The effects of viability status and treatment allocation on long-term mortality in patients with left ventricular dysfunction and coronary artery disease have not been determined. Several observational studies with significant limitations have addressed this issue, and a recent meta-analysis has attempted to combine these results to increase statistical power. However, the analysis did not test for an interaction between viability status and treatment type, and included extraneous studies. We provide an alternate meta-analysis of this primary literature, utilizing interaction statistical methodology on relevant data and factoring in multiple limitations.

We examined papers from this prior meta-analysis examining viable and nonviable patients undergoing surgical or medical therapy. We determined an interaction odds ratio for each study and used an empirical Bayes random-effects model to obtain a combined interaction odds ratio that was tested for statistical significance. We compared our results against an interaction odds ratio we estimated from the primary studies included in the previous meta-analysis.

Nine relevant studies with 1244 patients and 172 events were identified that utilized all 4 treatment/viability subsets. The interaction odds ratio was 2.76 (P = .0176, 95% CI 1.19-6.38), 2.5 times lower than our estimated interaction odds ratio of 7.27 for the prior meta-analysis.

We found a markedly reduced but statistically significant interaction between viability status and treatment allocation. However, numerous limitations in the primary studies and the application of meta-analysis along with significant improvements in medical therapies render a randomized controlled trial necessary to reach a definitive conclusion to this critical question.