Effect ofage and Breast Density on Screening Mammograms with False-Positive Findings

Transcription

1 Constance Emily Lehman1 Susan Peacock2 Mariann J. Drucker2 Nicole Urban2 4 Received April 8, 1999: accepted after revision June 2, Supported by grant ROl CA from the National Cancer Institute, National Institutes of Health, Bethesda, MD. 1 Department of Radiology, University of Washington Medical Center, RR21O Health Sciences Center, Box , Seattle, WA Address correspondence to C. 0. Lehman. 2Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. N., MP-804, P 0. Box 19024, Seattle, WA Department of Epidemiology, University of Washington, F-346A Health Sciences, Box357236, Seattle, WA Department of Health Services, University of Washington, H-664 Health Sciences, Box ,Seattle, WA AJR 1999:173: X/99/ American Roentgen Ray Society Effect ofage and Breast Density on Screening Mammograms with False-Positive Findings OBJECTIVE The objective of this study was to examine the effect of breast density and age on screening mammograms with false-positive findings. MATERIALS AND METHODS. The study sample was taken from the Washington State Mammography Tumor Registry. which links data from participating radiologists with the Puget Sound Cancer Surveillance System and the Washington State Cancer Registry. Participants (a = ) were women 35 years old and older who underwent screening mammography for which an assessment and a four-category density rating were coded. A total of 46,340 mammograms were sampled to avoid interpreter bias. In this study of false-positive mammograms, only women with no diagnosis of breast cancer within I 2 months of the index mammogram were included. Logistic regression was used to estimate the odds ratios of a false-positive mammogram being associated with each category of breast density or age. adjusting for the other factor as a covariate. RESULTS. After controlling for breast density, we found that the risk of a false-positive mammogram was not affected by age (p = 27). However, the trend of increasing risk of a false-positive mammogram with increasing breast density was highly significant (p <.001). Women with extremely dense breast tissue were almost two times more likely to have a false-positive mammogram than were women with fatty breast tissue. This effect persisted after controlling for age. CONCLUSION. Breast density. not age, is an important factor when predicting risk of a false-positive mammogram. Breast density should be considered when educating individual women on the risks and benefits of screening mammography. T he National Institutes of Health To our knowledge. no prior studies have 1997 consensus statement, Breast evaluated the effect of mammographic breast Cancer Screening for Women Ages ]. cites the rate of mammograms with false-positive findings as an important factor in assessing the benefits and risks of screening mammography in 40- to 49-year-old women. The percentage of falsepositive mammograms is defined as the ratio of positive mammograms to all mammograms in women with no known breast cancer diagnosis within 1 year of the mammogram (i.e., 100% minus specificity). Prior large-scale studies have reported that the percentage of falsepositive mammograms is higher among women years old than among women 50 years old or older 12-4]. In a recent study, Elmore et al. [51 reported that the percentage of false-positive mammograms decreased from 7.8% for women years old to 4.4% for women years old. density on the risk of a false-positive mammogram. It may be that screening mammography is less specific among younger women because their breast tissue is more radiographically dense. It has been reported that as breast density increases, the individual radiologist s certainty of interpretation decreases and the variability among reviewers increases [61. Although a certain proportion of false-positive mammograms is necessary to maximize breast cancer detection. false-positive examinations mean added financial costs and increased patient morbidity and anxiety [7-9J. Understanding factors affecting false-positive mammograms may be important in improving methods of screening for breast cancer. The aim of our study was to examine the influence of breast density and age on the risk of having a false-positive mammogram. AJR:173, December

2 Lehman et al. Materials and Methods Study Population The Washington State Mammography Tumor Registry was begun in 1994 to assess the feasibility and usefulness of routine linkage of mammography databases with a tumor registry. As part of the pilot study, 42 radiologists contributed data on 97,953 women and 190,647 mammograms between November 4, 1991 and November 17, Screening guidelines recommended by the participating radiologists were a baseline screening mammogram between the ages of35 and 40 years, a screening mammogram every 1-2 years for ages and a screening mammogram annually at ages 50 and older. The mammography databases are submitted electronically to the Mammography Tumor Registry for linkage with cancer data obtained from the Cancer Surveillance System and the Washington State Cancer Registry. The Cancer Surveillance System is a population-based tumor registry established in 1974 that is designed as one of 10 national sites that constitute the National Cancer Institute s Surveillance, Epidemiology, and End Results cancer registry. The registry documents cancers occurring in the 13 northwestern Washington State counties that surround Puget Sound and is considered to have close to 100% coverage of primary invasive and in situ breast cancers diagnosed within its region. The Washington State Cancer Registry is a populationbased registry established by the Washington State Department ofhealth in It provides the Mammography Tumor Registry with incident cases of breast cancer identified in all areas of Washington State not covered by the Cancer Surveillance Systern. Incident cancer cases are identified by the Cancer Surveillance System and the Washington State Cancer Registry through hospital records, outpatient surgery clinics, pathologists, oncologists, radiotherapists. and state death certificates. Linkage of rnammography information with a tumor registry is necessary to identify cancers occurring in a population of women participating in screening for the purpose of identifying true-positive, true-negative, false-positive, and false-negative mammograms. as recommended by Sickles [10] and Linveret al. [Ill. Our study sample included asymptornatic women 35 years old and older who had a screening mammogram associated with a breast density rating and an assessment code. Because women with a history of breast cancer (n = 1004) in our sample routinely received three-view mammograms for screening purposes, they were excluded from our study. Although some women in the database had more than one mammogram, only one mammogram per woman was included in the study. If a woman s initial screening mammogram did not have a density rating, the subsequent mammogram with a density rating was used. Three hundred fifty-one women diagnosed with breast cancer within 12 months of the index mammogram were excluded from calculations of falsepositive examinations. Finally, to ensure that our study sample fairly reflected the diversity of radiologists contributing to the Mammography Tumor Registry, mammograms were sampled so that no single radiologist represented more than 10% of the total sample. Thus. a random sampling of mammograms was taken from 73,247 mammograms meeting the inclusion criteria. A total of 46,340 women with a screening mammogram. breast density rating, interpretation. and I 2-month follow-up to document the absence of breast cancer constituted our final study population. Classification of Mammograms Age (yrs) Classification of screening mammograms was based on recommendations by the American College of Radiology s Breast Imaging Reporting and Data Ss-stem [121. Breast density ratings were grouped according to the lexicon: almost entirely fat. scattered 11- broglandular densities, heterogeneously dense. and extremely dense. Assessment codes followed the Breastimaging Reporting ant/data System lexicon: 1 = negative; 2 = benign finding. negative: 3 = probably benign finding; 4 = suspicious abnormality: 5 = highly suggestive of malignancy; and 0 = additional evaluation needed. A screening mammogram with positive findings was defined as one assigned an assessment code of 0, 3, 4, or 5. A screening mammogram with negative findings was defined as one given the assessment code of I or 2. This definition of positive mammograms maximizes the detection of falsepositives, even those leading to low-level follow-up recommendations (i.e., 6-month follow-up). False-positive mammograms were defined as screening mammograms with positive findings occurring in women with no known breast cancer diagnosis within 12 months from the index mammogram. True-negative mammograms were defined as screening mammograms with negative findings in women with no known diagnosis of breast cancer within 12 months of the index mammogram. StatisticalAnalysis The Pearson chi-square test was used to assess the relation between age and density [13]. Trends of false-positive examinations across categories of age and density were assessed by testing the statistical significance of a single trend variable coded as 1-4 for Breast Tissue Density of46,340 Patients Categorized by Age* No.of Patients 3,666 15,220 11,394 8,237 7,823 Predominantly 267(7.3) 1,243 (8.2) 1,550(13.6) 1,623(19.7) 1,602 (20.5) Fat Scattered Fibroglandular Breast Tissue Density 1,488(40.6) 6,816 (44.8) 6,238(54.8) 4,623(56.1) 4,424 (56.6) Heterogeneously Dense 1,508(41.1) 5,859 (38.5) 3,194(28.0) 1,806(21.9) 1,570 (20.1) Extremely 403 (11.0) 1, 302(8.6) 412(3.6) 185(2.3) 227 (2.9) All ages 46,340 6,285 (13.6) 23,589 (50.9) 13,937 (30.1) 2,529 (5.5) Note-Numbers in parentheses are percentages. a Chi-square test for difference in proportions: x2< g#{149} density and 1-5 for age. Density was defined as a categorical variable ofeither fatty, scattered fibroglandular, heterogeneously dense, or extremely dense tissue. Age was also defined as a categorical variable divided into increments (35-39 years old years old years old years old. and 70+ years old). Logistic regression was used to estimate the odds ratios of a false-positive mammogram associated with each category of breast density and age, after adjusting for the other factor as a covariate [14]. In our study, the crude (X1ds ratio estimates the odds orlikelihood that a woman with a given breast density or age will have a false-positive mammogram. The adjusted odds ratio provides the independent effect ofeach factor (i.e., breast density) on the probability of a falsepositive examination while controlling for the effect of the other factor (i.e.. age). Confidence intervals were based on the standard error of the coefficient and normal Results approximation. The distribution of the population by breast density and age is shown in Table 1. A significant association between parenchymal pattern and age was evident. Increasing age was associated with decreasing breast density, with the frequency of heterogeneously or extremely dense breast tissue being 52%, 47%, 32%, 24%. and 23% in the age groups 35-39, 40-49, 50-59, 60-69, and older than 70 years, respectively (p <.001). In our sample of 46,340 mammograms, 4091 mammograms had false-positive findings, yielding a false-positive rate of 8.8%, or an overall specificity of 91.2%. True-negafives, false-positives, and percentage of falsepositive screening mammograms by age are shown in Table 2 and Figure 1. Before controlling for the effect of breast density, the falsepositive rate of screening mammography was marginally affected by age (p for trend =.058). However, after controlling for density, age did not have a significant effect on the overall risk ofhaving a false-positive mammogram Dense 1652 AJR:173, December 1999

3 Effect of Age and Breast Density on Mammograms ssociation ofage with Percentage of Screening Mammograms Having Fals&Positive Findings True-Negative False-Positive Odds Ratios (Confidence Intervals) Age (yrs) No. of Mammograms (n= 42,249) (Specificity)a No. of Mammograms (n= 4,091) (1 Specificity) crudec Adjusted C,d ,399 13,828 10,331 1,505 7, SO-S Age (years) f:w;::,iii:4i%xssx. Fatty Scattered Heterogeneously Extreme y flbroglandular dense dens 6ssiae Fig. 1.-Bar graph shows percentage of false-positive mammograms for each age Fig. 2.-Bar graph shows percentage offalse-positive mammograms for each breast category. False-positive percentage was not significantly affected by age. density category. Women with dense breast tissue were significantly more likely to have false-positive mammograms than women with less-dense breast tissue. atrue.negative percentage = specificity = true-negatives I (true-negatives + false-positives(. bfalsepositive percentage = 1 - specificity = false-positives I (true-negatives + false-positivesl. c Odds of mammograms with false-positive findings in women of that group versus reference group. dodds ratio adjusted for age; evaluates influence of age on relationship of density and specificity ,392 1, (0.66, 0.88) 0.98 (0.90, 1.06) 0.95 (0.86, 1.05) 0.86 (0.78, 0.95) p for trends atruenegative percentage = specificity = true-negatives I true-negatives + false-positivesl. bfalsepositive percentage = 1 - specificity = false-positives I (true-negatives + false-positivesl. C Odds of mammograms with false-positive findings in women of that group versus reference group. dodds ratio adjusted for density; evaluates influence of density on relationship of age and specificity. #{149}WAssociadon of Breast Density with Percentage of Screening Mammograms Having False.Positive Findings Fatty Scattered Tissue Heterogeneously Extremely Density fibroglandular dense dense No.ofMammograms (n = 42,249) 5,879 21,513 12,591 2,266 Breast Density 0.73 (0.63, 0.84) 0.94 (0.86, 1.02) 0.98 (0.88, 1.08) 0.89 (0.80, 0.99) True-Negative False-Positive Odds Ratios (Confidence Intervals) % (Specificity)a No.ofMammograms (n = 4,091) 406 2,076 1, (1 crudec Adjustedcd (1.25, 1.56) 1.55 (1.38, 1.74) 1.68 (1.43, 1.98) pfortrend <.001 <.001 (p =.270). In fact. after controlling for density. the youngest age group (35-39) had a significantly lower risk of a false-positive mammogram than the age group. The association of breast density with a false-positive mammogram is shown in Table 3 and Figure 2. The trend of increasing risk of a false-positive mammogram with increasing breast density was highly significant. with women with heterogeneously or extremely dense breast tissue being more than 1.5 times more likely to have a false-positive mammogram than women with fatty breast tissue (p <.001 ). This effect persisted after controlling for age (p <.001). Furthermore. within each age group, density was significantly (p <.05) associated with an increased percentage of falsepositive mammograms. Discussion 1.40 (1.25, 1.56) 1.57 (1.40, 1.76) 1.73 (1.47, 2.05) Our study confirmed the well-documented association of mammographic pattern to age. A higher proportion of young women than older women have dense breast tissue [ I 5-19J. As Kopans [201 has described. breast density does not abruptly change at age 50. Rather, breast density tends to decrease with increasing age. AJR:173, December

4 Lehman et al. Kerlikowske et al. [18] reported that the sity and false-positive mammograms. We did performance of mammography by patient sensitivity of mammography decreases with increasing breast density in women older than not have consistent data from the facilities on estrogen replacement therapy or on family his- age should take into account the influence of breast density. Breast density, not age, may 50 years. Those researchers did not find a sig- tory of breast cancer. Analysis of the impact of be the more important predictor of mam- nificant association between breast density and these factors on the relationship of age and mography performance. Further studies to sensitivity of mammography in women breast density to risk of false-positive mammo- elucidate factors influencing specificity in years old. To our knowledge, ours is the first study to explore the effect of breast density on the rate of false-positive screening mammograms. We observed a higher percentage of false-positive mammograms in women with dense breasts, regardless of age. The rate of false-positive mammograms has been reported to be higher at ages in some series, but not in others [2-5, 21]. These studies did not consider the possible confounding effect of breast density on the rate of false-positive mammograms. We found only a marginal effect of age on the rate of false-positive mammograms, and no effect after controlling for breast density. Thus, the association of younger age with increased risk of a false-positive mammogram is explained by the large effect of breast density on the risk of false-positivity. In addition, the youngest age group (35-39) had the lowest false-positive rate. This age category included relatively few mammograms, and many women in this age group do not get mammograms. Thus, this finding may not be representative of all women years old. Laya et al. [22] reported on the effect of estrogen replacement therapy on the specificity of screening mammography. In their study, the specificity of mammography in women receiving hormone replacement therapy was significantly lower than in women not currently undergoing estrogen replacement. They suggested that this association might be caused in past by increased radiographic breast density associated with hormone use. Our data support their hypothesis by showing that increased breast density is associated with increased false-positive mammography (and thus decreased specificity). One limitation of this study is the variability (measurement error) in practice patterns, breast density, and assessment rating caused by the inelusion of multiple radiologists. This variability would have attenuated our findings. However, an advantage of using a community-based, multiple radiologist setting is its generalizability, because our study was conducted in community-based mammography practices. This was possible because we were able to link screening data with a tumor registry. Mother limitation of our study is the absence of informarion regarding potentially important factors affecting the relationship between breast den- grams is important in future studies. A third limitation is that we were not able to identify cases in which radiologists may have based their assessments on additional views taken at the same time as the four-view screening mammogram. Because this would serve to underestimate false-positive rates, exclusion of these cases would have strengthened our results. One practical application of our findings is to educate patients and referring physicians about the risks and benefits of screening mammography. A 40-year-old woman with fatty breast tissue should not be told she is at higher risk of having a mammogram with false-posilive findings than a 65-year-old woman with heterogeneously dense breast tissue. In fact, our data suggest that women in their 40s have no increased risk of a false-positive mammogram compared with women in their SOs. More importantly, after controlling for breast density, women in their 40s are no more likely to have a false-positive mammogram than women older than SO years. Mother application of our findings may be to direct women to have mammography at a time when breast density is decreased. In particular, having mammography during the follicular phase of the menstrual cycle [23] or short-term cessation of hormone replacement therapy before mammography [24] may prove to be useful strategies to reduce breast density and hence reduce the incidence of false-positive mammograms. Harvey et al. [24] describe results of discontinuing hormone replacement therapy for 2 weeks before mammography in a select group of patients. Their purpose was to avoid unnecessary biopsies of developing densities or new or enlarging circumscribed masses in women receiving hormone replacement therapy. Their results may have implications for general screening guidelines of select women undergoing hormone replacement. Finally, development of other techniques (e.g., digital mammography) to improve conventional mammography may contribute not only to improving sensitivity but also to improving specificity in women with dense breast tissue. We conclude that mammographically dense breast tissue is associated with an increased percentage of false-positive mammograms and that age is not a significant independent risk factor. Studies assessing the women with dense breast tissue may prove beneficial toward enhancing screening strategies for these women. Acknowledgments their We thank the participating radiologists for References contributions. I. Breast cancer screening for women ages 40-49: NIH consensus statement. Kensington, MD: National Institutes ofhealth, 1997 Jan 21-23;15: MillerAB, Baines Ci, To T, Wall C. Canadian national breast screening study. 2. Breast cancer detection and death rates among women aged 50 to 59 years. Can MedAssocf 1992:147: Miller AB, Baines CJ, To T, Wall C. Canadian national breast screening study. 1. Breast cancer detection and death rates among women aged 40 to 49 years. Can MedAssoc J 1992;147: Baker LH. Breast cancer detection demonstration project: five-year summary report. CA Cancer J Clin 1982:32: Elmore JG, Barton MB, Moceri VM, Polk 5, Arena PJ. Fletcher SW. Ten-year risk of false positive screening mammograms and clinical breast examinations. NEnglJMed 1998:338: Fajardo LL. Hillman BJ. Frey C. Correlation between breast parenchymal patterns and mammographers certainty ofdiagnosis. Ins estradiol 1988:23: Eddy DM. Hasselblad V, McGivney W. Hendee W. The value of mammography screening in women underage SOyears.JAMA 1988:259: Gram IT, Lund E, Slenker SE. Quality of life following a false positive mammogram. Br J Cancer 1990:62: Lerman C, Track B, Rimer BK, Boyce A, Jepson C. Engstrom PF. Psychological and behavioral implications of abnormal mammograms. Ann Intern Med 1991:1 14: Sickles EA. Quality assurance: how to audit your own mammography practice. Radio! Clin North Am 1992:30: I 1. Linver MN, Osuch JR. Brenner Ri, Smith RA. The mammography audit: a primer for the Mammography Quality Standards Act (MQSA). AJR 1995:165: Kopans DB, D Orsi CJ, Adler DD, et al. Breast imaging reporting and data system. Reston, VA: American College of Radiology I 3. Fisher LD, Van Belle G. Biostatistics: a methodologv ftr the health sciences. New York: Wiley, 1993: Breslow NE. Day NE. Statistical methods in cancer research. Vol. 1 : The analysis of case-control studies. Lyon, France: International Agency for 1654 AJR:173, December 1999

5 Effect of Age and Breast Density on Mammograms Research on Cancer Scientific Publications, 1980: Ciauo S. Zappa M. A prospective study of the value of mammographic patterns as indicators of breast cancer risk in a screening experience. Eur J Radiol 1993:17: Rook D, Gilhome RW. Harman J. Gravelle Webster DJ. Changes in Wolfe mammographic pattems with aging. Br) Radio! 1987:60: Wolfe in. Breast parenchymal pattems and their changes with age. Radiology 1976:121: Kerlikowske K. Grady D, Barclay J. Sickles EA, Ernster V. Effect of age, breast density, and family history on the sensitivity of first screening mammography. JAMA 1996:276: van Gils CH, Otten JD, Verbeek AL, Hendricks JH. Short communication: breast parenchymal patterns and their changes with age. Br J Radio! 1995:68: Kopans DB. Conventional wisdom: observation, experience. anecdote, and science in breast imaging.air 1994:162: Kerlikowske K, Grady D. Barclay J, Sickles EA, Ernster V. Likelihood ratios for modem screening mammography: risk of breast cancer based on age and mammographic interpretation. JAMA 1996:276: Laya MB, Larson EB, Taplin SH, White E. Effect ofestrogen replacement therapy on the specificity and sensitivity of screening mammography. J Nail Cancer Inst 1996:88: White E, Velentgas P. Mandelson MT. et al. Vanation in mammographic breast density by time in menstrual cycle among women aged years. J Nail Cancer Inst 1998:90: Harvey JA. Pinkerton JV. Herman CR. Short-term cessation of hormone replacement therapy and improvement of mammographic specificity. J Nail Cancer Inst 1997:89: AJR:173, December