Interobserver variability and positive predictive value for ultrasonographic BI-RADS categories requiring...

Area of Science:

• Radiology and Imaging
• Oncology
• Pathohistology

Background:

• Accurate assessment of breast lesions using Breast Imaging Reporting and Data System (BI-RADS) categories is crucial for patient management.
• BI-RADS categories 4A, 4B, 4C, and 5 require pathological evaluation, making interobserver agreement and positive predictive value (PPV) critical metrics.

Purpose of the Study:

• To analyze interobserver variability in the assessment of BI-RADS descriptors.
• To determine the positive predictive value (PPV) for BI-RADS categories 4A, 4B, 4C, and 5.
• To evaluate the clinical applicability of BI-RADS subcategories.

Main Methods:

• A retrospective study of 30 ultrasonographic reports with pathohistological verification was conducted.
• Ten observers (seven gynecologists, three radiologists) independently assessed each report using the BI-RADS atlas (4th edition).
• Interobserver variability was quantified using the kappa (k) coefficient.

Main Results:

• Conformity was highest for orientation (k=0.79), boundary (k=0.71), and shape (k=0.65).
• Moderate agreement was found for posterior features (k=0.54) and margins (k=0.41), with poor agreement for echogenicity (k=0.38).
• Overall interobserver agreement for BI-RADS categories 4A-5 was substantial (k=0.51), with higher agreement for category 5 (k=0.50) compared to 4C (k=0.37), 4B (k=0.32), and 4A (k=0.29).

Conclusions:

• Interobserver agreement for BI-RADS descriptors and final categories 4A-5 is generally good.
• The positive predictive value (PPV) supports the clinical justification and applicability of dividing lesions into categories 4 and 5, as well as subcategories 4A, 4B, and 4C.