Panel Discussion77sure �everything� and then come up with a hypothesis to test. That e, but is terribly expensive, and it very hard to get clinicians and academicians to agree to that because they will have their own thesis of what heart failure is and how to treat it and they like to lecture as opposed to doing your study. I think it is probably better to look at speci populations. Outside of congestive myopathy versus ischemic, which are obvious subgroups, additional subgrouping is very hard to determine. I talked of sudden death, but sudden death is different in people who have myopathic disease versus atherosclerotic disease, with patchy ischemic areas causing a reentrant arrhythmia and death. These latter patients may have calcium overload and they have EADs or DADs and triggered atomanticity, and different drugs will affect these things differently. Until you have some data to warrant your looking at target population or subpopulation, how to stratify is problematic. There are many possibilities. I might suggest that in addition to ischemic versus nonischemic populations, there are some patients who are relatively well adjusted to their very low ejection fraction; they are a totally different population than people who are markedly symptomatic with minimally elevated ling pressures. These, in turn, from a very different population from patients post-MI who have rales and who have arrhythmias and have another ischemic event triggering worse heart failure. A lot of patients may have ischemic events triggering heart failure. Patients with QT dispersion may be a group to separate out as well as those with heightened autonomic activity despite ACE inhibition therapy. The latter group may bene especially from beta-blockers. Dr. Lipicky: The intent of your question, Dr. Borer, as I understood it was that the trick is to have something that is easily identible, that has predictive value, that is measurable, and that clinicians can do whenever they want to. On that basis, if one identis a subgroup that one thinks responds differently, and randomizes that subgroup for clinical trials, indeed, that would be a reasonable approach because all morbidity/mortality trials should be analyzed by intent to treat. If one takes all comers into the trial, and
