Abstract / Summary
Low graft flow after coronary artery bypass grafting (CABG) may result from anastomotic stenosis or competitive flow, but differentiating these potential causes using intraoperative flow measurements remains challenging. This study aimed to explore graft-type-specific factors associated with low-flow phenotypes using postoperative computed fractional flow reserve (cFFR) as a model-based physiological reference and to characterize the corresponding hemodynamic patterns. This retrospective, single-center study included 157 patients who underwent CABG with intraoperative transit-time flow measurement (TTFM). A total of 130 left internal mammary artery (LIMA) grafts and 171 saphenous vein grafts (SVGs) were initially identified. After applying an intraoperative mean graft flow threshold of ≤ 20 mL/min, 86 low-flow LIMA grafts and 84 low-flow SVG grafts were identified. After excluding 6 LIMA and 4 SVG grafts with postoperative cFFR values in the gray zone of 0.75–0.80, 80 LIMA and 80 SVG grafts remained for the strict binary analyses. Patient-specific 0D–3D models were constructed from postoperative coronary computed tomography angiography, and cFFR and computational fluid dynamics (CFD) parameters were calculated. cFFR values of 0.75–0.80 were regarded as an equivocal gray zone, and supplementary analyses treated cFFR as a continuous outcome. The LIMA binary outcome was analyzed using multivariable logistic regression, whereas the SVG binary outcome was analyzed using a patient-clustered Firth-type penalized generalized estimating equation model. Internal validation was performed using 2,000 bootstrap resamples. In LIMA grafts, each 0.1-unit increase in preoperative cFFR was associated with higher odds of the competitive flow phenotype (odds ratio [OR] = 1.418, 95% confidence interval [CI] 1.086–1.851; P = 0.010) and with higher continuous cFFR ( β = 0.018, 95% CI 0.004–0.031; P = 0.013). Unadjusted comparisons showed lower WSS-mean and WSSG and higher RRT in the competitive flow phenotype; after Holm correction, only the differences in WSS-mean and RRT remained statistically significant. In the strict SVG binary analysis of 80 grafts from 59 patients, higher PI (OR = 0.725, 95% CI 0.543–0.969; P = 0.030) and DF (OR = 0.919, 95% CI 0.856–0.986; P = 0.019) were associated with lower odds of the competitive flow phenotype, whereas preoperative cFFR was not statistically significant. Continuous cFFR analysis similarly showed inverse associations with PI and DF. No CFD-derived parameter differed significantly between the SVG phenotypes after Holm correction. The optimism-corrected AUCs were 0.653 for LIMA and 0.703 for SVG; the SVG cluster-bootstrap fitting success rate was 67.85%. Low-flow LIMA and SVG grafts showed distinct associations between preoperative coronary physiology, intraoperative TTFM characteristics, and postoperative computational hemodynamics. Integrating these measurements provides a graft-specific framework for characterizing heterogeneous low-flow phenotypes after CABG. These findings are exploratory and require prospective external validation before clinical application.