July 28, 2011 — Computer-aided detection (CAD) software used for the analysis and interpretation of mammograms does not improve accuracy. So concludes a new analysis of a large database of digitized film mammography results from 90 facilities in the United States.
On a further negative note, CAD was associated with decreased specificity and an increased likelihood that a woman would be called back for further testing, say the authors, led by Joshua Fenton, MD, from the University of California Davis in Sacramento.
"Computer-aided detection does not act independently of human beings," Dr. Felton explained to Medscape Medical News. "It's supposed to be a 'second reader'."
"We discovered that having a second look didn't make much of a difference in changing the initial review," he said.
The study was published online July 27 in the Journal of the National Cancer Institute.
CAD software marks potential abnormalities for a radiologist to consider before making a final recommendation. The software uses algorithms to identify those abnormalities — that is, patterns associated with underlying breast cancers, the authors explain.
Therein lies a problem, says Donald Berry, PhD, from the University of Texas M.D. Anderson Cancer Center in Houston, in an editorial accompanying the study, which he describes as "important."
"Reality is that it is not easy to program a robot to recognize patterns," writes Dr. Berry. There is a notion that computers are "infallible," which is reinforced by much-publicized software such as Deep Blue, the chess player, and its "winning ways" against top-level competitors, he says.
So far, artificial intelligence has come up short in analyzing mammograms, he explains. "CAD is only as effective as its computer software," writes Dr. Berry, who believes the breast cancer algorithms might improve with "time and experience."
Despite having "equivocal health benefits," CAD is widely used in the United States, note Dr. Fenton and his coauthors.
Reportedly, 74% of all mammograms in the United States are read with the aid of CAD, adds Dr. Berry, who asks: Why is it so popular?
"An obvious reason is that it is built into digital mammography equipment, which is increasingly common in the United States," he says, answering his own question. "Another is financial: In 2008, Medicare's global reimbursement for CAD was $16.50. Still another is that CAD marks are comforting to the reader, even though the comfort may be misplaced."
The use of CAD for screening mammograms accounts annually for more than $30 million of direct Medicare costs, the study authors indicate. They also say that CAD was approved by the US Food and Drug Administration in 1998 on the basis of "small studies" that indicated it could increase breast cancer detection. The Medicare statute was subsequently amended to include coverage for its use.
No Help With Invasive Cancers
In their study, the authors analyzed data from 1.64 million film screening mammograms from more than 680,00 women carried out at facilities in 7 states from 1998 to 2006. The facilities participate in the Breast Cancer Surveillance Consortium, a federally supported network in which data quality is "rigorously monitored," say the authors.
Of 90 total facilities, 25 adopted CAD (27.8%) and used it for an average of 27.5 months during the study period. Screening mammograms for women 40 years and older were included and were defined as bilateral mammograms designated by radiologists.
The researchers collected information on women who had mammograms with and without CAD, including whether they were diagnosed with breast cancer (either ductal carcinoma in situ [DCIS] or invasive) within a year of the screening. The information included data on mammographic performance in facilities that switched to CAD (n = 25), comparing the periods before and after the switch. Also included were data on mammographic performance at facilities that did not switch to CAD (n = 65)
Odds ratios (ORs) and 95% confidence intervals (CIs) for CAD use and nonuse were estimated using random-effects logistic regression.
The ORs were adjusted for mammography registry, patient age, breast density, time since previous mammography, current hormone replacement therapy, and year of examination (1998 to 2002 or 2003 to 2006).
CAD use was associated with statistically significantly lower specificity (OR, 0.87; 95% CI, 0.85 to 0.89; P < .001), say the authors. They defined specificity as the proportion of screening mammograms that were negative in women who were not diagnosed with breast cancer.
On a positive note, CAD use was associated with an increase in overall sensitivity (OR, 1.06; 95% CI, 0.84 to 1.33; P = .62), but this was not statistically significant and was attributed to increased sensitivity for DCIS (OR, 1.55; 95% CI, 0.83 to 2.91; P = .17). Sensitivity was defined as the proportion of screening mammograms that were positive in women diagnosed with breast cancer, including interval breast cancers, within 1 year of screening mammography.
This increase in sensitivity for DCIS was expected, say the authors, because CAD is sensitive in detecting calcifications, which are highly characteristic of DCIS.
Sensitivity for invasive cancer was similar with or without CAD (OR, 0.96; 95% CI, 0.75 to 1.24; P = .77).
CAD was also associated with a statistically significantly lower positive predictive value (OR, 0.89; 95% CI, 0.80 to 0.99; P = .03). Positive predictive value was defined as the proportion of women diagnosed with breast cancer after a positive screening mammogram.
"CAD was not associated with higher breast cancer detection rates or more favorable stage, size, or lymph node status of invasive breast cancer," the authors report.
The fact that CAD did not improve detection of invasive cancer, including smaller-sized invasive cancers, is especially concerning to the authors.
"These findings raise concerns that CAD, as currently implemented in clinical practice, may have little or no impact on breast cancer mortality, which may depend on the earlier detection of invasive breast cancer," note the authors.
The study was supported by the National Cancer Institute–funded Breast Cancer Surveillance Consortium cooperative agreement, the National Cancer Institute, and the American Cancer Society. The collection of cancer data used in this study was supported in part by several state public health departments and cancer registries throughout the United States. The study authors and editorialist have disclosed no relevant financial relationships.
J Natl Cancer Inst. Published online July 27, 2011. Abstract, Editorial
On a further negative note, CAD was associated with decreased specificity and an increased likelihood that a woman would be called back for further testing, say the authors, led by Joshua Fenton, MD, from the University of California Davis in Sacramento.
"Computer-aided detection does not act independently of human beings," Dr. Felton explained to Medscape Medical News. "It's supposed to be a 'second reader'."
"We discovered that having a second look didn't make much of a difference in changing the initial review," he said.
The study was published online July 27 in the Journal of the National Cancer Institute.
CAD software marks potential abnormalities for a radiologist to consider before making a final recommendation. The software uses algorithms to identify those abnormalities — that is, patterns associated with underlying breast cancers, the authors explain.
Therein lies a problem, says Donald Berry, PhD, from the University of Texas M.D. Anderson Cancer Center in Houston, in an editorial accompanying the study, which he describes as "important."
"Reality is that it is not easy to program a robot to recognize patterns," writes Dr. Berry. There is a notion that computers are "infallible," which is reinforced by much-publicized software such as Deep Blue, the chess player, and its "winning ways" against top-level competitors, he says.
So far, artificial intelligence has come up short in analyzing mammograms, he explains. "CAD is only as effective as its computer software," writes Dr. Berry, who believes the breast cancer algorithms might improve with "time and experience."
Despite having "equivocal health benefits," CAD is widely used in the United States, note Dr. Fenton and his coauthors.
Reportedly, 74% of all mammograms in the United States are read with the aid of CAD, adds Dr. Berry, who asks: Why is it so popular?
"An obvious reason is that it is built into digital mammography equipment, which is increasingly common in the United States," he says, answering his own question. "Another is financial: In 2008, Medicare's global reimbursement for CAD was $16.50. Still another is that CAD marks are comforting to the reader, even though the comfort may be misplaced."
The use of CAD for screening mammograms accounts annually for more than $30 million of direct Medicare costs, the study authors indicate. They also say that CAD was approved by the US Food and Drug Administration in 1998 on the basis of "small studies" that indicated it could increase breast cancer detection. The Medicare statute was subsequently amended to include coverage for its use.
No Help With Invasive Cancers
In their study, the authors analyzed data from 1.64 million film screening mammograms from more than 680,00 women carried out at facilities in 7 states from 1998 to 2006. The facilities participate in the Breast Cancer Surveillance Consortium, a federally supported network in which data quality is "rigorously monitored," say the authors.
Of 90 total facilities, 25 adopted CAD (27.8%) and used it for an average of 27.5 months during the study period. Screening mammograms for women 40 years and older were included and were defined as bilateral mammograms designated by radiologists.
The researchers collected information on women who had mammograms with and without CAD, including whether they were diagnosed with breast cancer (either ductal carcinoma in situ [DCIS] or invasive) within a year of the screening. The information included data on mammographic performance in facilities that switched to CAD (n = 25), comparing the periods before and after the switch. Also included were data on mammographic performance at facilities that did not switch to CAD (n = 65)
Odds ratios (ORs) and 95% confidence intervals (CIs) for CAD use and nonuse were estimated using random-effects logistic regression.
The ORs were adjusted for mammography registry, patient age, breast density, time since previous mammography, current hormone replacement therapy, and year of examination (1998 to 2002 or 2003 to 2006).
CAD use was associated with statistically significantly lower specificity (OR, 0.87; 95% CI, 0.85 to 0.89; P < .001), say the authors. They defined specificity as the proportion of screening mammograms that were negative in women who were not diagnosed with breast cancer.
On a positive note, CAD use was associated with an increase in overall sensitivity (OR, 1.06; 95% CI, 0.84 to 1.33; P = .62), but this was not statistically significant and was attributed to increased sensitivity for DCIS (OR, 1.55; 95% CI, 0.83 to 2.91; P = .17). Sensitivity was defined as the proportion of screening mammograms that were positive in women diagnosed with breast cancer, including interval breast cancers, within 1 year of screening mammography.
This increase in sensitivity for DCIS was expected, say the authors, because CAD is sensitive in detecting calcifications, which are highly characteristic of DCIS.
Sensitivity for invasive cancer was similar with or without CAD (OR, 0.96; 95% CI, 0.75 to 1.24; P = .77).
CAD was also associated with a statistically significantly lower positive predictive value (OR, 0.89; 95% CI, 0.80 to 0.99; P = .03). Positive predictive value was defined as the proportion of women diagnosed with breast cancer after a positive screening mammogram.
"CAD was not associated with higher breast cancer detection rates or more favorable stage, size, or lymph node status of invasive breast cancer," the authors report.
The fact that CAD did not improve detection of invasive cancer, including smaller-sized invasive cancers, is especially concerning to the authors.
"These findings raise concerns that CAD, as currently implemented in clinical practice, may have little or no impact on breast cancer mortality, which may depend on the earlier detection of invasive breast cancer," note the authors.
The study was supported by the National Cancer Institute–funded Breast Cancer Surveillance Consortium cooperative agreement, the National Cancer Institute, and the American Cancer Society. The collection of cancer data used in this study was supported in part by several state public health departments and cancer registries throughout the United States. The study authors and editorialist have disclosed no relevant financial relationships.
J Natl Cancer Inst. Published online July 27, 2011. Abstract, Editorial
Δεν υπάρχουν σχόλια:
Δημοσίευση σχολίου