Effect of False-Positive Mammograms on Interval Breast Cancer Screening in a Health Maintenance Organization FREE

Screening for breast cancer can reduce mortality rates by 35% to 40% (1 - 3). Despite these benefits, however, a 1995 survey revealed that 30% of eligible U.S. women had not obtained a screening mammogram in the preceding 2 years (4). Consequently, several studies have evaluated factors that might improve screening adherence and further reduce the morbidity and mortality rate associated with breast cancer (5 - 9).

Knowledge of the importance of early cancer detection, receipt of a recommendation from a physician, younger age, and higher socioeconomic status are all associated with increased use of mammography (10). However, we do not know whether the result of mammography affects future screening behavior. It is estimated that approximately 1 in 2 women will have at least one false-positive mammogram after a decade of annual screening (11). It has been suggested that fear of a false-positive result will “dissuade women from obtaining the life-saving benefits of mammography” (12). Women who have had a false-positive mammogram have higher levels of anxiety about breast cancer than other women do, and they have a higher stated intention of obtaining repeated screening mammography at recommended intervals (13 - 15). It is not known, however, whether women follow through with these intentions. The anxiety caused by a false-positive experience may deter women from obtaining subsequent screening. Given the frequency of false-positive mammograms, deterrence from screening could be a serious public health problem.

We investigated whether having a false-positive mammogram alters the likelihood of obtaining breast cancer screening during the next recommended interval. We also examined whether the association between false-positive mammograms and future screening is confounded by age, other risk factors for breast cancer, or previous use of mammography.

We studied women 40 years of age or older who were enrolled in Group Health Cooperative of Puget Sound's breast cancer screening program. Group Health Cooperative is a not-for-profit health maintenance organization in Washington state that has had an organized breast cancer screening program since 1986 (16 - 18). Eighty-five percent of age-eligible women enroll in the program by completing a risk factor questionnaire. The information from the questionnaire is stored in a mainframe database linked to radiology information systems that record when women undergo mammography.

Women enrolled in this health plan may obtain a screening mammogram by referral from their primary care provider or by self-referral. Women who have not undergone screening mammography within their recommended screening interval are identified by routine monthly computerized reviews of the Group Health Cooperative radiology information systems. The women identified by these reviews are mailed letters reminding them to schedule a clinical breast examination and screening mammography through a screening center. These invitation letters summarize the woman's risk factors for breast cancer, inform women of their recommended screening interval, and are signed by the clinician in charge of the breast cancer screening center. A randomized trial of screening recommendation letters from the woman's primary care physician rather than from the director of the breast cancer screening center showed that these two methods had an equal effect (19). Between 1986 and 1992, the breast cancer screening program recommended screening intervals of 1 to 3 years on the basis of selected risk factors for breast cancer, including age 50 years or older, first-degree family history of breast cancer, age at menarche, age at menopause, nulliparity, and first birth after 30 years of age (17 - 18). Beginning in 1993, the tri-annual interval was eliminated and women were invited for repeated screening every 1 or 2 years on the basis of the same risk factors for breast cancer (20). Screening intervals of 1, 2, and 3 years were recommended initially for 2.4%, 81.1% and 16.5% of our cohort, respectively.

Our cohort was composed of women who had screening mammography through the plan's breast cancer screening program between 1 August 1990 and 31 July 1992. A screening mammogram obtained during this time was considered the index mammogram. We limited our cohort to women who were continuously enrolled in the health plan from 1 January 1989 through 31 January 1995. This observation period allowed us to collect data on mammography use before the index mammogram was obtained and ensured adequate follow-up after the index mammogram was obtained. The end of the follow-up interval was chosen to avoid overlap with a large randomized trial, started in February 1995, of interventions designed to increase compliance with mammography recommendations.

Breast cancer screening is conducted through specialized centers where women have a clinical breast examination, undergo bilateral screening mammography, and have an opportunity to review self-breast examination technique (18,20). A nurse at the breast cancer screening center informed patients by telephone of the radiologist's recommendations for additional evaluation and follow-up of abnormalities noted on the mammogram. The follow-up plan was discussed at that time, and a written record of the plan was sent to the primary care physician. Follow-up plans included additional imaging, repeated examination at a shortened interval (≤ 11 months), or surgical evaluation. Women enrolled in Group Health Cooperative may seek care from their primary care provider at any time and were encouraged to see that provider for the clinical breast examination in years in which they were not seen in a screening center.

Health plan databases and the SEER (Surveillance, Epidemiology and End Results) tumor registry were used to identify Group Health Cooperative enrollees with breast cancer (ductal carcinoma in situ, invasive breast cancer, or both).

Group Health Cooperative databases were used to determine whether women underwent repeated breast cancer screening during follow-up. Women were followed for at least 6 months after the recommended date of their next screening mammogram. The 36 mammograms obtained within 10 months of the index mammogram were assumed to be diagnostic, short-interval, follow-up mammograms. In these cases, the date of the index mammogram was reset to correspond to the date of the short-interval mammogram, and the date of the recommended subsequent screening mammogram was then adjusted to reflect this change.

Positive mammograms were those interpreted by the radiologist as indeterminate or positive (suggestive of breast cancer) or those with recommendations for additional evaluation, including follow-up mammography within a shortened interval (≤ 11 months), additional mammogram views, referral to a surgeon, or biopsy (similar to mammography Breast Imaging Reporting and Data assessment categories 0, 3, 4, and 5) (21). Although recommendations for ultrasonography were made, the report forms used by the radiologists during the study period did not include a category for this. We therefore included seven additional index mammograms as positive because they were followed by ultrasonography within 60 days.

False-positive mammograms were defined as positive mammograms in women without a diagnosis of breast cancer during the subsequent 12 months. True-negative mammograms were mammograms with negative or benign readings in women without a diagnosis of breast cancer during the subsequent 12 months.

We compared the baseline characteristics and rates of breast cancer screening up to 6 months after the next recommended screening interval for women with true-negative and those with false-positive index mammograms. The following variables were evaluated as predictors of subsequent screening mammography: result of the index mammogram (false positive or true negative), patient age (<50 years, 50 to 69 years, or ≥ 70 years), family history of breast cancer (yes or no), age of menarche (≤ 10 years or >10 years), age at first childbirth (≤ 30 years or >30 years), nulliparity, age at menopause (<55 years or ≥ 55 years), use of hormone replacement therapy (ever or never), number of previous mammograms, and history of a false-positive mammogram before the index mammogram (yes or no). Simple and multivariate logistic regression modeling were used to evaluate the individual and the simultaneous effect of all predictors. Statistical analyses, including calculation of odds ratios and 95% CIs, were performed by using SAS software (22).

Of the 5520 women 40 years of age or older who underwent screening mammography during the index period, 395 had a history of breast cancer and 19 had had other breast surgery (such as breast reduction or augmentation) and were therefore excluded from the analysis. Of the remaining 5106 women, 46 (1%) had true-positive index mammograms, 813 (16%) had false-positive index mammograms, 4246 (83%) had true-negative index mammograms, and 1 (<1%) had a false-negative index mammogram. Only women with false-positive or true-negative index mammograms (n = 5059) were included in further analyses. Ninety-five women had false-positive results on clinical breast examinations; all of these women also had false-positive index mammograms. Women with true-negative index mammograms and women with false-positive index mammograms did not differ significantly with regard to breast cancer risk factors (Table 1). However, fewer women with false-positive index mammograms had previous mammography compared with women with true-negative index mammograms (62% and 77%; P = 0.001). In addition, more women with a false-positive index mammogram had a previous false-positive mammogram (16% and 11%; P = 0.001).

Of the 5059 women, 3700 (73%) returned for screening mammography by 6 months after their next recommended interval. Of the 813 women with false-positive index mammograms, 602 (73%) returned within 6 months of their recommended screening interval, compared with 3098 (74%) of women with true-negative index mammograms. Crude and adjusted odds for obtaining screening mammography during follow-up are presented in Table 2. The strongest predictor of compliance with subsequent breast cancer screening was the number of mammograms before the index mammogram (unadjusted odds ratio, 1.87 [95% CI, 1.72 to 2.03]). Age 50 to 69 years, a history of a false-positive mammogram before the index mammogram, menopause at or after 55 years of age, and use of hormone replacement therapy were also predictors of compliance with subsequent interval breast cancer screening (Table 2).

Table 2 also shows the adjusted odds ratios of each predictor variable. After simultaneous adjustment for all predictors in the multivariate logistic model, women who had a false-positive index mammogram were more likely to undergo screening during the next recommended interval (adjusted odds ratio, 1.21 [CI, 1.01 to 1.45]). Results were similar when the analysis was limited to the 4414 women whose recommended screening interval was 2 years (adjusted odds ratio, 1.20 [CI, 0.98 to 1.47]). When the analysis was limited to women with no mammograms before the index mammogram (n = 1264), the association between a false-positive index mammogram and screening within the next recommended interval was even stronger (adjusted odds ratio, 1.66 [CI, 1.26 to 2.17]) (Table 3).

In the 3795 women with one or more mammogram before the index mammogram, neither a history of a false-positive mammogram before the index mammogram (n = 610) nor the occurrence of two sequential false-positive mammograms (the index mammogram and the one immediately preceding it) (n = 102) were predictive of compliance with the next screening recommendation (odds ratio, 1.01 [CI, 0.80 to 1.27] and 0.90 [CI, 0.56 to 1.46], respectively) (Table 3).

At Group Health Cooperative, women younger than 50 years of age are included in the breast cancer screening program only if they have one or more first-degree relatives with breast cancer or if they have both of the following risk factors: menarche before 10 years of age and nulliparity or first childbirth after age 30. These younger women may therefore not be representative of all women between 40 and 49 years of age. Some groups suggest that screening mammography in women older than 70 years of age may not be warranted because of suboptimal cost-effectiveness in this age group. Table 4 shows the adjusted odds ratios of each predictor variable stratified by age (40 to 49 years, 50 to 69 years, or 70 years or older). The overall compliance in each age group was 63%, 77%, and 72%, respectively. Although the adjusted odds ratios for a false-positive index mammogram were greater than 1.0 in all groups, all 95% CIs included 1.0.

Our broad definition of a false-positive mammogram included varying types of false-positive mammograms with potentially different implications for subsequent screening and different effects on patient compliance with recommendations for next screening. Subdividing results on the index mammogram into three categories (true-negative, false-positive with further imaging recommendation, and false- positive with biopsy recommendation) yielded results similar to those obtained when the result on the index mammogram was treated as a dichotomous variable (data not shown).

In our study, false-positive mammograms did not adversely affect breast cancer screening behavior during the next recommended interval. After adjustment for age, risk factors for breast cancer, and previous use of mammography, women who had a false-positive mammogram were more likely than women who had true-negative mammograms to return for screening mammograms within the next recommended interval. Women with no previous mammography use seemed to be influenced by a false-positive mammogram more strongly than other women; their adjusted odds ratio for obtaining screening within the next recommended interval was 1.66 (CI, 1.21 to 2.17). Although the study data do not support a decreased likelihood of adherence after a false-positive mammogram in women with one or more previous mammograms (adjusted odds ratio, 1.10), they do not preclude it at the 0.05 error level (CI, 0.86 to 1.40). Receipt of multiple false-positive mammograms, either serially or with an intervening normal mammogram, was not associated with interval screening compliance.

The rate of false-positive mammograms was higher for women without previous mammography. It is possible that radiologists are more conservative in their interpretations of a mammogram if comparison films are not available. A study of a community-based mobile mammography screening program in San Francisco found that the overall rate of abnormal examinations was 7.1% for women without available comparison examinations and 2.0% for the same women on subsequent screening (23).

An important methodologic feature of our study that enhances its validity is reliance on observed as opposed to self-reported behavior of subsequent screening activities. Although mammography self-reports have been shown to be accurate (24), the time of occurrence may be subject to foreshortening and the effect of false-positive results on retrospective recall has not been examined. Our analysis would not be susceptible to bias due to differential reporting of screening behavior in relation to previous test results because observed screening results and behavior were captured by using health plan data systems.

It is unlikely that barriers to screening prevented women with a true-negative index mammogram from having equal access to subsequent screening. The prepaid health plan in which this study was conducted offers mammography screening to eligible enrollees as a benefit without additional charge. Furthermore, all women who meet criteria for screening are invited to make an appointment for a mammogram, and we probably captured all or almost all subsequent screening mammograms done for these women because the plan is prepaid.

Our study had some potential limitations. First, the study was limited to women enrolled in Group Health Cooperative's breast cancer screening program. Like the surrounding community, enrollees in Group Health Cooperative are predominately white, but they are less likely to be in the lowest and highest income brackets than the rest of the local population; therefore, these results may have limited generalizability to populations with greater ethnic and economic diversity. In addition, we did not address whether a false-positive mammogram affects the efficacy of commonly used techniques to enhance compliance with mammography, such as mailed invitations for screening. This topic warrants further investigation.

We used a broad definition of a false-positive mammogram, ranging from repeated imaging either immediately or with a shortened interval to a recommendation for a biopsy. However, this broad definition has been used in other studies of mammography and, in particular, in some studies examining the effects of false-positive mammograms on women's perceived vulnerability to breast cancer and attitudes toward breast cancer screening (14 - 15,25 - 27). Our results did not differ when the analysis was repeated by using only index mammograms with biopsy recommendation as a more rigid definition of a false-positive mammogram.

Some variables found to be associated with screening compliance in previous studies were not available to us, including education and socioeconomic levels, marital status, perceived health status, source and frequency of health care utilization, and physician recommendation. We interpret our findings with some caution and recognize that one or more of these other factors may be responsible for the observed association between false-positive mammograms and subsequent screening. One of the most commonly cited factors, a recommendation from a primary care physician, was evaluated in a randomized trial among Group Health Cooperative enrollees and was not associated with increased screening compared with a recommendation from a breast cancer screening program (19). However, it is possible that recommendation from a primary care physician is influenced by the occurrence of a false-positive mammogram; this possibility may warrant further study. Given these caveats, we think it is unlikely that false-positive mammograms deter women from subsequent screening.

Breast cancer continues to be a common cause of illness and death in women. Rates of compliance with recommended screening intervals vary widely across the United States, and strategies to increase these rates continue to be explored. Given the high cumulative risk for a false-positive mammogram in women who regularly undergo mammography and the importance of regular mammography in decreasing breast cancer mortality rates, it was reassuring to find that a false-positive mammogram did not adversely affect subsequent screening in this cohort.