Artikel
Highest accuracy of BI-RADS 4/5 results by combined mammography and breast ultrasound - findings of a breast imaging reference center
Suche in Medline nach
Autoren
Veröffentlicht: | 20. März 2006 |
---|
Gliederung
Text
Background: The Breast Imaging Reporting and Data System (BI-RADS) of the American College of Radiology (ACR) is a diagnostic tool to catagorize the risk for malignancy of breast lesions. The systematic categorisation of morphological findings and malignant risk are thought to minimize inconsistencies of inter- and intra-individual estimations of breast imaging. Aim of our analysis was the evaluation of the accuracy of BI-RADS 4/5 results in a large collective.
Methods: From 1999 to 2003 5636 consecutive patients with BI-RADS 4/5 results were prospectively documented in the central database of the University Breast Center Franken (UBF). All patients underwent biopsy by an appropriate method. Positive predictive values (PPV) were calculated and adjusted for method of biopsy, morphological findings, and diagnostic method (mammography vs. ultrasound).
Results: Patients with a BI-RADS 4 result (n=3966) showed a malignant histology in 15,7% of cases (n=623). A BI-RADS 5 result showed a malignant histology in 86,6% (1447 out of 1670 results). Mammography alone proved 299 (14,4%) carcinomas to be false negative, ultrasound alone 367 (17,7%). Best discrimination of BI-RADS 4 vs. BI-RADS 5 results was seen in the group of patients with solid masses (10% vs. 92,9%) and sonografic guided biopsy (8,9 % vs. 91,6%). Findings with either architectural distortion or calcification resulted in more than 20% malignant histologies in BI-RADS 4 classified diagnoses.
Conclusion: BI-RADS categorisation is helpful to evaluate the accuracy of breast imaging results. The ACR-proposed probability of malignancy could not be achieved by using diagnostic method only. Concerning subgroups of morphological findings the PPV of BI-RADS 4 results was much higher than the recommended threshold. The complementary application of mammography and ultrasound seems to be mandatory. For the prevention of pitfalls special constellations of covariates have to be respected carefully.