| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Letter |
Charlotte Radiology Charlotte, NC 28203
WEB—This is a Web exclusive article.
My study [2], however, was designed to compare the effectiveness of three reading methods for detecting breast cancer on screening mammograms, and it does not purport to measure long-term morbidity or mortality. The study is based on commonly used indexes of reading effectiveness: recall rate, sensitivity, positive predictive value as a percentage of recalls that are true-positive, and cancer detection rate. In this respect, my study is similar to many other studies that measured double-reading and computer-aided detection (CAD) benefits. In particular, my study reports the same data categories that were included in the New England Journal of Medicine study by Fenton et al. [3].
My study [2] showed that CAD enhanced the performance of a single reader by increasing sensitivity (90.4% vs 81.4%) with only a modest increase in the recall rate (10.6% vs 10.2%). CAD also compared favorably with double reading by yielding a nonstatistically significant increased sensitivity (90.4% vs 88.0%, p = 0.205) with a significantly decreased recall rate (10.6% vs 11.9%). The same nine experienced mammography readers performed the 231,221 readings in this study. Although lymph node status and tumor size data might enhance the conclusions of the study, I do not agree with Dr. Ellis [1] that the absence of these factors negates the validity of comparing reader performance.
Dr. Ellis [1] stresses the importance of showing data for invasive and noninvasive cancers. My study [2] does report this information. Isolating invasive cancers only, the sensitivity of the three reading methods was 79.7% (first reader), 85.3% (double reading), and 87.2% (single reading with CAD). The resulting false-negative invasive cancers were 83 (first reader), 60 (double reading), and 50 (single reading with CAD). To Dr. Ellis' point that the purpose of screening is to find nonpalpable cancer, it is worth emphasizing that the 60 (double reading) and 50 (CAD reading) false-negative invasive studies became palpable within 1 year of the mammogram reading. Even without the surrogates of mortality that Dr. Ellis suggests, I believe the reduction of false-negative rates (e.g., higher sensitivity) and the other differences in reading data yield valuable conclusions about the three reading methods.
Finally, Dr. Ellis [1] indicates that it is necessary to list the reading performance of each reader. This was not included in our report in the interest of brevity. However, all nine radiologists showed an increased sensitivity with CAD compared with their first readings, with both mean and median absolute increases of 9% (range, 1–25%). The benefit of CAD to these experienced readers (mean, 15 years of experience; mean, 8,895 annual readings during the study period) calls into question the assertion of Dr. Ellis that CAD benefits only inexperienced readers.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
