A group of researchers affiliated with the National Institutes of Health’s Breast Cancer Surveillance Consortium provided a much-needed update to statistics on the performance of diagnostic mammography in the U.S., publishing the results in Radiology.
The transition from screen-film mammography to digital mammography was hailed as a vast improvement, resulting in an increased detection rate, easier image management and improved workflow. However, the transition happened quickly: 99 percent of facilities in 2012 were using digital mammography, while the previous benchmarks were based on exams conducted from 1996 to 2001, where 99 percent of exams were screen-film mammography.
Associate professor and lead study author Brian Sprague, PhD, of the University of Vermont, recognized the need for a refresh.
“Updated national statistics are needed to evaluate the current performance of diagnostic digital mammography and to provide contemporary benchmarks for facility auditing,” wrote Sprague et al. “Our results do not represent desirable goals or target levels of performance; rather, they describe the range of performance in clinical practice among a very large sample of providers in the United States. Thus, the data can be used by radiologists and facilities to put their performance statistics into context during the process of continuing quality improvement.”
The researchers gathered data on about 400,000 exams conducted between 2007 and 2013, totaling over 200,000 women. When compared to screen-film mammography, digital mammography showed increases in abnormal interpretation and cancer detection rates, but a decrease in positive predictive value. Sprague and colleagues theorize this may be due to improvements in imaging technology that do more to allow visualization of smaller lesions, rather than helping radiologists distinguish benign from malignant abnormalities. In addition, the circumstances of the exam greatly affects those statistics, according to the authors.
Most notably, radiologists were much more likely to find cancer if the exams are prompted by the presence of a lump. Exams performed for short-interval follow-up had much lower cancer detection rates, but this was expected.
“Facilities and radiologists conducting diagnostic outcomes auditing should be cognizant of these differences when evaluating their performance metrics if combining all diagnostic indications,” wrote Sprague et al.
Backing up these claims are the large amount of non-academic hospitals who contributed data, reflecting real-world radiology practices.
“Overall, our results suggest that radiologists in the United States excel at detecting cancers but are often not meeting expert-based goals for reducing false-positive biopsy recommendations,” they wrote. “The substantial degree of variation in performance metrics across radiologists suggests that ongoing quality improvement efforts are needed in diagnostic mammography.”