Assessment of Observer Variability in Determining Mammographic Breast Density as Per the American College of Radiology Birads 2013 Lexicon at Kenyatta National Hospital

Abstract

Background Breast density is a well-known and proven independent risk factor for breast cancer and can significantly affect the sensitivity of screening mammograms. Mammographic breast density is thus great importance in the evaluation of a mammogram because increased breast density is known to increase the risk of individual breast cancer (1) Analysis of breast density is done through the ACR- Birads lexicon 2013, which categorizes breast density into four categories A-D A- Almost entirely fatty, B- scattered areas of fibroglandular density, C- heterogeneously dense, which may obscure small masses and D- extremely dense, which lowers the sensitivity of mammography. Due to the subjective nature of the visual assessment during categorization, differences may arise when assessing the breast density between two observers. It is important to assess the degree of agreement or variability as this affects the reproducibility of the report. Aim The study aimed to evaluate and record the degree of inter and intra observer variability when assessing the mammographic breast density using the Breast Imaging Reporting and Data System (BIRADS), as defined by American College of Radiology. Methodology Cross-sectional study was carried out at Kenyatta National Hospital, mammography department of Radiology over a period of one year. Four radiologists were required to review mammograms that were of diagnostic value and classify the breast density based on the ACR classification 5th edition. After an interval of one month, they were asked to review the same mammograms again. Mammographic density category for each mammogram by each radiologist was recorded. Statistical Analysis The intra-observer variability was calculated for each radiologist and reported as weighted kappa values and 95% confidence intervals. Fleiss-Cohen Kappa coefficients and their corresponding 95% confidence intervals were used to calculate the inter-observer variability. The levels of agreement were assigned a kappa value between 0.0-1.0 where 0.0 was considered as no agreement and 1.0 was considered perfect agreement. Results x There was moderate overall inter-observer variability both at baseline and after the one-month review with kappa values being 0.49(95% CI 0.43,0.59) and 0.43 ( 95% CI 0.38,0.61)) respectively when using the ACR BI-RADS 2013 lexicon to assess breast density. The intra-observer variability was substantial with kappa values ranging from 0.61(95% CI 0.48- 0.72) to 0.77 (95% CI 0.67-0.87). Conclusion There was a moderate level of inter- and intra- observer variability demonstrated when assessing mammographic breast density but a substantial level of individual intra-observer variability.