ARCHIVED

Medical Policy

Policy Num:      06.001.021
Policy Name:    Computer-Aided detection in Conjuction with Digitized Screen-Film Mammography or Ful-Field Digital Mammography
Policy ID:          [06.001.021]  [Ar / B / M+ / P]  [6.01.39]

Last Review:      December 16, 2019
Next Review:      Policy Archived
Issue:                  12:2019

ARCHIVED

Related Policies BCBS:  6.01.34 Full-Field Digital Mammography

                                               6.01.39   Computer-aided detection using digital or digitized mammogram

Related Policies TSSS: 06.001.051  Digital Breast Tomosynthesis 

Computer-Aided detection in Conjuction with Digitized Screen-Film Mammography or Ful-Field Digital Mammography

Summary

Computer-aided detection (CAD) has been suggested as an adjunct to screening mammograms to decrease errors in perception, ie, failure to see an abnormality. The use of CAD systems requires a digital image, either generated by digitization of a prior screen-film mammogram (digitized mammogram), or generated directly (direct full-field digital mammogram). The effectiveness of CAD needs to be evaluated separately for these two types of digital images.

Commercially available CAD systems use computerized algorithms for identifying suspicious regions of interest on the digital image. The locations of the abnormalities are marked such that the reader can then reference the same areas in the original mammogram for further review. The intent of CAD is to aid in detection of potential abnormalities for the radiologist to re-review. The radiologist, not CAD, makes the diagnosis if a clinically significant abnormality exists and whether future diagnostic evaluation is warranted. The distinction between digitized screen-film mammograms (SFM) and direct full-field digital mammograms (FFDM) is important. Since these two images are generated in different ways, the associated diagnostic performance of adjunctive CAD must be considered separately. Conceptually, the CAD systems used with digital mammography are very similar to those used with film mammography. The computer analyzes the digital images collected directly by the FFDM system, applies a set of algorithms that capture characteristics known to be associated potentially with malignancies, and produces an image with markings that show the site of suspicious findings. Sometimes, different marks are used for suspected masses and suspected microcalcifications. The major difference between CAD for FFDM and CAD for SFM is the extensive data set provided by the former and its interaction with the CAD algorithms.

Objective

This purpose of this evidence review is to evaluate if the use of Computer-Aided Detection with digital mammography  to reduce risk of missing an abnormality on 

Policy Statements

Computer-aided detection devices as an adjunct to single-reader interpretation of digitized screen-film mammograms may be considered medically necessary.

Computer-aided detection devices as an adjunct to single-reader interpretation of direct, full-field digital mammography is considered investigational.

Policy Guidelines

There are a series of 3 HCPCS G codes (G0202, G0204, G0206) that specifically describe direct (de novo) full-field digital mammography. According to this policy, all 3 of these codes would be considered investigational, if used in conjunction with CAD. There is no HCPCS code that describes the generation of digitized mammogram, either with or without adjunctive CAD.

In 2004, two CPT codes were introduced that specifically describe computer-assisted detection of mammography regardless of whether the image is digitized from screen film mammograms or direct full field de novo digital mammography:

CPT 77051: Computer-assisted detection (computer algorithm analysis of digital image data for lesion detection) with further physician review for interpretation, with or without digitization of film radiographic images; diagnostic mammography (List separately in addition to code for primary procedure) CPT 77052: As above, but for screening mammography.

According to this policy, the above CPT codes would be considered medically necessary if unaccompanied by HCPCS codes G0202, G0204, or G0206.

There is no CPT code that describes direct full-field digital mammography, either with or without adjunctive CAD.

Benefit Application

BlueCard/National Account Issues

While there is no national Medicare coverage decision regarding digital mammography, coverage of digital mammography was addressed legislatively in the Benefits Improvement and Protection Act 2000 (BIPA). Under a provision in this act, payment for technologies that directly take digital images would equal 150% of the amount that would otherwise be paid for a bilateral diagnostic mammography. State or federal mandates (eg, FEP) may dictate that all devices approved by FDA may not be considered investigational. Therefore, FDA-approved devices may be assessed on the basis of their medical necessity.

Background

N/A

Regulatory Status

Several CAD systems for use with mammography have been approved by the Food and Drug Administration through the premarket approval process. These include: · R2 ImageChecker (Hologic): Approved in 1998 for use with film-based mammography and in 2001 for use with digital mammography.

· Second Look (Cadx Medical Systems): Approved in 2001 to identify and mark regions of interest on standard mammograms.

· Kodak Mammography CAD Engine (Eastman Kodak): Approved in 2004 for use in identifying and marking areas of interest on film mammograms

Rationale

The policy regarding computer-aided detection is based on a 2002 TEC Assessment. (1) The Assessment focused on 2 techniques: (1) computer-aided detection (CAD) as an adjunct to a digitized screen-film mammogram and (2) CAD as an adjunct to a direct full-field digital mammogram. The Assessment offered the following observations and conclusions.

· The available evidence suggests that the use of CAD as an adjunct to the radiologist’s interpretation of screen-film mammography improves net health outcomes compared with single reader radiologist interpretations by increasing true positive rate without a disproportionate increase in the false positive rate.

 · However, there were no published studies identified using CAD in full-field digital mammography. It is important to separately study the clinical effectiveness of CAD systems as applied in full-field mammography and not to generalize results of CAD in screen-film mammography. Thus, the Assessment concluded that there is insufficient evidence to permit conclusions on the effect on health outcomes of using CAD as an adjunct to the radiologists’ interpretation of full-field digital mammography.

In February 2006, the TEC Assessment on CAD with full-field digital mammography was updated. (2) The Assessment noted that the results of the American College of Radiology Imaging Network’s Digital Mammographic Imaging Screening Trial (DMIST) were released in the fall of 2005.

(3,4) The trial showed with reasonable certainty that there was no difference in the accuracy of direct full-field digital mammograms (FFDM) and digitized screen-film mammograms (SFM) for asymptomatic women in general. For three subgroups of women, however, FFDM performed better than SFM: women younger than age 50; pre- or perimenopausal women; and women with heterogeneously dense or extremely dense breasts. Now that the performance of FFDM has been demonstrated, the question is whether the use of CAD can improve its sensitivity and specificity, which continue to be lower than optimal even for FFDM.

The updated TEC Assessment did not find any high-quality articles in peer-reviewed journals that assessed the use of CAD as an adjunct to FFDM. The Assessment concluded it is not possible to judge the impact of CAD on single readings of digital mammograms at this time. While there is evidence that the use of CAD with SFM is equal to or better than a single reading of the SFM images, there is scant information on the performance of CAD with FFDM. Logically, it might seem that CAD should play the same role with FFDM as with SFM, but the differences between film and digital mammography—which show up in the greater accuracy of FFDM in certain populations—preclude extrapolating from the impact of CAD with SFM to CAD with FFDM. The large increase in the magnitude of the data collected by FFDM, the ability to fine-tune the digitally acquired images, and the elimination of the digitization step make FFDM sufficiently different from SFM that separate studies on the impact of CAD on FFDM are needed. Until those results become available, the benefits of CAD with FFDM remain a promising possibility. Therefore, the impact of CAD on cancer detection, treatment, and survival is unknown and the policy statement remains unchanged.   

2007–2008 Update The policy was updated with a literature search using MEDLINE conducted in April 2008.

Digitized Film-Screen Mammography. Similar to earlier studies reported in the TEC Assessment, (1) Dean reported on results of a prospective study of CAD with SFM in 9520 consecutive mammograms that demonstrated improved detection of cancers in a screening and diagnostic situation. (5) The greatest impact was on detection of ductal carcinoma in situ. Morton reported on a prospective study of 21,349 screening mammograms obtained in 18,096 women interpreted first without and then with review of CAD images to determine the effect of CAD analysis. (6) On the basis of pre-CAD interpretations, the breast cancer detection rate per 1000 screening mammograms was 4.92 (105/21,349 mammograms), the recall rate was 9.84% (2101/21,349 mammograms), and the positive predictive value (PPV) for biopsy was 41.0% (105/256 biopsies). After CAD image review, 199 additional patients were recalled, 21 additional biopsies were performed, and 8 additional cancers were detected. The effect was a 7.62% (8/105) increase in the number of breast cancers detected, an increase in the recall rate to 10.77% (2300/21,349 mammograms), and a slight decrease in the PPV to 40.8%. The authors concluded that use of CAD improved the detection of breast cancer with an acceptable increase in the recall rate. In contrast, Fenton et al conducted a retrospective study at 43 facilities (429,345 mammograms) that included a pre/post comparison of CAD at 7 facilities (16%). (7) This study concluded that the use of CAD was associated with reduced accuracy of interpretation of screening mammograms and also noted that the increased rate of biopsy with CAD was not clearly associated with improved detection of invasive breast cancer. While these results are intriguing, questions have been raised about this study including the retrospective design, the pre/post analysis, limited time for radiologists to gain experience with CAD, and the high sensitivity of mammography reported in this study. This study also did not report details about how CAD was applied in the groups where it was used; that is, whether or not it was used after the radiologists’ interpretations as described in the studies by Freer, Dean, and Morton described here. While this 1 study could reflect the impact of CAD as it is disseminated into practice, this study is not sufficient to refute the findings of several prospective studies. Thus, for CAD with SFM, the policy statement is unchanged.

Full-field digital mammography (FFDM). None of the articles identified through the MEDLINE search provided data on use of CAD that adequately address the issues raised by the TEC Assessment. (2) Specifically, no large prospective studies were identified that report on the addition of CAD with FFDM in a typical population that would receive screening mammography. Brancato reported on a European study of 3425 women who were self-referred to a breast clinic. (8) In this study, addition of CAD (CAD-X Systems’ Second Look®) to FFDM increased the cancer detection rate from 107 to 112 cases (p=0.07) and increased the recall rate from 11.9% to 13.1%. However, for a study of screening mammography, this is a relatively small study. In addition, long-term follow-up was not conducted to assess overall accuracy.

Skaane reported on a study from Norway that compared SFM and FFDM using CAD or double reading. (9) While this study reports results on 3683 patients, it appears that CAD was only used in a limited subset of patients. Thus, while CAD was reported to increase the sensitivity for both techniques, since it was applied to only a subset of patients the overall impact in a screening population cannot be determined. Wei et al (University of Michigan) studied CAD with both film-screen and digital images in the same patients. (10) Masses were found in 131 cases (27 cancers), and 98 cases had no masses. Using their CAD system, which is being developed (not commercially available), they found similar levels of performance for CAD with both types of mammograms. These authors comment that further study is underway to collect a larger dataset and to improve the performances of both [CAD] systems. A U.S. Food and Drug Administration Radiological Devices Panel Meeting on CAD was held in March 2008. (11) Some of the questions addressed by this meeting related to use of CAD with FFDM; differences between FFDM devices were also discussed. Comments were made that it was difficult to know about the image processing links between CAD and FFDM. Thus, the policy statement on CAD in FFDM is unchanged.

2009 Update The policy was updated with a literature search using MEDLINE conducted in October 2009.

Digitized Film-Screen Mammography.

Two new meta-analyses of studies on CAD with SFM in a screening population were identified in the search. Taylor and Potts identified 10 studies that compared a single reading with CAD to a single reading alone. (12) The main outcomes were the cancer detection rate and the recall rate. The pooled analysis found a nonsignificant difference (odds ratio [OR], 1.04; 95% confidence interval [CI], 0.96 to 1.13; p=0.35). The main analysis combined matched studies (compared interpretation with and without CAD in the same mammograms) and unmatched studies (compared mammography outcomes in programs before and after the introduction of CAD). An analysis limited to the 6 matched studies also found a nonsignificant increase in the cancer detection rate with CAD (OR=1.09; 95% CI, 0.92 to 1.29). Seventeen studies comparing single reading without CAD to double reading were included. The pooled estimate of the impact of double reading on the cancer detection rate was statistically significant (OR=1.10; 95% CI, 1.06 to 1.14). A pooled analysis of the 10 studies on single reading with CAD found a statistically significant increase in the recall rate (OR=1.13; 95% CI, 1.05 to 1.23). There was strong evidence of heterogeneity in the analysis of the impact of CAD on recall rate so a random effects model was used. Findings were unchanged when the Fenton et al (7) study, which found an unusually high recall rate, was omitted. There was less heterogeneity in the double-reading studies for this outcome. In a pooled analysis of all of the double-reading studies, there was a statistically significant increase in the recall rate (OR=1.17; 95% CI, 1.15 to 1.18). However, in an analysis limited to the 8 studies that used arbitration/consensus in double reading, the recall rate was not significantly increased (OR=0.94; 95% CI, 0.92 to 0.96). The authors concluded that the evidence on the use of double reading to enhance SFM is stronger than for CAD; however, although double reading is common in Europe where the authors are based, it is rarely used in practice in the United States.

The other meta-analysis, by Noble et al, identified 7 studies on the diagnostic performance of CAD for screening mammography. (13) A pooled analysis of 3 studies (n=347,324 women) found a sensitivity of a single reader plus CAD of 86% (95% CI, 84.2% to 87.6%) and a pooled specificity of 88.2% (95% CI, 88.1% to 88.3%). Among women screened with a single reading with CAD compared to a single reading alone, the pooled analysis of 5 studies (n=51,162 women) found an incremental cancer detection rate of 50 (95% CI, 30 to 80) per 100,000 women screened. (Note that the 2 largest studies were excluded from this analysis). A pooled analysis of recall rate found a rate of 1190 (95% CI, 1090 to 1290) additional healthy women per 100,000 screened with CAD were recalled for further testing. Approximately 4.1% (95% CI, 2.7% to 6.3%) of the women who were recalled were subsequently diagnosed with cancer. The incremental biopsy rate of healthy women was 80 (95% CI, 60 to 110) per 100,000 screened.

One of the largest studies evaluating CAD with SFM was by Fenton et al, (7) discussed in the previous policy update. The current search identified another large retrospective study, by Gromet et al. (14) Both the Fenton and Gromet studies were included in the two meta-analyses discussed here. Gromet reported on a review of 231,221 screening mammograms performed at 10 facilities. Data were collected over a 6- year period when the centers were switching from double reading to a single reading with CAD (Image Checker CL). In this sample, 112,413 mammograms underwent double reading and 118,808 underwent single reading with CAD. Radiologists who had subspecialties in mammography conducted all first readings on the double readings and all single readings with CAD. Arbitration/consensus was not used for double reading—women were recalled if the first reader determined them to be positive regardless of the second reading, or if the second reader, a generalist, said they were positive and this was confirmed by a second subspecialist. False negatives included cancers identified by a second reader or diagnosed within 1 year of an overall negative screening interpretation. Positive mammograms were followed up to obtain pathology results to distinguish between true and false positives. The main findings are as follows:

The difference in sensitivity between CAD and double reading was not statistically significant. However, there was significantly higher sensitivity with a single reading with CAD compared to the first reader (p<0.001), and this analysis more closely resembles practice in the United States. There was a significantly lower recall rate with CAD compared with double reading without arbitration/consensus (p><0.0001). Differences in the cancer detection rate were not statistically significant. Gromet concluded that CAD enhances the performance of a single reader, increasing sensitivity with only a small increase in the recall rate. The author points out several strengths of this study compared to the Fenton et al study: (1) there were nearly 4 times the number of CAD cases; (2) the radiologists were more experienced at reading with CAD: the mean CAD experiences was 13,201 compared with 821 in the Fenton et al study; and (3) a direct data analysis was conducted rather than statistical modeling, which Fenton et al used. A limitation of the Gromet study was that it was a pre/post comparison and as such did not compare single reading with CAD to single reading without CAD in the same mammograms. Moreover; women in the CAD group were slightly younger than those in the non-CAD group, and more women in the CAD ><0.001), and this analysis more closely resembles practice in the United States. There was a significantly lower recall rate with CAD compared with double reading without arbitration/consensus (p<0.0001). Differences in the cancer detection rate were not statistically significant. Gromet concluded that CAD enhances the performance of a single reader, increasing sensitivity with only a small increase in the recall rate. The author points out several strengths of this study compared to the Fenton et al study: (1) there were nearly 4 times the number of CAD cases; (2) the radiologists were more experienced at reading with CAD: the mean CAD experiences was 13,201 compared with 821 in the Fenton et al study; and (3) a direct data analysis was conducted rather than statistical modeling, which Fenton et al used. A limitation of the Gromet study was that it was a pre/post comparison and as such did not compare single reading with CAD to single reading without CAD in the same mammograms. Moreover; women in the CAD group were slightly younger than those in the non-CAD group, and more women in the CAD ><0.0001), Differences in the cancer detection rate were not statistically significant. Gromet concluded that CAD enhances the performance of a single reader, increasing sensitivity with only a small increase in the recall rate. The author points out several strengths of this study compared to the Fenton et al study: (1) there were nearly 4 times the number of CAD cases; (2) the radiologists were more experienced at reading with CAD: the mean CAD experiences was 13,201 compared with 821 in the Fenton et al study; and (3) a direct data analysis was conducted rather than statistical modeling, which Fenton et al used.

A limitation of the Gromet study was that it was a pre/post comparison and as such did not compare single reading with CAD to single reading without CAD in the same mammograms. Moreover; women in the CAD group were slightly younger than those in the non-CAD group, and more women in the CAD group had a prior mammogram in the past year. These differences in patient characteristics could affect outcomes.

In summary, the new evidence affirms the policy statement that CAD as an adjunct to single-reader interpretation of digitized SFM is medically necessary. Although the Taylor and Potts meta-analysis did not find a statistically significant improvement in the cancer detection rate, the degree of improvement found in the Taylor meta-analysis, when balanced with the recall rate, is clinically reasonable. Moreover, the large Gromet study found significant improvement in the sensitivity with CAD used with single reading compared to a single reading alone.

Full-field digital mammography (FFDM).

The search did not identify any study that met all of the criteria for the ideal design; that is, large, prospective, reported diagnostic accuracy with FFDM with and without CAD, and conducted in a screening population. Van den Biggelaar et al included 1048 consecutive women who referred to the radiology department of a medical center in the Netherlands. (15) Only 219 (21%) were referred for screening mammography; the remaining 829 (79%) received diagnostic mammography. The addition of CAD (Second Look digital CAD system, iCAD Inc.). The reference standard was 12 months’ follow-up. A total of 51 malignancies were found in 50 patients. There were only 4 malignancies in the screening population; this study is too small to do an analysis of the impact of the CAD system. Overall, sensitivity rates did not change after application of the CAD software. There was a mean sensitivity of 92% for all 3 technologists and 84% for the single radiologist.

A large study by Karssemeijer et al included 367,000 screening mammograms, of which 56,518 were digital. (16) The FFDM images, but not the SFM images, were analyzed with a CAD system (ImageChecker R2/Hologic). Breast cancer was detected in 1927 women. At initial screening, the detection rate was 77% with FFDM plus CAD and 62% with SFM (p=0.11). The recall rate was significantly higher with FFDM plus CAD than with SFM (4.4% vs 2.3%, p<0.001). This study does not compare FFDM with and without CAD assistance and does not compare FFDM with CAD to digitized SFM with CAD, and thus its applicability to this policy is limited. Due to the paucity of data comparing FFDM without CAD to FFDM with CAD, the policy statement is unchanged. ><0.001). This study does not compare FFDM with and without CAD assistance and does not compare FFDM with CAD to digitized SFM with CAD, and thus its applicability to this policy is limited.

Due to the paucity of data comparing FFDM without CAD to FFDM with CAD, the policy statement is unchanged.

Population Reference No. 1 

The evidence affirms the policy statement that CAD as an adjunct to single-reader interpretation of digitized SFM is medically necessary. Although the Taylor and Potts meta-analysis did not find a statistically significant improvement in the cancer detection rate, the degree of improvement found in the Taylor meta-analysis, when balanced with the recall rate, is clinically reasonable. Moreover, the large Gromet study found significant improvement in the sensitivity with CAD used with single reading compared to a single reading alone.

Supplemental Information

N/A

Practice Guidelines and Position Statements

American College of Obstetricians and Gynecologists (ACOG, 2019) The ACOG recommends routine screening with use of digital mammography for women diagnosed with dense breasts. They do not recommend routine use of alternative or adjunctive tests to screening mammography in women with dense breasts who are asymptomatic and have no additional risk factors. The College strongly supports additional research to identify more effective screening methods that will enhance meaningful improvements in cancer outcomes for women with dense breasts and minimize false-positive screening results. ACOG also recommends that health care providers comply with state laws that may require disclosure to women of their breast density as recorded in a mammogram report.

Medicare National Coverage

Medicare covers screening and diagnostic mammography when criteria are met; see the National Coverage Determination (NCDs) for Mammograms (220.4) and the NCD for FDG PET for Breast Cancer (220.6.10). Medicare does not have NCDs for the following procedures: digital mammography, Breast Magnetic Resonance Imaging (MRI), Magnetic Resonance Elastography (MRE), Breast Specific Gamma Imaging (BSGI) (also known as scintimammography), Electrical Impedance Scanning (EIS), Computer Aided Detection (CAD) for MRI of the breast, breast ultrasound, Computer Aided Detection (CAD) for Ultrasound, computer aided tactile breast imaging and Automated Breast Ultrasound System (ABUS).

Local Coverage Determinations (LCDs) exist; see the following LCDs: Breast Imaging: Breast Echography (Sonography)/Breast MRI/Ductography, Breast Imaging Mammography/Breast Echography (Sonography)/Breast MRI/Ductography, 3D Interpretation and Reporting of Imaging Studies, Screening and Diagnostic Mammography, Noncovered Services, Non-Covered Category III CPT Codes and Services That Are Not Reasonable and Necessary. (Accessed February 28, 2019

References

1. 2002 TEC Assessment: Computer-aided Detection in Mammography.

2. 2006 TEC Assessment: Computer-aided Detection (CAD) with Full-Field Digital Mammography.

3. Pisano ED, Gatsonis CA, Yaffe MJ et al. American College of Radiology Imaging Network Digital Mammographic Imaging Screening Trial: objectives and methodology. Radiology 2005; 236(2):404- 12.

4. 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(17):1773-83.

5. Dean JC, Ilvento CC. Improved cancer detection using computer-aided detection with diagnostic and screening mammography: prospective study of 104 cancers. AJR Am J Roentgenol 2006; 187(1):20- 8.

6. Morton MJ, Whaley DH, Brandt KR et al. Screening mammograms: interpretation with computeraided detection--prospective evaluation. Radiology 2006; 239(2):375-83.

7. Fenton JJ, Taplin SH, Carney PA et al. Influence of computer-aided detection on performance of screening mammography. N Engl J Med 2007; 356(14):1399-409.

8. Brancato B, Houssami N, Francesca D et al. Does computer-aided detection (CAD) contribute to the performance of digital mammography in a self-referred population? Breast Cancer Res Treat 2007; Oct 16 (E-pub).

9. Skaane P, Kshirsagar A, Stapleton S et al. Effect of computer-aided detection on independent double reading of paired screen-film and full-field digital screening mammograms. AJR Am J Roentgenol 2007; 188(2):377-84.

10. Wei J, Hadjiiski LM, Sahiner B et al. Computer-aided detection systems for breast masses: comparison of performances on full-field digital mammograms and digitized screen-film mammograms. Acad Radiol 2007; 14(6):659-69.

11. http://www.fda.gov/ohrms/dockets/ac/08/briefing/2008-4349b1- 01%20FDA%20Radiological%20Devices%20Panel%20Meeting%20Introd.pdf

12. Taylor P, Potts HW. Computer aids and human second reading as interventions in screening mammography: Two systematic reviews to compare effects on cancer detection and recall rate. Eur J Cancer 2008; 44(6):798-807.

13. Noble M, Bruening W, Uhl S et al. Computer-aided detection mammography for breast cancer screening: systematic review and meta-analysis. Arch Gynecol Obstet 2009; 279(6):881-90.

14. Gromet M. Comparison of computer-aided detection to double reading of screening mammograms: Review of 231,221 mammograms. AJR Am J Roentgenol 2008; 190(4):854-9.

15. van den Biggelaar FJ, Kessels AG, van Engelshoven JM et al. Computer-aided detection in full-field digital mammography in a clinical population: performance of radiologist and technologists. Breast Cancer Res Treat 2009 May 6 [Epub ahead of print].

16. Karssemeijer N, Bluekens AM, Beijerinck D et al. Breast cancer screening results 5 years after introduction of digital mammography in a population-based screening program. Radiology 2009; 253(2): 353-8.

Codes

Appplicable Modifiers

N/A

Policy History