Observer variation in the interpretation of xeromammograms

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Screening for breast cancer using xeromammography is crucial for early detection.
• Multicenter randomized controlled trials (RCTs) are essential for evaluating screening efficacy.
• Observer variability in image interpretation can impact trial outcomes.

Purpose of the Study:

• To assess inter-observer variability in the interpretation of xeromammograms among radiologists participating in a Canadian breast cancer screening RCT.
• To identify specific areas of diagnostic disagreement in xeromammogram readings.

Main Methods:

• Radiologists reviewed 100 xeromammograms with known outcomes (10 cancers, 40 benign abnormalities, 50 normal films).
• Data on diagnostic categories (suspicion of cancer, cancer, benign abnormality, normal) and recommended procedures (biopsy, aspiration) were collected.
• Statistical analysis focused on agreement levels across different diagnostic categories.

Main Results:

• Wide variation observed in the frequency of "suspicion of cancer" or "cancer" classifications (10-55%).
• Recommendations for biopsy or aspiration ranged from 21% to 53%.
• Highest agreement was for cancer diagnosis, lowest for benign abnormalities, and intermediate for normal findings.

Conclusions:

• Significant observer variation exists in xeromammogram interpretation, necessitating methods to enhance consistency.
• Maintaining diagnostic sensitivity and validity while reducing variability is critical for screening programs.
• Robust quality control strategies are vital for multicenter breast cancer screening trials.