Women s Imaging Original Research

Transcription

1 Women s Imaging Original Research Irshad et al. Women s Imaging Original Research Abid Irshad 1 Rebecca Leddy 1 Susan Ackerman 1 Abbie Cluver 1 Dag Pavic 1 Ahad Abid 2 Madelene C. Lewis 1 Irshad A, Leddy R, Ackerman S, et al. Keywords: BI-RADS guidelines, breast density, density assessment, observer agreement, reader agreement DOI: /AJR Received April 5, 2016; accepted after revision May 25, Based on a presentation at the ARRS 2016 Annual Meeting, Los Angeles, CA. 1 Department of Radiology and Radiological Sciences, Medical University of South Carolina, 169 Ashley Ave, Charleston, SC Address correspondence to A. Irshad (irshada@musc.edu). 2 University of South Carolina School of Medicine, Columbia, SC. This article is available for credit. AJR 2016; 207: X/16/ American Roentgen Ray Society Effects of Changes in BI-RADS Density Assessment Guidelines (Fourth Versus Fifth Edition) on Breast Density Assessment: Intra- and Interreader Agreements and Density Distribution OBJECTIVE. The objective of our study was to determine intra- and interreader agreements for density assessment using the fifth edition of the BI-RADS guidelines and to compare with those for density assessment using the fourth edition of the BI-RADS guidelines. MATERIALS AND METHODS. Five radiologists assessed breast density four times in 104 mammographic examinations: twice using the fourth edition of the BI-RADS guidelines and twice using the fifth edition. The intra- and interreader agreements for density assessment based on each guideline were determined and compared. The density distribution pattern under each of the four BI-RADS density categories using each guideline was also noted and compared. RESULTS. The intrareader agreement for density assessment using the fifth-edition criteria was lower than that using the fourth-edition criteria (p = ). The overall intrareader agreement (weighted kappa) using the old criteria was 0.84 (95% CI, ), and the individual intrareader agreement values in five readers ranged from 0.78 (95% CI, ) to 0.92 (95% CI, ). The overall intrareader agreement using the new BI-RADS criteria was 0.77 (95% CI, ), and the individual intrareader agreement values in five readers ranged from 0.74 (95% CI, ) to 0.99 (95% CI, ). The interreader agreement values obtained using the fifth-edition criteria were also lower than those obtained using the fourth-edition criteria (p = 0.006). The overall interreader agreement using the old BI-RADS criteria was 0.65 (95% CI, ), whereas the overall interreader agreement using the new BI-RADS criteria was 0.57 (95% CI, ). Overall a higher number of dense assessments were given when the fifth-edition guidelines were used (p < ). CONCLUSION. Compared with the intra- and interreader agreements obtained using the fourth edition of the BI-RADS guidelines, the intra- and interreader agreements were lower using the fifth-edition guidelines. An increased number of dense assessments were given when the fifth-edition guidelines were used. M ammographic breast density has been a topic of interest for several decades. Increased density is known to limit the sensitivity of mammography for cancer detection because of the masking effect of dense breast tissue [1 4]. Additionally, several studies have shown that increased breast density is associated with an increased risk (up to sixfold) of developing breast cancer [5 13]. Breast density became a particularly hot topic when community-led efforts resulted in the first breast-density legislation in the State of Connecticut in This legislation mandated that patients should be informed of their breast density after their annual mammographic examinations. Additionally, it was legislated that patients with dense breasts should be provided other options as supplemental methods of screening. Since then, a number of states have adopted similar laws in which mammography providers are mandated to inform patients of their breast density. Although these laws provide patients important information about their breast density, there are some concerns in the radiology community about the implications of these laws [14, 15]. The American College of Radiology (ACR) raised some concerns in a position statement issued in 2012 [14]. Because density assessment is subjective and is not reliably reproducible, there is a concern that inconsistent breast density assessments reported to patients from 1 year to the next may result in patient confusion. Additionally, inconsistent insurance coverage for supplemen AJR:207, December 2016

2 tal screening may lead to a health care disparity. Nevertheless, accurate, consistent, and reproducible classification of breast density is important not only for the stratification of patients into various risk prediction models that use breast density as a risk factor [16, 17] but also for the identification of patients with dense breasts who would possibly benefit from supplemental screening methods. Conventionally, mammographic breast density has been visually assessed and categorized by radiologists into one of four density categories according to the ACR s BI-RADS density assessment criteria [18]. Although quantitative technologies are also available for measuring breast density, these technologies are not widely used and visual assessment is currently the most commonly used method for density assessment. Before the fifth edition of the BI-RADS guidelines [18], which was published in 2013, breast density was assessed and categorized using the fourth edition of the BI-RADS guidelines [19], which was published in According to the density assessment criteria provided in the fourth-edition guidelines, breast density was categorized into one of four categories using the estimated percentage of fibroglandular tissue seen on mammography compared with the total size of the breast. The categories were numbered from 1 to 4 from least dense to most dense as follows: 1, almost entirely fat (< 25% glandular density); 2, scattered fibroglandular densities (25 50% glandular density); 3, heterogeneously dense (51 75% glandular density); and 4, extremely dense (> 75% glandular density). The heterogeneously dense and extremely dense categories (i.e., breasts with > 50% glandular density) are generally labeled as dense breasts, whereas the first two categories (breasts with 50% glandular density) are considered nondense. In 2013 with the publication of the new BI-RADS guidelines (fifth edition), the guidelines for density assessment were significantly changed. Although similar density categories are used in the fifth edition (categories A, B, C, and D), the percentage system was removed from the density assessment methods, and more emphasis was given to the masking effects of the dense tissues. According to the new guidelines, a breast could still be classified as dense even if it is < 50% glandular but the radiologist is concerned about an area of dense tissue that could potentially be masking an underlying cancer. The new guidelines provide more freedom to the individual readers for classifying breast density in a more meaningful way. However, by removing the quantifying percentage value system, one may be concerned about a possible increase in interreader variability in density assessment. Greater variability in density assessment can potentially affect the breast density information in the letters that are sent to patients each year after their annual mammographic examinations. Several previous studies have shown a wide range of intra- and interreader agreements in density assessment using the fourth-edition BI-RADS criteria [20 29], with agreements varying from substantial [21] to slight [22]. However, these agreements have not yet been thoroughly evaluated using the fifthedition BI-RADS criteria. To our knowledge, only one recently published study [30] has evaluated reader agreements using the new (fifth-edition) BI-RADS criteria; however, the authors of that study determined the agreements using the new BI-RADS criteria without a controlled comparison with the old (fourth-edition) BI-RADS criteria. In our study, we determined and compared intra- and interreader agreements within the same group of radiologists when using both the new and old density assessment criteria. We also evaluated differences in the density assessment patterns when the old and new BI-RADS criteria were used. Materials and Methods Patients Approval from the institutional review board was obtained for this study, and informed consent was waived because of the retrospective nature of this study. This study was HIPAA-compliant. Only screening mammographic examinations were selected for this study. Digital mammograms of 104 consecutive women were selected from our screening mammography database for the purpose of density assessment. The mean patient age was 47 years (range, years); 46 women were premenopausal, and 58 were postmenopausal. Imaging All digital mammograms were obtained on fullfield digital mammography units (Lorad Selenia, Hologic). The digital DICOM images were stored on a PACS (Impax, Agfa Healthcare) and were evaluated on dual-screen 5-megapixel LCD monitors. The breast density in each case was assessed from the craniocaudal and mediolateral oblique images of both breasts. If the density of one breast was different from that of the other breast, the breast with the higher density was categorized. Five fellowship-trained radiologists (breast imagers with 3 17 years of experience) who were blinded to the originally reported density independently reviewed and assessed the breast density in each case. All radiologists were originally trained under the fourth-edition BI-RADS guidelines and had clinical experience of practicing using the fourth-edition guidelines. Each radiologist evaluated the breast density of 104 mammographic examinations four times (416 assessments): twice using the fourth-edition BI-RADS criteria and twice using the fifth-edition criteria. Thus, a total of 2080 readings were performed by the five radiologists. The cases were first evaluated using the fourthedition criteria. The first two readings (readings 1 and 2) using the fourth-edition BI-RADS criteria were separated by a period of 4 weeks to reduce the potential effect of the radiologists memory of cases. Density was classified in each case into one of the four categories on the basis of density percentage values (almost fatty, < 25% glandular density; scattered densities, 25 50% glandular density; heterogeneously dense, 51 75% glandular density; and extremely dense, > 75% glandular density). Then, after a period of more than 6 months during which the radiologists had been using the fifth edition of the BI-RADS guidelines in clinical practice, the radiologists reassessed density twice (readings 3 and 4) using the fifth-edition BI-RADS density assessment criteria. The last two readings were also separated by a period of 4 weeks. For the last two readings, no percentage values were used to assess breast density, and the radiologists subjectively assessed and classified each breast into a density category using the new guidelines. They classified a breast as dense if they perceived that an area of glandular tissue might be dense enough to obscure an underlying cancer, as indicated in the fifth-edition guidelines. To determine the agreements under the fourthedition guidelines, we calculated the intrareader agreements between readings 1 and 2 for each reader and the interreader agreements across reading 1 by all readers. The intrareader agreements under the fifth-edition guidelines were calculated between readings 3 and 4 by each reader and across reading 3 by all readers. The agreements calculated under each guideline were compared. Additionally, the density distribution pattern under each of the four density categories was noted and compared. The data were also analyzed for differences when breasts were divided into nondense (categories 1 and 2 according to fourth-edition criteria and A and B according to fifth-edition criteria) and dense (categories 3 and 4 according to fourth-edition criteria and C and D according to fifth-edition criteria) groups. AJR:207, December

3 Irshad et al. TABLE 1: Interreader Agreement for Breast Density Assessment Using Fourth-Edition BI-RADS Guidelines Fleiss-Cohen (Quadratic) Weighted κ (95% CI) Statistical Analysis Because breast density can be rated on an ordinal scale from 1 to 4, the agreement between readers is reported as the Fleiss-Cohen weighted kappa coefficient and associated 95% CI. Differences in this proportion were tested using a chi-square test for equality of proportions. Intra- and interreader agreements were calculated across subjects for both the fourth- and fifth-edition BI-RADS ratings. Rating levels ranged from low density to high density across four ordered categories (1 4 or A D); thus, a weighted kappa statistic [31, 32] and CI were calculated so that the cells with greater distance from agreement are weighted at a lower level than those closer to agreement. Both within and across fourth- and fifth-edition BI-RADS ratings were compared to examine reader reliability on the same scale and the impact of the fifth-edition BI-RADS guidelines on density ratings for the same images. Kappa values were compared across readers using Z scores calculated using the difference in kappa values and the associated asymptotic standard errors. The subject level and overall rating scale distribution were compared between the fourth and fifth edition of the BI-RADS guidelines using a multinomial logistic regression model that accounted for clustering on each examination. Similarly, changes from nondense (A and B) to dense (C and D) categories from the fourth edition to the fifth edition of the BI-RADS guidelines were assessed using a logistic regression model that also accounted for clustering on each examination. All analyses were performed ( ) 0.82 ( ) 0.86 ( ) 0.77 ( ) ( ) 0.71 ( ) 0.67 ( ) ( ) 0.87 ( ) ( ) 5 Note Dash ( ) indicates not applicable. Overall Fleiss-Cohen (quadratic) weighted κ = 0.65 (standard error = 0.02). using statistics software (SAS, version 9.4, SAS Institute), and no corrections for multiple comparisons were performed. Results Intrareader Agreement When breast density was assigned using the fourth-edition BI-RADS criteria (reading 1 vs 2), the overall intrareader agreement (weighted kappa) was 0.84 (95% CI, ), and the individual intrareader agreements in five readers ranged from 0.78 (95% CI, ) to 0.92 (95% CI, ) (Fig. 1). The overall intrareader agreement using the fifth-edition BI-RADS criteria (reading 3 vs 4) was 0.77 (95% CI, ), and the individual intrareader agreements in five readers ranged from 0.74 (95% CI, ) to 0.99 (95% CI, ). The difference between the intrareader agreements obtained using the fourthedition criteria and the fifth-edition criteria was statistically significant (p = ). Interreader Agreement The overall interreader agreement (weighted kappa) using the fourth-edition BI-RADS criteria was 0.65 (95% CI, ), whereas the overall interreader agreement using the fifth-edition BI-RADS criteria was 0.57 (95% CI, ) (Tables 1 and 2). The difference between the interreader agreements obtained using the old and new BI-RADS criteria was statistically significant (p = 0.006). Density Assignment Distribution The differences in the overall density assignment patterns using the fifth-edition BI-RADS criteria compared with the fourthedition criteria were analyzed (Table 3). Compared with the density assignments using the fourth-edition criteria, a significant change in the density assignment pattern was noted when the fifth-edition criteria were used. Overall, a denser assessment was given more frequently with the fifth-edition criteria (p < ). When the total assessments with the new criteria (n = 1040) were compared with the total assessments with the old criteria (n = 1040), the density category remained unchanged in 750 (72.1%) cases, a higher density category was given in 245 (23.6%) cases, and a lower density category was given in 45 (4.3%) cases. We analyzed the cases in which a higher density category was given using the fifth-edition criteria and found that the assessment changed from category 1 to 2 in 24 (2.3%) cases, from 1 to 3 in one (0.1%) case, from 1 to 4 in 0 (0%) cases, from 2 to 3 in 122 (11.7%) cases, from 2 to 4 in eight (0.8%) cases, and from 3 to 4 in 90 (8.7%) cases. When breast density was divided into nondense (categories 1 and 2 or A and B) and dense (categories 3 and 4 or C and D) groups, assignment shifts from nondense to dense were noted in 131 (12.6%) cases and from dense to nondense in 29 (2.8%) cases. Overall, 47.5% of the cases were assigned TABLE 2: Interreader Agreement for Breast Density Assessment Using Fifth-Edition BI-RADS Guidelines Fleiss-Cohen (Quadratic) Weighted κ (95% CI) ( ) 0.77 ( ) 0.74 ( ) 0.72 ( ) ( ) 0.75 ( ) 0.79 ( ) ( ) 0.85 ( ) ( ) 5 Note Dash ( ) indicates not applicable. Overall Fleiss-Cohen (quadratic) weighted κ = 0.57 (standard error = 0.02) AJR:207, December 2016

4 TABLE 3: Distribution of Density Categories Assigned Using Fourth- and Fifth-Edition BI-RADS Guidelines Stratified by and Overall Density Category (No. of Cases) Fourth-Edition (Old) BI-RADS to the dense group using the fourth-edition criteria, whereas 57.3% of the cases were assigned to the dense group using the fifth-edition criteria (Table 4). Discussion For more than a decade, radiologists have used the fourth-edition BI-RADS density assessment criteria that categorize density by estimating the percentage of glandular density in relation to the whole breast on a mammogram. The fourth-edition criteria provided radiologists with a quantifying guideline to help them categorize breast density in a somewhat uniform and consistent way. However, the visual estimation of breast density is not a perfect science. In addition to individual perceptual differences, a number of technical and positional factors may affect the perceived density on a mammogram. Previous studies using the old criteria have shown a wide range of interreader agreements ranging from substantial to slight [20 29]. Ooms Fifth-Edition (New) BI-RADS A B C D < < < < Overall < a For change from old to new BI-RADS guidelines. et al. [20] used 57 film-screen mammograms evaluated by four radiologists and reported substantial interreader agreement (κ = ). Similarly, Bernardi et al. [21] also reported substantial interreader agreement (κ = ) when six radiologists read 100 digitized film-screen mammograms. Several other studies have shown moderate interreader agreements using film-screen and digital mammograms [23 29]. However, Redondo et al. [22] reported only slight interreader agreement (κ = 0.37) between 21 radiologists. The fourth-edition BI-RADS density assessment criteria had an intrinsic tendency for the reader to rely more on how the glandular tissue spread across the area of the breast and less on its masking effect and possibility of obscuring an underlying mass. Because increased breast density is known to decrease the sensitivity of mammography for cancer detection due to its masking effect, a density assignment taking into account only the area occupied by the glandular tissue and ignoring focal areas of dense tissue can sometimes either provide a false sense of security or raise a false alarm. The fifthedition density assessment guidelines are geared toward correctly identifying patients who can possibly benefit from supplemental screening methods. In addition to achieving the goal of stratifying patients in a more meaningful way, the fifth-edition guidelines also allow radiologists to express their limitation for detecting cancer on mammograms even if a breast is considered nondense by the fourth-edition criteria. Expressing this limitation may provide a sense of security to the radiologists from a medicolegal perspective in some cases of dense assessment. One might think that removing the estimated percentage value system from the density assessment guidelines would result in a reader s observation becoming even more subjective. This potential increase in subjectivity could lead to a significant drop in both intra- and interreader agreements. However, TABLE 4: Distribution of Density Assignments in Nondense and Dense Categories Using Fourth- and Fifth-Edition BI-RADS Guidelines Stratified by and Overall Fourth-Edition (Old) BI-RADS Density Category (No. of Cases) Fifth-Edition (New) BI-RADS Nondense a Dense b Nondense c Dense d < < < Overall < a Density categories 1 and 2. Density categories 3 and 4. Density categories A and B. Density categories C and D. e For change from old to new BI-RADS guidelines. p e p a AJR:207, December

5 Irshad et al. Intrareader Agreement (κ) Fourth-edition criteria Fifth-edition criteria Fig. 1 Bar graph shows comparison of intrareader agreements (κ) between fourth-edition (old) and fifth-edition (new) BI-RADS density assessment criteria. Proportion of Cases Assigned to Dense Categories Ekpo et al. [30] in their recent study showed substantial interreader (κ = ) and almost-perfect intrareader (κ = ) agreements between five readers when they used the fifth-edition BI-RADS guidelines. Our study showed an overall almost-perfect intraobserver agreement using the fourthedition criteria. Despite a small decrease in intrareader agreement with the fifth-edition criteria (0.07 decrease in kappa value), the agreement still remained substantial. The interreader agreement that was substantial using the old criteria (κ = 0.65) decreased to moderate using the new criteria (κ = 0.57). The decreases in agreements were expected considering the fact that the assessment using the fifth-edition criteria is relatively more subjective and that the fifth-edition criteria had been used in clinical practice for a relatively short time at the time of the study. Our study showed a significant increase in the assignment of a dense category using the fifth-edition BI-RADS criteria (Fig. 2). This finding can also be considered an expected result. Because having > 50% glandular density is no longer a requirement to place a breast in the dense category, one would expect a higher number of dense assignments. The overall dense assignments increased from 47.5% of the study group according to the fourth-edition criteria to 57.3% according to the fifth-edition criteria. A maximum shift (11.7%) was noted from the scattered glandular densities category (B) to the heterogeneously dense category (C). Among the readers, all showed a significant increase in dense assignments with the fifth-edition criteria (p < ) except one reader (reader 2) who did not show a significant change in dense assignments (Fig. 2) between the two guidelines (p = 0.25). This reader in the study who followed a similar density assessment pattern using both criteria despite clearly understanding the differences between the two criteria was considered to be an outlier. These results may suggest that some radiologists may not show any significant change in the way they have been practicing using the fourth-edition guidelines for density assessment. Additionally, one of the five readers (reader 5) showed better intrareader agreement using the fifthedition criteria (Fig. 1). One limitation of the study was its design for readers to focus all their attention on breast density, making density the most important finding on the mammograms, which is not the case in real practice in which density is usually a secondary focus of attention. Furthermore, the visual density assessment in general is a subjective process at its core. Although commercially available volumetric quantitative density assessment methods may not be as subjective as visual assessment, it still needs to be established how density assessment by quantitative software tools will correlate with density assessment using the fifth-edition BI-RADS guidelines. The high agreement seen in our study indicates the capability of radiologists to perform well when attention is paid to breast density. However, this finding may or may not fully translate in routine clinical practice because density is usually not a primary focus of attention. Further retrospective large clinical data analysis will show if there has really been a practical Fourth-edition criteria Fifth-edition criteria Fig. 2 Bar graph shows proportions of cases assigned to dense categories by each reader using fourth-edition BI-RADS (old) and fifth-edition (new) criteria. Dense categories are categories 3 and 4 of fourth-edition criteria and categories C and D of fifth-edition criteria. change in the density assessment pattern in routine clinical practice after the guidelines have been revised. The other limitation is that we had an outlier reader that might have affected some overall results; however, the very high agreement among the four other readers minimizes this effect. Conclusion Substantial intrareader and moderate interreader agreements were observed using the new fifth-edition BI-RADS density assessment guidelines. Agreements were slightly lower compared with almost-perfect intrareader and substantial interreader agreements when the fourth-edition guidelines were used. An approximately 10% overall increase was seen in the assignment of a dense category using the fifth-edition guidelines: The most frequent shift was from the scattered fibroglandular category (B) to the heterogeneously dense category (C). After adoption of the fifth-edition BI-RADS guidelines, one can expect an overall shift toward higher density assessments, which will lead to an overall increase in the patient population who may be interested in supplemental screening. References 1. van Gils CH, Otten JD, Verbeek AL, Hendriks H, Holland R. Effect of mammographic breast density on breast cancer screening performance: a study in Nijmegen, The Netherlands. J Epidemiol Community Health 1998; 52: Buist DS, Porter PL, Lehman C, Taplin SH, White E. Factors contributing to mammography failure in women aged years. J Natl Cancer Inst 2004; 96: AJR:207, December 2016

6 3. Jackson VP, Hendrick RE, Reig SA, Kopans DA. Imaging of the radiographically dense breast. Radiology 1993; 188: Pisano ED, Hendrick RE, Yaffe MJ, et al.; DMIST Investigators Group. Diagnostic accuracy of digital versus film mammography: exploratory analysis of selected population subgroups in DMIST. Radiology 2008; 246: Wolfe JN. Risk for breast cancer development determined by mammographic parenchymal pattern. Cancer 1976; 37: Ciatto 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: Vachon CM, Kuni CC, Anderson K, Anderson CE, Sellers TA. Association of mammographic defined percent breast density with epidemiologic risk factors for breast cancer (United States). Cancer Causes Control 2000; 11: Boyd NF, Byng JW, Jong RA, et al. Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study. J Natl Cancer Inst 1995; 87: Boyd NF, Lockwood GA, Byng JW, Tritchler DL, Yaffe MJ. Mammographic densities and breast cancer risk. Cancer Epidemiol Biomarkers Prev 1998; 7: Pinsky RW, Helvie MA. Mammographic breast density: effect on imaging and breast cancer risk. J Natl Compr Canc Netw 2010; 8: Byrne C, Schairer C, Wolfe J, et al. Mammographic features and breast cancer risk: effects with time, age, and menopause status. J Natl Cancer Inst 1995; 87: Boyd NF, Rommens JM, Vogt K, et al. Mammographic breast density as an intermediate phenotype for breast cancer. Lancet Oncol 2005; 6: Harvey J, Bovbjerg V. Quantitative assessment of mammographic breast density: relationship with breast cancer risk. Radiology 2004; 230: American College of Radiology website. ACR statement on reporting breast density in mammography reports and patient summaries. org/about-us/media-center/position-statements/ Position-Statements-Folder/Statement-on-Reporting- Breast-Density-in-Mammography-Reports-and- Patient-Summaries. Published April 24, Accessed April 1, Price ER, Hargreaves J, Lipson JA, et al. The California Breast Density Information Group: a collaborative response to the issues of breast density, breast cancer risk, and breast density notification legislation. Radiology 2013; 269: Tice JA, Cummings SR, Smith-Bindman R, Ichikawa L, Barlow WE, Kerlikowske K. Using clinical factors and mammographic breast density to estimate breast cancer risk: development and validation of a new predictive model. Ann Intern Med 2008; 148: Vachon CM, van Gils CH, Sellers TA, et al. Mammographic density, breast cancer risk and risk prediction. Breast Cancer Res 2007; 9: D Orsi CJ, Sickles EA, Mendelson EB, et al. ACR BI-RADS Atlas, Breast Imaging Reporting and Data System, 5th ed. Reston, VA: American College of Radiology, D Orsi CJ, Mendelson EB, Ikeda DM, et al. Breast Imaging Reporting and Data System: ACR BI-RADS breast imaging atlas, 4th ed. Reston, VA: American College of Radiology, Ooms EA, Zonderland HM, Eijkemans MJ, et al. Mammography: interobserver variability in breast density assessment. Breast 2007; 16: Bernardi D, Pellegrini M, Di Michele S, et al. Interobserver agreement in breast radiological density attribution according to BI-RADS quantitative classification. Radiol Med 2012; 117: Redondo A, Comas M, Macia F, et al. Inter- and intraradiologist variability in the BI-RADS assessment and breast density categories for screening mammograms. Br J Radiol 2012; 85: Baker JA, Kornguth PJ, Floyd CE Jr. Breast imaging reporting and data system standardized mammography lexicon: observer variability in lesion description. AJR 1996; 166: Berg WA, Campassi C, Lagenberg P, Sexton MJ. Breast Imaging Reporting and Data System: interand intraobserver variability in feature analysis and final assessment. AJR 2000; 174: Ciatto S, Houssami N, Apruzzese A, et al. Categorizing breast mammographic density: intra- and interobserver reproducibility of BI-RADS density categories. Breast 2005; 14: Kerlikowske K, Grady D, Barclay J, et al. Variability and accuracy in mammographic interpretation using the American College of Radiology Breast Imaging Reporting and Data System. J Natl Cancer Inst 1998; 90: Lehman C, Holt S, Peacock S, White E, Urban N. Use of the American College of Radiology BI-RADS guidelines by community radiologists: concordance of assessments and recommendations assigned to screening mammograms. AJR 2002; 179: Harvey JA, Gard CC, Miglioretti D, et al. Reported mammographic density: film-screen versus digital acquisition. Radiology 2013; 266: Keller BM, Nathan DL, Gavenonis SC, et al. variability in breast density estimation from full-field digital mammograms: the effect of image postprocessing on relative and absolute measures. Acad Radiol 2013; 20: Ekpo E, Ujong U, Mello-Thoms C, McEntee M. Assessment of interradiologist agreement regarding mammographic breast density classification using the fifth edition of the BI-RADS Atlas. AJR 2016; 206: Fleiss JL, Cohen J, Everitt BS. Large sample standard errors of kappa and weighted kappa. Psychol Bull 1969; 72: Cohen J. Weighed kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull 1968; 70: FOR YOUR INFORMATION This article is available for CME and Self-Assessment (SA-CME) credit that satisfies Part II requirements for maintenance of certification (MOC). To access the examination for this article, follow the prompts associated with the online version of the article. AJR:207, December