Digital and Film Mammography
http://www.100md.com
《新英格兰医药杂志》
To the Editor: Pisano et al. (Oct. 27 issue)1 report that digital mammography is more accurate in women under the age of 50 years, women with radiographically dense breast tissue, and premenopausal or perimenopausal women. They showed that digital and film mammography are similar with respect to the overall diagnostic accuracy. The latter finding is consistent with previous reports.2,3,4 We are disturbed that the authors did not provide a stratified analysis for the remainder of women, who make up more than half the study population (women 50 years of age or older, women with nondense breast tissue, and postmenopausal women).
We have calculated the mammographic sensitivity for invasive breast cancers on the basis of the data in the article by Pisano et al. (Table 1). The sensitivity of digital mammography is lower than that of film mammography among women 50 years of age or older, postmenopausal women, and women with nondense breasts. This fact merits thorough discussion.
Table 1. Sensitivity of Digital vs. Film Mammography for Detecting Invasive Breast Cancers.
Pavel Crystal, M.D.
University of Toronto
Toronto, ON M5S 3E2, Canada
pavel.crystal@utoronto.ca
Selwyn Strano, M.D.
Rachel Nash Comprehensive Breast Clinic
95484 Jerusalem, Israel
References
Pisano ED, Gatsonis C, Hendrick E, et al. Diagnostic performance of digital versus film mammography for breast-cancer screening. N Engl J Med 2005;353:1773-1783.
Lewin JM, D'Orsi CJ, Hendrick RE, et al. Clinical comparison of full-field digital mammography and screen-film mammography for detection of breast cancer. AJR Am J Roentgenol 2002;179:671-677.
Skaane P, Young K, Skjennald A. Population-based mammography screening: comparison of screen-film and full-field digital mammography with soft-copy reading -- Oslo I study. Radiology 2003;229:877-884.
Skaane P, Skjennald A. Screen-film mammography versus full-field digital mammography with soft-copy reading: randomized trial in a population-based screening program -- the Oslo II Study. Radiology 2004;132:197-204.
To the Editor: If digital mammography is more sensitive in women under 50 years of age but not better overall, in what populations is its sensitivity worse than that of film mammography? In Figure 1 of the study by Pisano et al., the three variables — breast density, menopausal status, and age — appear to be correlated; was this correlation taken into account in the analysis? Also, statistics were generated for the area under the curve (AUC). Since any radiologist who had a false positive rate or a recall rate of 20 percent would probably be asked to stop reading mammograms, isn't the appropriate statistical analysis the one for the area to the left of a 20 percent false positive rate, or partial AUC?1
Finally, were the academic readers deliberately reading at a low recall rate to improve specificity, judging by the clustering of data points? The sensitivity noted is only 51 percent for one year, which is below the population-based average of 68 percent.2
John D. Keen, M.D., M.B.A.
John H. Stroger Hospital of Cook County
Chicago, IL 60513
jkeen@ccbhs.org
References
Obuchowski NA. ROC analysis. AJR Am J Roentgenol 2005;184:364-372.
Carney PA, Miglioretti DL, Yankaskas BC, et al. Individual and combined effects of age, breast density, and hormone replacement therapy use on the accuracy of screening mammography. Ann Intern Med 2003;138:168-175.
To the Editor: If 10 of 10,000 premenopausal or perimenopausal women have invasive cancer, digital mammography, as compared with film mammography, will detect two additional cancers (sensitivity, 72 percent vs. 51 percent). If only one of the two patients benefits from earlier detection — which is an optimistic estimate — and the digital mammogram costs $100 more (the authors note that digital mammography is 1.5 to 4 times as expensive as film), we will have spent $1 million to save one additional life. This cost is far higher than the usual cutoff for cost-effective medical technology. And of course, it is likely that the use of digital mammography will not be limited to just one of these subgroups but will be used for all women, making its use even less cost-effective.
Mark H. Ebell, M.D.
Michigan State University
East Lansing, MI 48824
The authors reply: Table 1 in the letter from Drs. Crystal and Strano, which compares the sensitivity of digital and film mammography, is fatally flawed. It does not account for multiple comparisons, nor does it take into consideration the correlation in the test data. This correlation results from the paired design used in the study, according to which each participant underwent both types of mammography. The joint test results for each participant are not presented in our article. However, as pointed out in the Methods section, for the comparisons of all measures of diagnostic performance (AUC, sensitivity, and specificity), we used methods appropriate for paired test data. As for subgroup analyses, we followed two commonly accepted practices: we specified the subgroups in advance and controlled for multiple comparisons in the analysis.
In response to Dr. Keen: we note that the full AUC is the most common summary of the receiver-operating-characteristic curve. Its use obviates the need to choose an arbitrary minimum value for specificity, as would be required for computation of a partial AUC.
We also compared the results within strata defined according to age, breast density, menopausal status, race, risk of breast cancer, and each of the four digital-machine manufacturers. These variables were prespecified and were considered one at a time in the analysis. We did not pursue further subgroups defined according to combinations of these variables. Because of space limitations, we did not report the AUC for subgroups in which the comparison was not statistically significant. In line with our stated goal to treat the reporting on sensitivity, specificity, and predictive value as descriptive, we did not control for multiple comparisons in this part of the analysis. Finally, in response to Dr. Ebell: a formal cost-effectiveness analysis of the Digital Mammographic Imaging Screening Trial is under way.
Constantine Gatsonis, Ph.D.
Brown University
Providence, RI 02912
Etta D. Pisano, M.D.
University of North Carolina at Chapel Hill
Chapel Hill, NC 27599
etta_pisano@med.unc.edu
Edward Hendrick, Ph.D.
Northwestern University
Evanston, IL 60208
We have calculated the mammographic sensitivity for invasive breast cancers on the basis of the data in the article by Pisano et al. (Table 1). The sensitivity of digital mammography is lower than that of film mammography among women 50 years of age or older, postmenopausal women, and women with nondense breasts. This fact merits thorough discussion.
Table 1. Sensitivity of Digital vs. Film Mammography for Detecting Invasive Breast Cancers.
Pavel Crystal, M.D.
University of Toronto
Toronto, ON M5S 3E2, Canada
pavel.crystal@utoronto.ca
Selwyn Strano, M.D.
Rachel Nash Comprehensive Breast Clinic
95484 Jerusalem, Israel
References
Pisano ED, Gatsonis C, Hendrick E, et al. Diagnostic performance of digital versus film mammography for breast-cancer screening. N Engl J Med 2005;353:1773-1783.
Lewin JM, D'Orsi CJ, Hendrick RE, et al. Clinical comparison of full-field digital mammography and screen-film mammography for detection of breast cancer. AJR Am J Roentgenol 2002;179:671-677.
Skaane P, Young K, Skjennald A. Population-based mammography screening: comparison of screen-film and full-field digital mammography with soft-copy reading -- Oslo I study. Radiology 2003;229:877-884.
Skaane P, Skjennald A. Screen-film mammography versus full-field digital mammography with soft-copy reading: randomized trial in a population-based screening program -- the Oslo II Study. Radiology 2004;132:197-204.
To the Editor: If digital mammography is more sensitive in women under 50 years of age but not better overall, in what populations is its sensitivity worse than that of film mammography? In Figure 1 of the study by Pisano et al., the three variables — breast density, menopausal status, and age — appear to be correlated; was this correlation taken into account in the analysis? Also, statistics were generated for the area under the curve (AUC). Since any radiologist who had a false positive rate or a recall rate of 20 percent would probably be asked to stop reading mammograms, isn't the appropriate statistical analysis the one for the area to the left of a 20 percent false positive rate, or partial AUC?1
Finally, were the academic readers deliberately reading at a low recall rate to improve specificity, judging by the clustering of data points? The sensitivity noted is only 51 percent for one year, which is below the population-based average of 68 percent.2
John D. Keen, M.D., M.B.A.
John H. Stroger Hospital of Cook County
Chicago, IL 60513
jkeen@ccbhs.org
References
Obuchowski NA. ROC analysis. AJR Am J Roentgenol 2005;184:364-372.
Carney PA, Miglioretti DL, Yankaskas BC, et al. Individual and combined effects of age, breast density, and hormone replacement therapy use on the accuracy of screening mammography. Ann Intern Med 2003;138:168-175.
To the Editor: If 10 of 10,000 premenopausal or perimenopausal women have invasive cancer, digital mammography, as compared with film mammography, will detect two additional cancers (sensitivity, 72 percent vs. 51 percent). If only one of the two patients benefits from earlier detection — which is an optimistic estimate — and the digital mammogram costs $100 more (the authors note that digital mammography is 1.5 to 4 times as expensive as film), we will have spent $1 million to save one additional life. This cost is far higher than the usual cutoff for cost-effective medical technology. And of course, it is likely that the use of digital mammography will not be limited to just one of these subgroups but will be used for all women, making its use even less cost-effective.
Mark H. Ebell, M.D.
Michigan State University
East Lansing, MI 48824
The authors reply: Table 1 in the letter from Drs. Crystal and Strano, which compares the sensitivity of digital and film mammography, is fatally flawed. It does not account for multiple comparisons, nor does it take into consideration the correlation in the test data. This correlation results from the paired design used in the study, according to which each participant underwent both types of mammography. The joint test results for each participant are not presented in our article. However, as pointed out in the Methods section, for the comparisons of all measures of diagnostic performance (AUC, sensitivity, and specificity), we used methods appropriate for paired test data. As for subgroup analyses, we followed two commonly accepted practices: we specified the subgroups in advance and controlled for multiple comparisons in the analysis.
In response to Dr. Keen: we note that the full AUC is the most common summary of the receiver-operating-characteristic curve. Its use obviates the need to choose an arbitrary minimum value for specificity, as would be required for computation of a partial AUC.
We also compared the results within strata defined according to age, breast density, menopausal status, race, risk of breast cancer, and each of the four digital-machine manufacturers. These variables were prespecified and were considered one at a time in the analysis. We did not pursue further subgroups defined according to combinations of these variables. Because of space limitations, we did not report the AUC for subgroups in which the comparison was not statistically significant. In line with our stated goal to treat the reporting on sensitivity, specificity, and predictive value as descriptive, we did not control for multiple comparisons in this part of the analysis. Finally, in response to Dr. Ebell: a formal cost-effectiveness analysis of the Digital Mammographic Imaging Screening Trial is under way.
Constantine Gatsonis, Ph.D.
Brown University
Providence, RI 02912
Etta D. Pisano, M.D.
University of North Carolina at Chapel Hill
Chapel Hill, NC 27599
etta_pisano@med.unc.edu
Edward Hendrick, Ph.D.
Northwestern University
Evanston, IL 60208