Quality control of mammography screening in the Veneto Region. Evaluation of four programs at a local health unit level--analysis of the frequency and diagnostic pattern of interval cancers

Service mammography screening has been reported to have suboptimal performance compared to controlled trials. The aim of this study was to evaluate the sensitivity of the mammography screening program in four Local Health Units (ASL) and the possible causes of diagnostic error in cases further surfacing as interval cancers. Interval cancers were identified by cross checking of screened women databases with hospital discharge records reporting breast cancer. Proportional interval cancer incidence (observed interval cancers/expected invasive cancers) was determined by matching the database of women screened during 1999-2002 to the hospital discharge records databases during 1999-2003. The ratio of observed interval cancer rate to underlying incidence was compared to international standards and with literature data. Screening mammograms reported as negative and followed by interval cancers were randomly mixed with true-negative controls, and the resulting set underwent blind review by an external radiologist who applied the conventional criteria recommended for the classification of the type of diagnostic error (occult, minimal signs, screening error). Matching of screening archives with the hospital discharge records databases allowed for the identification of 154 invasive interval cancers compared to 480 expected. The proportional observed/expected interval cancer incidence in the first and second year of the interval was 21% and 46%, respectively (ASL 1 = 14% or 38%, ASL 2 = 19% or 48%, ASL 3 = 30% or 53%, ASL 4 = 25% or 49%). Radiological review included 38 further interval cancer cases, identified after the time limits defined for proportional interval cancer incidence assessment, and could not include 18 interval cancers, not retrieved from ASL 4 archives: overall, 174 interval cancers were reviewed, of which 135 were classified as occult (77.3%) (ASL 1 = 83.3%, ASL 2 = 71.1%, ASL 3 = 78.6%, ASL 4 = 75%), 12 (6.9%) as minimal signs (ASL 1 = 6.6%, ASL 2 = 11.5%, ASL 3 = 2.4%, ASL 4 = 5%), and 27 (15.5%) as screening error (ASL 1 = 8.3%, ASL 2 = 17.3%, ASL 3 = 19.0%, ASL 4 = 25%). Observed proportional interval cancer incidence was lower than commonly reported for service screening programs and currently recommended (< 30% in the first, < 50% in the second year of the interval). The analysis of interval cancer causes showed a screening error rate below the maximum acceptable standard (< 20% of interval cancers should be classified as screening error) in t