[A multicenter study of diagnostic criteria in ultrasound mammography. Statistical pitfalls and ways out of the data jungle].

This paper demonstrates the statistical processing of data from a multicentric study. Seventy-seven patients with a suspicious, solid breast mass were included in a two-center study, using clinical breast examination, mammography, and ultrasound. The assessment was formalized using a uniform evaluation sheet. The prospective results of the examinations were compared with the histologic report. The degree of interdependence between an examiner's diagnosis and diagnostic criteria used was assessed using the kappa statistics. We found that despite standardized assessment, the criteria used for differential diagnosis were at least partially different. The criteria valid for diagnosis at both centers were whether a mass fitted well into the surrounding tissue or led to destruction or architectural distortion, whether it was fixed to the surrounding tissues, as assessed by sonopalpation maneuvers, and whether it borders were sharp or unsharp.