Single X-Ray Absorptiometry for Measuring Breast Density Compared with Visual Inspection by Radiologists: Inter- and Intraobserver Variability

Abstract:

PURPOSE: Breast density is one of the strongest risk factors associated with increased breast cancer risk. Yet, there is no uniformly accepted clinical method of determining breast density. We measured the level of agreement in assessing breast density by visual inspection and correlation of visual inspection with an automated method using single x-ray absorptiometry (SXA).MATERIAL AND METHODS: The SXA method measures breast density by comparing extent of opacity present on mammograms to fat and glandular reference standards imaged with the craniocaudal breast view. SXA breast density (BD-SXA) is calculated using mathematical derivations previously described. We applied this automated technique to 286 consecutive film screen mammograms and then digitized the 286 screens to assess BI-RADS breast density. Blinded to BD-SXA results, 2 radiologists assigned BI-RADS density categories (1-fatty; 2-scattered fibroglandular densities; 3-heterogeneously dense; 4-extremely dense) to soft-copy digitized film screen on two separate occasions. The second reading occurred at least 1 month after the first reading. Radiologist A was a fellowship-trained mammographer with 7 years of experience. Radiologist B was a fourth-year radiology resident. A weighted kappa statistic was used to assess the variability in assigning BI-RADS categories between radiologists and within radiologists. Statistical significance was shown using 95% confidence interval (CI). Disagreement was considered to exist if there were any categorical differences between readings. Correlation between BI-RADS breast density category and BD-SXA was determined for each radiologist.RESULTS: Agreement between two radiologists using BI-RADS breast density was substantial (kappa 0.79; 95% CI: 0.72-0.85). Agreement on repeat reading by the same radiologist using BI-RADS was less (kappa 0.67 for Radiologist A [95% CI: 0.59-0.75] and 0.53 [95% CI: 0.45-0.61] for Radiologist B) than agreement between radiologists. Correlation of BI-RADS assessment with BD-SXA (kappa-values of 0.47 [95% CI: 0.73-0.57] and 0.42 [95% CI: 0.33-0.51] for radiologist A and B, respectively) was moderate.CONCLUSION:There is moderate to substantial agreement in assigning BI-RADS breast density on screening mammography examinations. Correlation of BI-RADS and BD-SXA was modest. This suggests that BD-SXA is measuring breast density differently than a qualitative assessment. Future studies will examine whether BD-SXA is more strongly associated with breast cancer risk than BI-RADS breast density.¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬¬_______________________________________________________________________KappaWeighted KappaIntraobserver variabilityJessica vs. Jessica0.530.67Jing vs. Jing0.310.53Interobserver variabilityJessica vs. Jing0.670.79Jessica vs. BD-SXA0.280.47Jing vs. BD-SXA0.220.420.59-0.750.45-0.610.72-0.850.37-0.570.33-0.51

• Tagged
• XML
• BibTex