AN ASSESSMENT OF THE INFLUENCE OF CLINICAL BREAST EXAMINATION REPORTSON THE INTERPRETATION OF MAMMOGRAMS IN A BREAST SCREENING-PROGRAM

The population-based Ontario Breast Screening Program (OBSP) provides two-yearly screening by both nurse examiner clinical breast examinatio n (CBE) and two-view mammography to women aged 50 to 69. CBE alone acc ounts for about 5% of cancer detection. The purpose of this study was to determine whether CBE information affects radiologists' interpretat ion of mammography. Interpretation was defined by 1) radiologists' ref erral rates for diagnostic evaluation and 2) radiologists' accuracy in distinguishing cancer from non-cancer on mammograms. Mammograms were obtained from women randomly selected from those screened in the OBSP between 1990 and 1992. Selection was stratified by whether or not the nurse examiner had independently referred for diagnostic evaluation. A dditional women diagnosed with breast cancer were selected to increase the precision of the receiver-operating characteristic (ROC) curve. E ach participating OBSP radiologist read a set of mammograms twice, app roximately three weeks apart. The first reading was based on mammogram s alone. At the second reading, the CBE report was included on the rep orting form. Among 620 women referred by the nurse, radiologist referr al rate increased from 37.7% to 40.8% (p = 0.079) when CBE information was available. Among the 637 not referred by the nurse? radiologist r eferral rate decreased from 29.8% to 25.6% (p = 0.005). There was litt le effect on the sensitivity and specificity of radiologist referral a nd the areas under the two ROC curves (with and without CBE informatio n) were not significantly different (p = 0.571). Although there was so me effect of CBE information on radiologists' pattern of referral, the re was no effect on accuracy of cancer detection.