J Thorac Cardiovasc Surg 2001;121:601
© 2001 The American Association for Thoracic Surgery
Transmyocardial laser channeling, coronary artery bypass, statistical analyses, and their interpretations
Michael Nathanson, MD, PhD,
Kai Ihnken, MD
Santa Clara Valley Medical Center
751 S Bascom Ave
San Jose, CA 95128
To the Editor:
Statistical analyses of experiments designed to prove significant differences between treatment modalities may yield uninterpretable results if the experimental design is inadequate and data are insufficient. In comparing the treatment effects of coronary artery bypass grafting (CABG) only with CABG plus transmyocardial laser revascularization (TMR), Allen and coworkers 1 found no long-term significant differences between the two. In this "multicenter, blinded, prospective, randomized, controlled trial," at 12 months after the operation there was equality in freedom from mortality and myocardial infarction (P = .09), residual class III/IV angina (P = .11), as well as in improvement in exercise treadmill testing (P = .9). At odds with these data is the finding in a multivariate analysis that CABG alone was a predictor of death, myocardial infarction, and residual class III/IV angina (odds ratio 2.2; P = .03). Equally puzzling is a glaring disparity in operative mortality (CABG 7.6%, CABG/TMR 1.5%; P = .02), which adversely affected the overall mortality rates at 12 months postoperatively. The article does not provide analysis of differences, if any, between survivors and nonsurvivors. The two treatment groups had similar numbers of grafts per patient: CABG alone 3.4, CABG/TMR 3.1 (P = .07) (How many arterial?). Both had a similar frequency of bypassed main arteries (left anterior descending 90%, right coronary artery 75%, circumflex 60%), so, obviously more than one graft per system was constructed in some patients. In others, most likely, TMR was carried out in grafted territories (systems). The article does not provide an inventory of coronary vessels bypassed versus systems bypassed vis-à-vis the sites of laser channeling and relating this information to frequency of sentinel events at 1 year after the operation. Specifically, was there a statistically significant outcome difference (in either group) between patients who had all three major arteries bypassed and those who did not? Was the addition of TMR to triple bypass (left anterior descending/right coronary/circumflex) significantly meaningful and did it have the same effect if added to a triple CABG that did not include one of the three major coronary arteries, particularly the left anterior descending? Lack of angiographic inventory 1 year after the operation further limits the scope of interpretation of results, since the impact of stenosed or occluded grafts (as well as progression of native vessel disease) was not sorted out. These issues are important in that in reviewing the literature to date, no evidence was found to support a notion that more than one bypass graft per system (the "benchmark" triple CABG) does improve long-term freedom from key sentinel events.
Finally, and not in the least, is the issue of study design involving 24 centers. Each participating institution had patient populations widely varying from the mean (1-39, mean = 11, SD = 10.2), which makes it hard to compare treatment effects between and within each and every institution. By contrast, in another study 2 the authors assigned a similar number of patients to each treatment group and to each of the institutions. Data simply pooled from a disparate group of patient cohorts cannot be statistically meaningful, because replicated experiments at different sites must be analyzed by treatments, sites (institutions), the confounding effect of treatments x institutions, and by calculating the effects of pooled experimental errors. 3 Although the primary goal of this multicenter study was to compare two treatment effects, the impact of the individual institutions was not brought out (notwithstanding the authors' claim to negative institutional bias), and the corroborative value of such a design element is missing in this study. The role of TMR channeling as an adjunct to genuine or perceived incomplete revascularization or to a complete one (one graft per system) deserves closer scrutiny. We therefore agree with Allen's observation that a larger (perhaps much larger) validation study is required. Additionally, proper patient stratification as part of the experimental design elements and more rigorous statistical analysis will contribute to our understanding of the contribution of TMR to the surgical armamentarium.
Allen KD, Dowling RD, Del Rossi AJ, Realyvasques F, Lefrak EA, Pfeffer TA, et al. Transmyocardial laser revascularization combined with coronary artery bypass grafting: a multicenter, blinded, prospective, randomized, controlled trial. J Thorac Cardiovasc Surg 2000;119:540-9.[Abstract/Free Full Text]
Allen KD, Dowling RD, Fudge TL, Schoettle GP, Selinger SL, Gangahar DM, et al. Comparison of transmyocardial revascularization with medical therapy in patients with refractory angina. N Engl J Med 1999;341:1029-36.[Medline]
Cochran WG, Fox GM. Experimental designs. New York: John Wiley; 1957. p. 545-68.