«PacifiCare’s medical management guidelines represent the recommendation of the PacifiCare Medical Management Guideline (MMG) committee. They are ...»
The American College of Radiology (ACR) issued a statement regarding the study by Fenton et al.
The ACR states that "the reported decreased accuracy of mammography when using CAD was due to increased false positives, not to fewer cancers being detected. Refinements and improvements in CAD software are continually being made and it is possible that newer versions of CAD systems have better performance parameters. The study was designed in such a way that it was impossible to determine how many cancers in their series would have been missed without the use of CAD." The ACR goes on to state that there is a fair amount of evidence outside the current study by Fenton et al.
to suggest that CAD may be worthwhile and it seems unwarranted to consider abandoning this potentially valuable tool. (ACR Statement on NEJM Study Regarding CAD in Mammography, 2007) A study by Gromet (2008) compared the efficacy of single reading with CAD with double reading without CAD in a retrospective review of 231,221 screening mammograms. Results of the study included a lower recall with CAD compared with double reading (10.6% vs. 11.9%). Single reading with CAD demonstrated a higher sensitivity of 90.4% with a smaller increase in the recall rate from 10.2% to 10.6%. Gromet concluded that with manpower constraints limiting the use of double reading, CAD appears to be an effective alternative that provides potentially greater benefits. CAD also provides improved patient outcomes when added to a single reading.
Breast Imaging- Commercial Medical Management Guideline
In a prospective study, mammograms were obtained in 6,111 women and were analyzed with CAD and were also independently double read. CAD prompted 51 (84%) of 61 radiographically detected cancers. Sensitivity of a single reading was 90.2%, sensitivity of a single reading with CAD was 91.5%, and sensitivity of double reading without CAD was 98.4%. The investigators concluded that CAD increases sensitivity of single reading by 1.3%, and double reading increases sensitivity by 8.2%. (Khoo et al., 2005) Gur et al. computed recall and cancer detection rates for 24 radiologists who interpreted 115,571 screening mammograms with (n=59,139) or without (n=56,432) the use of CAD and found that the use of CAD was not associated with significant changes in recall and breast cancer detection rates.
(Gur et al., 2004) A prospective study assessed 5,016 mammograms that were interpreted with and without CAD and found that use of CAD significantly increases recall rates, has no significant effect of PPV, and can increase cancer detection rate by at least 4.7% and sensitivity by at least 4%. (Ko et al., 2006) Gilbert et al. conducted a study which included a sample of 10,267 mammograms. Mammograms that were double read in 1996 were randomly assigned to be re-read using CAD. The cancer detection and recall rates from double reading and single reading with CAD were compared. Single reading with CAD led to a significantly improved cancer detection rate and an increased recall rate (8.6% for single reading with CAD versus 6.5% for double reading). (Gilbert et al., 2006) A prospective clinical trial of 8,682 patients who underwent screening mammograms with CAD resulted in a 7.4% increase (from 27 to 29) in cancers detected. (Birdwell et al., 2005) A study was conducted to evaluate radiologists' performance before CAD (n=7,872) and with CAD (n=19,402). The investigators reported that increased detection rate, earlier stage of invasive cancer detection, and younger age at diagnosis are positive results of screening with CAD. (Cupples et al.,
2005) In a clinical trial, CAD was applied to 2,389 patients' screening mammograms. Of these patients, 11 had mammographically detected non-palpable breast cancers, and 10 of these 11 cancers were identified with CAD. CAD sensitivity was 91%. (Helvie et al., 2004) One large prospective study of 12,860 women at a community breast center reported that the use of computer-aided detection increased the cancer detection rate by 19.5%, without undue effect on the recall rate or the positive predictive value (PPV) for biopsy. (Freer and Ulissey, 2001) Two prospective studies found similar results when evaluating mammography with CAD. CAD use resulted in improved detection of cancer with an increased recall rate, but little or no deterioration in PPV of biopsy. (Dean and Ilvento, 2006; Morton et al., 2006) Brem et al. evaluated 906 sequential mammographically detected breast cancers and 147 normal mammograms and found that breast density does not impact overall CAD breast cancer detection.
Breast Imaging- Commercial Medical Management Guideline There is no statistically significant difference in breast cancer detection in dense and non-dense breasts. However, the detection of breast cancer manifesting as masses is impacted by breast density and CAD may be advantageous in dense breasts in these cases. (Brem et al., 2005)
Hayes reviewed the following studies:
In a retrospective study, Warren Burhenne et al. (2000) reported data obtained from screening mammography undertaken with and without CAD at 13 centers in the U.S. The study compared recall rate between mammograms obtained from 23,682 patients and single read before installation of CAD and mammograms obtained from 14,817 patients and single screened with CAD after installation of CAD. The addition of CAD to single reading decreased the recall rate from 8.3% before CAD installation of CAD to 7.6% after installation of CAD. The authors retrospectively examined 1083 mammograms that led to detection and subsequent biopsy proof of breast cancer. In these cases, the sensitivity of single reading with CAD was 84% for detecting any breast cancer but was substantially greater for detecting malignant microcalcifications (99%) than for detecting malignant masses (75%) and was not compared with single reading alone. Prior mammograms (n=427) available in a subgroup of these cases were obtained 9 to 24 months earlier, were deemed negative for cancer at that time after single reading alone, did not involve interpretation with CAD at that time, but, for the current study, were retrospectively reexamined with and without CAD by 5 independent radiologists blinded to all data. In the retrospective review of prior mammograms, 286 (67%) of the 427 subsequently diagnosed malignant lesions were deemed visible by ³ 1 of 5 radiologists; 115 (40%) of the visible lesions were deemed actionable (warranting recall) by ³ 1 of 5 radiologists; and CAD correctly marked 89 (77%) of the 115 lesions deemed actionable. According to the authors, these findings suggest that, if CAD had been applied at the time these mammograms were obtained and interpreted, it may have helped detect 89 breast cancers earlier, potentially leading to a 77% (89/115) absolute decrease in the false-negative rate. However, this conclusion assumes that radiologists will respond correctly to every correct CAD prompt.
Findings from a recent analysis by Jiang et al. (2007) suggest that, due to the low cancer rate in screening populations, as well as the high inter-radiologist variation in cancer detection rate, clinical trials comparing standard mammography with a new modality for breast cancer screening may require a large number of patients, as well as a relatively large number of interpreting radiologists, in order to demonstrate a significant improvement in cancer detection rate. Jiang et al. (2007) analyzed data from 2,289,132 screening mammograms read by 510 radiologists in the U.S. who participated in and read at least 500 mammograms for the Breast Cancer Surveillance Consortium between January 1996 and December 2002. Among this selection of mammograms, there were 9030 cancers, for an overall cancer detection rate of 3.94 cancers per 1000 screening examinations. For individual radiologists, the mean cancer detection rate was 3.91 cancers per 1000 examinations overall but increased with case volume (3.49, 3.94, 4.04, 4.18, and 4.30 cancers per 1000 examinations for radiologists reading 500 to 1000, 1000 to 2000, 2000 to 3000, 3000 to 4000, and 4000 to 6000 mammograms, respectively). Using statistical projections, Jiang and colleagues determined that, in order to achieve 80% power to detect one additional cancer per reader, a clinical trial comparing a Breast Imaging- Commercial Medical Management Guideline new modality and standard mammography would require at least 25 radiologists each reading images of 8000 examinations or 91 radiologists each reading images of 1000 to 2000 examinations.
Hayes reviewed data associated with screening with digital mammography and CAD. (Hayes Directory, Computer-Aided Detection for Screening Mammography, 2007) They found two retrospective studies involving a total of 257 asymptomatic patients that showed that the addition of CAD to single reading may improve the sensitivity of single reading alone in detecting breast cancer but is associated with false-positive CAD marks in most cases. (Skaane et al., 2007; Sang et al.,
2007) However, these studies provided no data for other outcome measures and are insufficient to establish the value of CAD for digital mammography. (Hayes Directory, Computer-Aided Detection for Screening Mammography, 2007) Kim et al. (2008) evaluated 93 women with breast cancer who had a CAD system applied to their initial and follow-up digital mammograms. According to the investigators, the results of the study demonstrated that the sensitivity of the CAD system was consistently high for detection of breast cancer on initial and short-term follow-up digital mammograms. A retrospective study included 3,425 women who had digital mammography reported without and then with CAD. The investigators found that while the increase in cancer detection rate with the use of CAD only approached statistical significance, the incremental number of cancers detected justified the incremental recall and benign surgical biopsy attributable to CAD use.
(Brancato et al., 2007) Yang et al. (2007) applied a CAD system to the digital mammograms of 103 women with 103 breast cancers detected with screening. The CAD system correctly marked all 31 lesions of ductal carcinoma in situ (DCIS), all 22 lesions of invasive ductal carcinoma with DCIS, the single invasive lobular carcinoma lesion, and 45 (92%) of 49 lesions of invasive ductal carcinoma. The investigators concluded that the CAD system correctly marks most (96.1%) asymptomatic breast cancers detected with digital mammographic screening, with acceptable falsepositive marks.
Retrospective case or case-controlled studies evaluating CAD with full-field digital mammography, found that CAD systems were able to detect the lesions with high sensitivity, but had a high rate of false-positives (0.3 to 2.2 false-positives per image). (ECRI, Computer-aided detection with full-field digital mammography, 2006) The Technology Evaluation Center (TEC) Blue Cross Blue Shield Association (BCBSA) completed a technology assessment for computer-aided detection (CAD) with full-field digital mammography (FFDM). The report states the studies on the use of CAD with FFDM are inadequate to determine the incremental impact on sensitivity, specificity, and biopsy rates. The use of CAD systems as an adjunct to single reading of full-field digital mammography images did not meet TEC criteria.
(Technology Evaluation Center (TEC) Blue Cross Blue Shield Association, May 2006b)
American Cancer Society (ACS): The ACS guidelines for breast cancer screening indicate that for CAD with screen-film mammography or full-field digital mammography, there is some clinical Breast Imaging- Commercial Medical Management Guideline evidence for effectiveness or equivalence to screen-film mammography for screening (Evidence Level B). (Smith et al., 2003) National Cancer Institute (NCI): The NCI states that since there are no mortality studies regarding CAD, the impact of CAD on breast cancer mortality is uncertain. (NCI, 2007) American College of Radiology (ACR): The ACR practice guideline for the performance of screening mammography states that CAD may increase the sensitivity of mammogram interpretation and may be used even though cost and workforce issues may make this difficult to accomplish. (ACR Practice Guideline for the Performance of Screening Mammography, 2006)
Computer-Aided Detection for MRI of the Breast:
The published evidence on CAD MRI consists of 2 small clinical studies (Demartini et al., 2005;
Lehman et al., 2006) performed with the CADStream and a few studies that reported on the development and validation of 3TP. (Hauth et al., 2006; Kelcz et al., 2002) and non-commercial CAD systems (Deurloo et al; Pediconi et al., 2005; Meinel et al., 2007) DeMartini et al. (2005) performed a study to determine the utility of CADStream applied to breast MRI in 15 newly diagnosed breast cancer patients (16 lesions) undergoing neoadjuvant chemotherapy. Prior to chemotherapy, all tumors demonstrated CAD-assessed significant enhancement. Following chemotherapy, 7/16 tumors showed no residual significant enhancement, but all had residual disease at pathology. The authors concluded that CAD may be helpful in assessing changes in MRI enhancement profiles of tumors following chemotherapy. However, CADassessed significant enhancement following chemotherapy can be falsely negative for residual malignancy, and CAD tumor sizes are less accurate than those measured by the radiologist in predicting size of residual malignancy. CAD may complement but should not replace the radiologist's assessment of tumors in this patient population.
Lehman et al. (2006) compared the accuracy of breast MRI interpretations of 33 consecutive lesions seen only on MRI (9 malignant, 24 benign) with and without CADstream. For benign lesions, the false-positive rate was reduced by 25% to 50%, depending on the enhancement thresholds used for the analysis. The authors concluded that CADstream accurately showed significant enhancement in all the malignant lesions while depicting 12 of 24 benign lesions as showing insignificant enhancement. The authors further stated if these results are validated by a larger study, the number of unnecessary biopsies of MR lesions could be reduced without a concomitant decrease in cancer detection.
The Technology Evaluation Center (TEC) Blue Cross Blue Shield Association (BCBSA) completed a technology assessment for CAD with MRI. The report concluded that there are no high quality, current published studies of the impact of commercially available CAD systems on the sensitivity and specificity of MRI of the breast. The few studies and abstracts available focus primarily on the