| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorials |
1 From the Departments of Radiology and Biomedical Engineering, University of North Carolina at Chapel Hill, 4030 Bondurant Hall, CB 7000, Chapel Hill, NC 27599 (E.D.P.); Department of Radiology, Northwestern University, Chicago, Ill (R.E.H.); Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada (M.Y.); Department of Radiology, University of Pennsylvania, Philadelphia, Pa (E.C.); and Department of Community Health and Applied Mathematics, Brown University, Providence, RI (C.G.). The University of North Carolina has a research agreement with GE Healthcare; R.E.H. has received honoraria from GE Healthcare for continuing medical education lectures. Received February 27, 2007; final version accepted February 28. Address correspondence to E.D.P. (e-mail: etpisano{at}med.unc.edu).
The Digital Mammographic Imaging Screening Trial (DMIST) enrolled 49 528 women in 33 centers in the United States and Canada (1), with 42 760 evaluable patients (86%) included in the analysis. Its results showed no difference in the performance of film and digital mammography as measured by area under the receiver operating characteristic curve (AUC) for the entire population studied. We did find, however, that the diagnostic accuracy of digital mammography was superior to that of film mammography for women younger than 50 years of age, women with dense breasts, and pre- and perimenopausal women (2). These results were statistically significant, with AUC differences ranging from 0.106 for women with dense breasts (P = .0033) to 0.151 for women younger than 50 years of age (P = .0023) in these subsets. Importantly, the diagnostic performance of digital mammography showed improvements in sensitivity over film mammography in these subgroups, without differences in specificity (2).
Gur (3) raises doubts about the reliability of DMIST findings because of the relative performance of digital mammography in younger and older women. He uses the published DMIST results that the AUC of digital mammography overall was 0.78, while the AUC of digital mammography in women younger than 50 years of age was 0.84, to argue that the results of DMIST are implausible. He goes on to suggest that because the AUC of digital mammography was 0.06 higher in younger women than in the overall population, the results of DMIST are "inconsistent" and should not be the primary reason for sites to convert to digital mammography (even though he argues that sites should convert to digital mammography). Dr Gur makes an interesting observation, but not one that necessarily casts doubt on DMIST's comparison of digital and film mammography.
Dr Gur's observation that digital mammography performed better in younger women than in older women needs to be held to the same scientific scrutiny as the DMIST comparison of digital and film mammography. That is, first, careful statistical analysis needs to be performed to see if there is a statistically significant difference between the performance of digital mammography in younger and older women and between women with dense and fatty breasts. DMIST investigators are currently performing additional exploratory analyses of DMIST data.
The fact that the DMIST digital mammography results fail to meet Dr Gur's a priori expectations based on film mammography in younger versus older women is not a valid reason to disregard DMIST results for the comparison of digital and film mammography. It should be noted that digital mammography need not follow the behavior of film mammography by having higher sensitivities or greater AUC values for older women or women with fatty breasts. Screening studies of digital mammography before DMIST included few women (and very few breast cancers among women) younger than 50 years of age, and none of those prior studies had sufficient power to permit analysis of results according to breast density (4–6).
There may be valid reasons that digital mammography performed as well or better in women with dense breasts. Image processing algorithms in digital mammography are known to significantly affect the visibility of breast lesions (7). It has been demonstrated that radiologists prefer different algorithms for different diagnostic functions and machine types (8). To compensate for the shortcomings of film mammography, which is known to be less sensitive for breast cancers in dense breasts (2,9,10), it is conceivable that manufacturers of the first-generation digital mammography systems evaluated in DMIST emphasized image processing and presentation algorithms that performed well for dense breasts, at the expense of performance in women with fatty breasts.
In distinction to Dr Gur's view, we believe that the DMIST results in younger versus older women (or, more specifically, in denser vs fatty breasts) may be telling us something important about the image processing and presentation algorithms of first-generation digital mammography systems. DMIST results may be signaling the need for manufacturers to use more sophisticated image processing and presentation schemes that differ on the basis of breast density.
We are in the midst of exploring some of these factors in a careful review by experienced radiologists of the DMIST cancer cases on both film and digital mammograms. We hope that this review will determine the factors that led to improved lesion detection for women in various population subsets. These analyses will include a study of the effects of lesion type, machine type, and other factors on the detectability of DMIST cancers.
Another article that is currently undergoing peer review is the report of the DMIST cost-effectiveness analysis, of data specifically collected for this purpose. We anticipate that radiologists seeking to decide whether to convert to digital mammography will find that study useful in that it takes the viewpoint of federal payers who will be deciding how much to pay providers for the more expensive digital technology.
In summary, neither we nor Dr Gur have all the answers on the advantages and disadvantages of digital mammography. We strongly believe that results of carefully performed clinical research are the best basis for medical decision making. Further analysis of DMIST results and future studies of digital mammography in light of the DMIST results are the best ways to learn exactly how digital mammography should be used.
 |
FOOTNOTES |
|---|
|
References |
|---|
 TOP References |
|---|
Related Article
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
