluable assistance.