ROGER HERDMAN, M.D., Director, National Cancer Policy Board, Institute of Medicine: I want to compliment all the speakers for a wonderful job. They have obviously prepared very carefully, and their talks were uniformly of high quality and relevant to the subject matter of this report. Our last tasks in this wrap-up session are to ask Dr. Pisano and Dr. Bohmer to review with us the bottom lines of the group discussions this afternoon, the implications for the recommendations and findings of the report, and any other relevant comments that they would like to make. Then we are very privileged to have Carolina Hinestrosa here to represent the perspective of survivors; we give the survivors the last word.
DR. PISANO: Dr. Bohmer and I are going to briefly summarize the talks in each of our group discussions because all of us at this symposium could attend only one of the concurrent sessions. Then we will try to come to some conclusions.
In our session on delivering services, we heard first from Dr. Taplin. He spoke about the organization of screening. He presented a system of improved screening services in which he had personally participated that affected both providers and patients. As an example of affecting women’s behavior, he reported that women who were surveyed and discovered to have an increased risk of breast cancer because of family history were more likely to return for recommended mammograms if they were informed of their elevated risk. In the population that he studied, 52 percent of late stage breast cancers were in women who had never been screened. The multidisciplinary approach was very important to his model in which the radiologists, pathologists, surgeons, primary care physicians, and oncologists sat together and designed a system of care. Dr. Taplin’s presentation raised the subject of motivating practitioners to perform within these sorts of models and how to pay for them. Dr. Tunis’s comments that Medicare was interested in addressing this payment question were encouraging.
The next speaker was Dr. Smith-Bindman. She spoke about access to, and accuracy of, screening mammography and provided data comparing the U.K. to the U.S. system. She pointed out that self referrals for screening mammography probably do not reflect reality; non-whites seemed to be getting regular screening mammograms at the same rate as self referrals, but in fact, they probably are not if you look at population registries. She observed that even though the rate of recall for mammography is about 15 percent in the U.S. versus about 7 percent in Britain, it is quite variable here in different regions of the country. Actual tissue sampling is remarkably similar at about two to three percent of all women screened. In addition, in the U.S. there is a two to three time’s higher rate of negative open biopsies than in the U.K.
Next we heard about the U.S. Food and Drug Administration’s (FDA’s) implementation of the Mammography Quality Standards Act from Dr. Finder. He presented the requirements. He mentioned that mammography facilities have indeed decreased since 2000 from 10,000 to 9,100, approximately 900 fewer facilities in a four year period. But the number of available x-ray units has actually increased slightly because of the increase in average numbers of units per facility from 1.2 to 1.5. He also described how mammography facilities have improved on MQSA inspections. Now about 70 percent of facilities do not get a violation of any type, whereas when Mammography Quality Standards Act (MQSA) first started it was about 30 percent.
Dr. Borgstede of the American College of Radiology described some practical considerations regarding improvement of mammography screening services. He referenced the 2002 U.S. General Accounting Office (GAO) study that reported that mammography capacity was adequate at that time. Because the aging of our population is increasing demand for mammography, there is concern that that report may already be out of date. In addition, he cited a study about difficulties of improving both accuracy and access. Improving accuracy by not certifying as interpreting physicians those radiologists who do not provide a good standard of accuracy will result in a reduction in manpower sufficient to impair patient access (Beam et al., 2003).
He spent quite a bit of time on our recommendation about prescreening, observing that reimbursement may suffer if we implement the recommendation. It is ironic that a second radiologist reading a mammogram is paid nothing, but a computer doing computer assisted diagonosis (CAD) generates a payment. We would hope that a trained technologist who provided a second set of eyes to read mammograms would get an additional fee. He also questioned how radiologists would supervise breast imaging technologists and posed questions that, as a practical matter, have to be answered if this recommendation is to be implemented, although in theory it is implementable right now. Current FDA regulation allows for a non-radiologist second reader. The committee envisioned extending capacity with specially trained technologists the way we do now through the use of residents and highly-qualified fellows.
Dr. Borgstede next noted that recent insurance industry data identify missed breast cancer as the number one reason for malpractice litigation in the U.S., and second only to birth-damaged infants as the most expensive condition. He concluded by providing data on mammography reimbursement, showing there is a difference between reimbursement and actual costs, especially for hospitals, less so for offices, but still negative for both.
During the discussion period, we talked about approaching Medicare for changes in reimbursement, paying for new technologies within protocols, and funding demonstration projects for some of our recommendations.
DR. BOHMER: I will report our presentations on developing and delivering new technologies out of order and begin with Dr. Vosburgh. He reminded us that mammography’s sensitivity and specificity were still good. It would take quite a substantial trial spanning many years to give us sufficient evidence to justify completely replacing mammography. In the short term, therefore, new technologies are likely to be complements, not substitutes, for the existing technology. Some of the more promising advances are likely to come from the use of contrast agents or collaboration between radiologists and biologists.
We heard from Drs. Baugh and Flamm from Blue Cross and Blue Shield. They noted that they faced several kinds of insurance coverage decisions. The easy decisions were in situations in which a technology is clearly a substantial advantage for patients in all respects—then there is no difficulty in agreeing to cover it—or those technologies that add no benefit whatsoever and will not be covered. Technologies that in some respects are the same as, or better than, existing technologies are the ones that require a well defined process for medical assessment before a coverage decision can be made. Ideally, it would be nice to have a trial that compared a new technology with the incumbent technology in all respects, but often such trials do not exist. Dr. Flamm pointed out that in this situation insurers have to rely on indirect evidence, for example, (a) comparisons of a new technology’s sensitivity and specificity with data on existing modalities; (b) how the performance of the new technology is likely to affect clinical decision making and, therefore; (c) the possible effect on patient outcomes. Thus, making a coverage decision in this situation can involve coordinating three sets of data.
Once the medical assessment is done, the Blues then engage in a rigorous determination of whether or not the technology should be covered, and how. At that point there are several possible strategies. The first is to cover the new technology and then apply various utilization management techniques, claims edits, or retrospective review; these need to be applied conservatively because they are burdensome and annoying to practitioners. In other circumstances, the Blues have decided to fund the clinical costs of trials to answer questions about a technology with a view to making a definitive decision at a later date.
Dr. Petitti gave us a view of a similar sort of analysis from Kaiser Permanente. This organization has the benefit of combining both the payor and the
provider function and can take a more systemic view of a technology than the Blues. She described Kaiser’s recent evaluation of CAD, which resulted in a decision not to deploy CAD. The analysis examined the merits of CAD versus the benefits of alternative use of the resources for the Kaiser population.
Such benefits might include increasing breast cancer screening rates from 80 to 93 percent, reorganizing services to improve the performance of radiology (as was done in a case study from Colorado) or improving the performance of individual radiologists. Based on the available evidence, Kaiser concluded that CAD was no better than an experienced radiologist, so they would be better off training and deploying experienced radiologists than funding CAD. This context-based decision took account of system-wide concerns about the available resources within Kaiser, the kind of processes that would be more or less applicable within Kaiser, and the kinds of organizational changes that Kaiser could make. A different organization might have come to a different conclusion on reviewing the same evidence. It stands in contrast to large insurers, such as Blue Cross Blue Shield, which cannot take such local contextual factors into account in a decision about whether or not to adopt a new technology.
Our first presenter, Dr. Sullivan addressed the question of where the evidence that allowed a clearer comparison of one technology with another might come from. He cited three possibilities. First, the National Institutes of Health (NIH) is beginning to promote multidisciplinary collaboration in the conduct of trials (for example, by having more than one principal investigator on a study) so as to improve collaboration across disciplines. Second, he discussed interagency collaboration. For example, the DMIST trial involved NIH, Centers for Medicare and Medicaid Services (CMS), and FDA in the design so that the data needs of each of those three agencies would be met by this trial which compared digital with screen-film mammography. The concerns of CMS about the practical application of virtual (that is, CT) colonoscopy explain why the current National Cancer Institute (NCI) trial was designed with power to test for inter-site variability. Its aim was to address the effectiveness concerns that organizations like Kaiser worry about when deciding to adopt a new technology.
Finally, Dr. Sullivan talked about further research on new technology adoption. He described four phases in technology assessment—discovery, development, maturation, and dissemination. He pointed out that the ways to use a new screening technology are still poorly understood. Dr. Sullivan compared the problem of signal-to-noise ratios in the fields of radiology and air traffic control. In radiology, the signal is buried in other valid signals, but in air traffic control, the signal (an airplane) is buried in white noise, an entirely different problem. Hence, lessons from the airline industry may not be relevant to the problems of breast image interpretation.
At the end, the group discussed funding the kinds of trials called for in the report. We thought that funding ought to be fairly shared among payors. All payors as well as society at large realize significant benefits from effective-
ness/comparison-oriented trials. For example, an early trial of autologous bone marrow transplantation for metastatic cancer might have shown this to be a treatment not worth pursuing well in advance of the time that became understood by physicians and the public. We also observed that there is a big difference between efficacy and effectiveness, a difference not fully understood. It is an area into which many of the problems that we were convened to comment on fall. We are beginning to talk about very specific interventions to try and shrink that difference.
Final Remarks Carolina Hinestrosa, M.A., M.P.H., Executive Vice President for Programs and Planning National Breast Cancer Coalition
The organization that I work with, the National Breast Cancer Coalition, is committed to ending breast cancer through action and advocacy. The Coalition and I think it is important to look at the report’s recommendations and the problem of breast cancer from the consumer perspective.
There are many recommendations in this report, but it seems that the media have focused on the recommendations on better use of mammography. We are concerned that we are yet again reducing the approaches to breast cancer to the mammography question. As we all know, and the report says, the committee that preceded the current one, and that I was a member of, looked at mammography and concluded that it is an imperfect tool. It is the tool that we have, but it is imperfect. It is not going to solve the breast cancer problem. We are hoping that from the work of this committee, we will see more of a push to address breast cancer in a fundamental way. We believe that yes, we need to be able to find breast cancer early, but we want to know what early means and what implications that has for a patient.
It was very interesting to hear Dr. Fletcher’s presentation addressing the recommendation to target women who are at high risk for breast cancer. It underscored some of the concerns I have, for example, how do we go from models of risk to telling a woman what her individual risk is? Are we going to be able to get to that point, and how useful will that be? We know that our knowledge of what causes breast cancer, what puts us at risk for breast cancer, is very limited right now. Perhaps this is one of the reasons risk stratification is so imperfect.
Furthermore, as we try to identify women and stratify them in accordance to risk, we must face a health care system that we already know is not accessible to everyone. If women are classified as high risk, will that lead to discrimination in employment or for health insurance?
Regarding, the crisis in access to mammography screening—which the previous committee also heard about—we still see that the information we have comes in bits and pieces, and what we need is a thorough and rigorous study. I urge us to be careful. I am not convinced that we have a nationwide crisis. So we welcome studying this issue, looking at all the aspects of reimbursement, of people entering the profession, of centers opening and closing, and of numbers of units, because we need to drive our policy decisions on evidence, and we need to have clear evidence of a problem.
We strongly support the committee’s recommendations to look at the system of screening and move away from the ad hoc system we have, where depending on who you are, depending on whether you have insurance or not, or depending on where you live, you get screened or you don’t get screened. I think that the recommendations to look at quality improvement overall, look at screening as a system, look at benchmarking, are important.
We are moving in exciting directions in breast cancer research. We are hoping, as we share the goal of saving women’s lives and reducing mortality from this disease, that we put our best efforts toward going in the directions pointed out by new research findings. We have been focusing too much on a technology like mammography that has too many limitations, and we need to put our energy and resources into moving forward. We hope we can improve mammography, but let’s be clear that the major impact in breast cancer is still to be made, and it is not all about mammography.
So I thank you for listening to this perspective. I want to remind you that we are expecting a lot more from you, and we will continue to push you. We hope that in this dialogue we get to our objective—to have a real impact on breast cancer.
DR. HERDMAN: I think some of our panelists and committee members are prepared to answer questions if there are any from the audience.
DR. SULLIVAN: I want to comment on benchmarking that was just mentioned in the last presentation because we need to emphasize the importance of having sensitivity data available for the population that is being looked at. I think that was not emphasized enough in the report, at least in my quick read. I think the Colorado example is useful, because I suspect they have a population registry in Colorado (Colorado is a member of the NCI Breast Cancer Screening Consortium and has a mammography registry) that gives sensitivity information, but most of what we have for benchmarking relates to specificity, that is, false positives. It is not clear from looking at changes in specificity if the radiologist’s sensitivity is staying the same; actual sensitivity data are needed. Cancer yield information is a surrogate, and is helpful, but it is an imperfect surrogate.
I mention this because we know that the next Institute of Medicine report will be looking to inform the 2005 reauthorization of the MQSA, and I am sure it will come up in that discussion. It is not obligatory that an increase in specificity (fewer false positives) will come at the cost of a decrease in sensitivity for a
given radiologist, but without actually measuring and knowing sensitivity, you don’t know what is happening, whether a radiologist is moving to a different ROC curve.
DR. NORTON: I think it is important to identify that reports like this are done to initiate conversation, not end conversation. I think that was the intention of this committee. I wasn’t on the committee, I was an advisor to the committee, and I tried very much to stay on the periphery except when asked for advice on this. But I think it is a very hard-hitting report, and I think it is a shame that some of the public perception of it mis-identifies some of the points that were made.
Mammography clearly saves lives. I don’t think we need any more studies to show that. It does not, and will not, save all lives; it is a technology that has its limitations, but a very careful review of the data shows that it has specificity and sensitivity that is very useful. If more people had access to it and it was organized better more lives would be saved and more people would take advantage of it.
There are countries that are organizing screening mammography services in an entirely different way with an impact on survival statistics and cure rates from breast cancer. We saw this morning very impressive data showing that a big difference can be made. We are all working hard on the molecular etiology and progression of breast cancer, and that is where I am focusing a lot of my energies, probably 80 percent of my research, but to throw the baby out with the bath water and say that mammography has not answered all the questions and so we should just get rid of it—I think one of the important observations in this report is that it is going to be very hard to do the trials with the resources that are, or could be, available to replace mammography.
We are all hoping that a better test will emerge and mammograms can be eliminated, but this group of experts finds that that may be extremely difficult or impossible without dramatic and unexpected changes in technology. The real advances are going to come from integrating other approaches with mammography, such as molecular diagnostics, risk assessment, and allocating patients to proper screening on that basis.
If what you have is imperfect, you have to decide whether to throw it out and replace it or to build on it. I think the conclusion is that it is something we can build on, improve its quality, improve its availability, and build on it through increasing knowledge of cancer. As far as I am concerned, it is one of the core messages of this report.
DR. HERDMAN: That was well said, Dr. Norton. We have an event like today’s symposium because no report can say everything. There are different ways of looking at breast cancer research and care and different emphases that can be assigned. That is why we will report this symposium and make every effort to distribute it as widely as the original report. We consider it an important