Result quality of breast cancer detection in QuaMaDi with respect to levels of the diagnostic process chain

Introduction: In Schleswig-Holstein (SH) the Quality assured Mamma Diagnostic programme (QuaMaDi) is offered to women symptomatic or at risk for breast cancer of all ages. The standardized process includes a clinical examination (gynaecologist), independent double-reading of mammograms (radiologists), and expert reading (reference centres) followed by an assessment if needed. Aim of the present analysis was to compare diagnostic parameters with respect to levels of the QuaMaDi process chain. Methods: From 2001–2008, 295327 QuaMaDi processes were documented. This cohort has been linked to the cancer registry SH allowing confirmation of suspicious findings in cohort members. Each process was assigned a final diagnosis (breast cancer occurred within 12 months after the QuaMaDi process: yes/no). To set up a „gold standard“, data from various sources (QuaMaDi, cancer registry, gynaecologists' questionnaire) were combined. Physicians applied BI-RADS categories for ratings (1 normal, 2 benign, 3 probably benign, 4 suspicious, 5 highly suspicious, 6 malignant). Sensitivity, specificity, and predictive values were computed for summarized negative (BI-RADS 1–3) and positive findings (BI-RADS 4–6). Results: Overall tumour rate was 152/1000 QuaMaDi processes. Across the process chain sensitivity increased (gynaecologist: 24.6%, radiologists: 90.1%; expert reading: 97.9%) and specificity decreased (99.8%, 94.7%, 77.9%, respectively). Assessment plus final recommendation had a high sensitivity (97.6%) and balanced specificity (95.1%). This last step showed the highest positive predictive value (84.3%, false positive rate 4.9%). The lowest PPV (20.3%) resulted after first/second mammogram reading. Core biopsy had a PPV of 45.9% (malignant to benign findings 1:1.18). Discussion: Expert reading of mammograms increased sensitivity while specificity decreased compared to first/second readings. The diagnostic accuracy of assessment reached a close to maximum sensitivity/specificity >95%. The false positive rate was relatively high due to our approach of BI-RADS 4 results comprised to positive findings. Nonetheless, a high quality of diagnostic procedures was achieved, particularly in advanced levels of QuaMaDi.