Evidenced-based policy-making goes beyond numbers to consider "subjective factors that have to do with things other than evidence," said Tunis. "We shouldn't put ourselves in the box of thinking that somehow, either clinical decisions or policy decisions are regimented and strictly translated into policy." This view, according to Tunis, is consistent with a 1992 manifesto for practicing evidence-based medicine, which "deemphasizes intuition, unsystematic clinical experience, and pathophysiologic rationale as sufficient grounds for clinical decision making and stresses the examination of evidence from clinical research."2

We shouldn't put ourselves in the box of thinking that somehow, either clinical decisions or policy decisions are regimented and strictly translated into policy.

Defined this way, Tunis said, evidenced-based medicine does not entirely represent a direct translation of data into practice. "None of you clinicians do that as part of your clinical practice. You use scientific evidence matched with your judgments and experience to make clinical decisions that are appropriate for your patients." Evidenced-based policy making is the same way, he said.

"In an ideal world, clinical decisions would be made by informed patients in consultation with informed clinicians," he said. The economic impact on third-party payers, providers, and patients "would have nothing to do with it. . . . That's the perfect world," Tunis said. He then displayed a variegated map of the US. "This is the real world."

The map illustrated the population-adjusted frequency of PCI procedures throughout the country, exposing widespread regional variation by multiples of two or three. "This suggests there is a potentially huge amount of underuse, huge amount of overuse, and certainly inefficient use of important medical technologies, and that is a situation that should concern us all," Tunis said.

There are clearly opportunities for better clinical benefit with the same amount of resources.

"There are clearly opportunities for better clinical benefit with the same amount of resources," he said. "The clinical community really needs to take this seriously. Because if the clinical community doesn't act to get this addressed, then other actors in the system will do it."

The economic implications of clinical practice should be part of a national dialogue, Tunis said. "Medical professionals have an obligation to the medical profession. Industry is interested in patient care and also has an obligation to their shareholders. Medicare and other payers are interested in patient care and also have an obligation to the taxpayers and the trust fund. So we all come at this from different perspectives, and we're not going to get very far if we point our fingers at each other," he said. "Resource constraints are a shared challenge."

We're not going to get very far if we point our fingers at each other.

Much of the controversy, however, has focused on the appropriateness of the putative risk-stratifier that the CMS added to its reimbursement criteria. At the HRS sessions, Tunis explained why the QRS restriction was added: "At the time, MADIT II was the single and only trial looking at the efficacy of the ICD in this population. There were questions about possible selection bias. . . . How representative was the population of the universe of patients who might meet MADIT-II criteria?"

There were questions about possible selection bias.

Also, "part of the rationale was that in six months there was going to be another very large trial coming out, which had not an identical population but one with some similarity in terms of prophylactic use of defibrillators."

CMS agreed to reconsider its reimbursement decision as soon as those data were released, Tunis said. "And that's in fact what has happened with SCD-HeFT. Although we don't have the complete data, we've certainly begun meeting with the investigators and biostatisticians. . . . So we've started that process as we promised we would."

And then there are data that can be interpreted as supporting the QRS restriction, he noted. Subgroup analyses from a range of ICD trials all show "directionally similar" trends suggesting the benefit of ICD therapy in patients with ischemic HF is concentrated among those with wide QRS intervals, Tunis said. The trials he cited included not only MADIT II, but also COMPANION, DEFINITE, and SCD-HeFT.3,4

This is a subgroup of a subgroup of a subgroup in SCD-HeFT, very hard to know what that means.

Tunis cited one analysis of SCD-HeFT patients who met MADIT-II entry criteria that showed an ICD-related mortality hazard ratio of 0.65 for wide-QRS patients and 0.92 for narrow-QRS patients.

"This is a subgroup of a subgroup of a subgroup in SCD-HeFT, very hard to know what that means," Tunis said. "But it's interesting to find that essentially, no benefit was demonstrated in SCD-HeFT in the [narrow-QRS] MADIT-II-like patients that were enrolled."

Dr Jeanne E Poole (Source: University of Washington)

The day after Tunis made these remarks, however, SCD-HeFT investigator Dr Jeanne E Poole (University of Washington, Seattle) took center stage at the late-breaking trials session to question the validity of the very same QRS-based SCD-HeFT interpretation. She showed how the trial's numbers can be crunched so as to favor an ICD benefit in either wide- or narrow-QRS patientsbut not the otherdepending on minor changes to the cut points used to separate patients according to QRS interval [see the May 24, 2004heartwire story ].

For his part, Tunis had acknowledged that the post-hoc MADIT-II finding of greater benefit in long-QRS patients was merely "hypothesis generating." Still, it was worth considering, he said, expressing frustration with the view that "subgroup analyses are never reliable, statistically invalid, you can't do anything with the informationI just reject that interpretation."

Subgroup analyses are never reliable, statistically invalid, you can't do anything with the informationI just reject that interpretation.

With such analyses, Tunis said, "How much confidence you place in the finding is based on what you believed before you saw the finding." The greater the result's "biologic and pathophysiological plausibility," the more consideration it deserves. "This is why prespecified subgroup analyses are more meaningful than post-hoc subgroup analyses. . . . The QRS data in MADIT II were [from] a post-hoc analysis, and therefore relatively low confidence can be placed in them. Not no confidence, but relatively low," Tunis said.

Because of such inherent limitations of retrospective subgroup analyses, he went on, "you should consider formally evaluating the question in a specifically designed trial. That trial would probably be, if someone would do it, narrow-QRS patients randomly assigned to ICD vs placebo." In the absence of such a trial, argued Tunis, retrospective subgroup analyses are all there are to clarify the risk stratification significance of the QRS interval in ICD candidates.

But such analyses presented at the HRS sessions typically argued against the CMS official's case for using QRS duration to select patients for ICDs.

For example, one study found that a QRS interval >120 ms, as compared with <120 ms, was indeed associated with a significantly greater risk of appropriate ICD shock delivery over about three years among 352 patients with CAD and LV dysfunction.

Dr Lynda E Rosenfeld

"However, the shock rate was high even in patients with a narrow QRS," said Dr Lynda E Rosenfeld (Yale University, New Haven, CT) in her presentation. In multivariate analysis, QRS duration was not significantly related to VT/VF inducibility in prior EP studies but was associated with history of sustained VT and baseline degree of LV dysfunction and beta-blocker use.

In a subgroup of 106 patients without prior clinical arrhythmias, the rate of shock delivery for VT/VF events was about 53% for those with longer QRS intervals and 25% for those with shorter intervals (p<0.05). Thus, even short-QRS patients receiving ICDs for primary prevention were at significant risk of events, Rosenfeld said, as the negative predictive value of a QRS interval <120 ms was a mere 75%.

Dr Douglas L Packer

Elsewhere at the HRS sessions, Dr Douglas L Packer (Mayo Clinic, Rochester, MN) reported a similar post-hoc finding for QRS-related risk among 1232 MADIT-II-like patients enrolled in the MUSTT randomized trial and registry.5 The analysis corroborated MADIT II by finding an ICD-associated 70% reduction in relative risk for sudden death and 67% relative risk reduction for all-cause mortality over five years.

Among ICD recipients alone, those with QRS intervals >120 ms were at higher risk than patients with intervals <120 ms, who still had substantial event rates. Moreover, patients who did not receive ICDs were at substantially higher risk than those who did.

"ICD therapy appears to be highly effective in reducing arrhythmic death, cardiac arrest, and total mortality in patients with CAD and an ejection fraction <30%, Packer said. "All this remains true for patients with short QRS durations. It really is difficult to see how that particular indicator isolates patients who have high risk vs those who don't."

It really is difficult to see how that particular indicator isolates patients who have high risk vs those who don't.

Small wonder, given such data, that many clinicians see the QRS qualifier to Medicare reimbursement as misguided, if not out of line.

"It's not really within the FDA and CMS mandate to tell physicians how to practice medicine," said Dr Gerald V Naccarelli (Pennsylvania State University, Hershey). If CMS policies lead to inappropriately low reimbursement for ICD therapy, he noted, hospitals and clinicians will be pressured to underprescribe the devices or use inappropriate less-expensive models in patients.

Revamping the government process for establishing reimbursement policies would solve at least some of the problem, he suggested. For example, the delay between released data and a CMS ruling on reimbursement "can take anywhere from six to nine to 12 or more months," Naccarelli said. Shortening these delays would help alleviate the burden they impose on patient access to appropriate care. It would also limit the liability physicians potentially assume if lack of reimbursement somehow prevents clinically recommended treatment, he noted.

"This is the quandary physicians find themselves in. From my perspective as a physician, the shorter the time delay is once the data are out, the less of an issue this becomes," Naccarelli said.

Part of the problem: one federal hand doesn't know what the other is doing. There is a "disconnect" between government regulatory and reimbursement agencies that leads to diverging, often inconsistent sets of rules for physicians, according to Naccarelli. "When a device receives FDA approval and commercial release, coding and reimbursement rulings need to be set from the first day."

Finally, he noted, there is something clinicians can do. "The house of medicine has done a poor job of educating the CMS, private payers, and patients regarding the cost efficacy of ICDs," Naccarelli noted. "Since sudden death isn't sexually transmitted, it doesn't make the front pages," so the medical community should take it upon itself to publicize the technology's cost-effectiveness.

Beyond the QRS: Clinical criteria to target ICDs? Dr Arthur J Moss (Source: University of Rochester, NY) Speaking recently on theheart.org's feature The EP Show, Dr Arthur J Moss (University of Rochester, NY) said that a yet-unreleased analysis of patients in MADIT II disclosed a series of clinical features that identifies a very-low-risk group in whom ICDs are unlikely to prolong survival. The patients with an LVEF >25% who were in sinus rhythm and free of diabetes and were in NYHA class 1-2 "had a very good riskthat is, at three or four years they had only a 1% or 2% mortality," Moss said. This group, "which made up 20% of the population, achieved no benefit whatsoever from the ICD." He added, "On the other hand, each factor of LVEF <25%, atrial fibrillation, diabetes, and NYHA class 3-4 "made a significant and independent contribution to risk. . . . The benefit from the ICD was entirely in those patients who carried one, two, or three or more risk factors."