To Cover or Not to Cover

The day of decision is coming. The Centers for Medicare and Medicaid Services (CMS) will decide soon whether to go along with the recommendation of the independent review panel, MEDCAC, against coverage of lung cancer screening using low dose CT-scans, or to bow to political pressure  to cover the test (see my post on June 23, Politicians: Keep Out). A spate of recent commentaries in medical journals weigh in on the debate. Most seem to support coverage, although they add a few caveats. Two of these caveats are intriguing: using a registry to actually generate data about what happens to older patients who are screened with lung CT scans and mandating the use of a formal shared decision making process.

Writing in a recent issue of JAMA, Harold Sox, a distinguished physician with a long career promoting evidence-based medicine, recognizes the weaknesses of the available data, principally the National Lung Cancer Screening Trial (NLST). The NLST simply was not representative of the Medicare population because it included only a small number of patients over age 65 and only a very small number over 70. Moreover, the patients followed—who had a lower risk of death if they underwent the screening with CT scans compared to screening with regular xrays—in all likelihood did not have multiple other chronic conditions, the norm for Medicare patients. But Sox is clearly worried about depriving those older patients who might benefit from screening if Medicare categorically refuses to cover the test. He therefore veers towards recommending coverage, but with the proviso that a registry be maintained that tracks all patients who are screened to see what happens to them. In principle, this information could be used to modify the rules about coverage later on. Taking away coverage for something that was previously offered is certainly possible—coverage of bone marrow transplants for breast cancer was rescinded after it was found to be both tremendously toxic and no more beneficial than conventional, less risky treatment—but it happens rarely because of pressure from patient advocacy groups, regardless of the data.

A second opinion piece in the same issue of JAMA argues that CMS should provide coverage but should address MEDCAC’s concerns by mandating “shared decision-making.” Shared decision-making rests on patients’ understanding the risks and benefits of the various options and deciding, in concert with their physician, which approach is most consistent with their own values and preferences. Now I am a fan of shared decision-making. And I support the use of decision aids, as proposed by the authors of this article, which are formal tools that systematically describe the risks and benefits of the alternatives in clear, readily understandable language illustrated by simple graphics. But for shared decision making to make sense, we have to know the risks and benefits of the alternatives. The reason that patients should have a choice to make is that some patients may be willing to accept one risk (say losing their independence) in exchange for a particular benefit (say a small chance of extending their life) while other patients might not. In the case of CT-scanning for lung cancer, the problem is that we simply don’t know what the risks and benefits are in older patients in general and in older patients with multiple co-morbidities in particular. So it’s just not possible to design clear and accurate decision aids to help patients think about screening for lung cancer.

The real issue is whether and under what circumstances it is appropriate to extrapolate from situations where we do have data to situations where we don’t. Sometimes there are plenty of good reasons for making such a leap. It seems reasonable to provide a treatment to a mature 17-year-old, for example, even though he is technically a child, and even though studies of the treatment were done in patients over age 18. After all, physiologically and socially, the 17-year-old is almost identical to the 18-year-old. Sometimes what initially seems reasonable proves not to be: it turns out that the data on first generation left ventricular assist devices (a high tech device that is essentially a partial artificial heart) could not be extrapolated from men to women because the devices were often too big to fit into a woman’s chest. Sometimes, however, there is good reason to believe that data obtained from one group cannot be extrapolated to another. There is good a priori reason to believe that neither safety nor efficacy data will be the same in old, sick patients as in young, healthy patients. It is eminently plausible to believe that the 75 year old with heart disease and diabetes might run into problems when undergoing open chest surgery to remove a lung cancer. The only way to know is to do a study involving older patients: there aren't any good shortcuts. 

Agreeing to pay for the test would no doubt be politically expedient. It would not be a wise decision.