We appreciate the letters by Poullis and Pullan and Picard and Ly commenting on our recent publication on predictors of internal mammary artery (IMA)-left anterior descending (LAD) artery graft failure. Poullis and Pullan identified differences in the findings from our Project of Ex-vivo Vein Graft Engineering via Transfection (PREVENT) IV publication and their unpublished data from a large single-center cohort in Liverpool, UK. In their analysis, they found that (1) the use of a diagonal graft in addition to a left IMA-to-LAD graft was associated with worse long-term survival, and (2) this association was only found in patients with diabetes mellitus. Assuming that survival is, at least in part, affected by the patency of the left IMA-LAD graft, the authors speculate that perhaps these different findings are the result of differences in sample size and related statistical power. We agree that the incongruence between these findings is interesting. The data from Poullis and Pullan suggest that there may be an interaction between diabetes mellitus and additional diagonal grafting for IMA-LAD graft failure. We performed interaction testing using data from PREVENT IV and did not find a significant interaction between diabetes mellitus and additional diagonal bypass grafting for IMA-LAD graft failure (P-interaction=0.59). An alternative explanation for these different findings may lie in the different outcomes used in the 2 studies. Their analysis used survival as the main outcome, whereas our analysis included direct measurement of graft failure by means of angiographic follow-up in >1500 patients, irrespective of symptom status. Survival is almost certainly related to factors beyond IMA-LAD graft failure. Picard and Ly question whether medical therapy was adequately optimized, and postulate that perhaps suboptimal medical regimens could have played a role in IMA graft failure. They wonder whether we included medication use in the statistical modeling. We agree with the authors that optimizing medical therapy remains a challenge in both clinical practice and in clinical trials. In our study, we did not include information about postprocedural medication use in our prediction models of IMA graft failure. We did provide data on medication use at 30 days and 1 year following surgery in a supplemental table (Table I in the online-only Data Supplement). Although the use of dual-antiplatelet therapy, nitrates, angiotensin-converting enzyme inhibitors, and angiotensin receptor blockers did differ, we did not find significant differences between patients with and without IMA failure in the use of aspirin, β-blockers, lipid-lowering agents, or diuretics. Including postbaseline variables such as medication use in statistical models to predict IMA graft failure can be problematic. The major challenge is that we do not know when graft failure occurred, other than before 12 to 18 months. Changes in postoperative medication use could just as easily be the result as the cause of IMA graft failure. Thus, given the aforementioned limitations, including medications in the IMA graft failure prediction models would not be informative. Picard and Ly raise the prospect of using fractional flow reserve as a tool to guide (surgical) revascularization. We agree with their comments that evaluating the hemodynamic significance of lesions can be challenging in serial coronary lesions (such as combined left main and proximal LAD lesions). The authors wonder whether we looked at intermediate LAD lesions as stand-alone lesions or whether we adjusted for combinations of levels of severity of left main and LAD lesions as predictors of IMA graft failure. In our prediction models, we took into account both the severity of the stenosis in the LAD and the left main coronary artery, because these factors could affect graft patency. Unfortunately, we collected limited detail regarding baseline native coronary anatomy in PREVENT IV and cannot definitely determine whether a lesion was flow limiting or whether competitive flow was present. Theoretically, it makes sense that fractional flow reserve assessment may contribute to reclassification of invasive treatment strategy or help guide surgical revascularization; however, this strategy has never been tested in patients undergoing coronary artery bypass grafting or to select patients for coronary artery bypass grafting versus percutaneous coronary intervention. Fractional flow reserve–guided revascularization strategies should be specifically tested in coronary artery bypass grafting surgery before they are used in this setting.