When Bypass Grafts Slow Down: New Study Separates Competitive Flow from True Blockage

When a coronary artery bypass graft delivers less blood than surgeons expect, the operating room falls into an uncomfortable silence. Is the graft itself failing—a technical problem at the anastomosis that demands immediate revision—or is another vessel quietly feeding the same territory, siphoning flow away in a phenomenon known as competitive flow? Distinguishing between these two causes has long been one of the most frustrating ambiguities in cardiac surgery, because the transit-time flow measurement probes applied during the operation produce numbers that look disturbingly similar in both situations. A new retrospective study published in BioMedical Engineering OnLine by Chen Kai, Mao Boyan and colleagues takes a systematic swing at this diagnostic dilemma, using postoperative computational physiology as a reference standard to work out which intraoperative and anatomical signals actually point toward competition rather than obstruction.

The research team assembled a single-center cohort of 157 patients who had undergone coronary artery bypass grafting with intraoperative transit-time flow measurement, or TTFM, the ultrasound-based technique that records mean graft flow, pulsatility index and diastolic filling in real time. From this cohort the investigators initially identified 130 left internal mammary artery grafts, the arterial workhorse of modern bypass surgery, and 171 saphenous vein grafts, the venous conduits typically reserved for targets that arterial grafts cannot reach. Applying an intraoperative mean flow threshold of 20 milliliters per minute or less, they flagged 86 low-flow LIMA grafts and 84 low-flow vein grafts for deeper analysis. This threshold matters: it is the point at which surgeons begin to worry, and it is precisely the population in which the competitive-flow-versus-stenosis question becomes clinically urgent.

To establish what was actually happening inside each graft, the team turned to computed fractional flow reserve, or cFFR, derived from postoperative coronary computed tomography angiography. Rather than relying on a pressure wire pushed into the artery during a second catheterization, the researchers built patient-specific zero-dimensional and three-dimensional models of each patient’s coronary circulation and computed the fractional flow reserve numerically. Values between 0.75 and 0.80 were treated as an equivocal gray zone and excluded from the strict binary analyses, leaving 80 arterial grafts and 80 vein grafts with unambiguous physiology. This modeling approach is elegant in principle: by fusing the anatomy from CT with lumped-parameter representations of the downstream microcirculation, the models yield a physiologic index that behaves like the invasive gold standard without the risks of a repeat procedure.

The results split cleanly along graft-type lines, and that split is the study’s central message. In the left internal mammary artery grafts, the preoperative physiology of the target coronary artery dominated the picture. Each 0.1-unit increase in preoperative cFFR was associated with significantly higher odds that a low-flow graft represented competitive flow rather than a stenotic problem, with an odds ratio of 1.418 and a confidence interval running from 1.086 to 1.851, statistically significant at P equals 0.010. In plain terms, the healthier and less diseased the native vessel downstream of the anastomosis, the more likely it was to compete with the graft and drag its measured flow down. A continuous analysis of cFFR as an outcome confirmed the same relationship, reinforcing that the arterial graft’s low-flow state tracks the native vessel’s preserved physiology.

The hemodynamic fingerprints of the two phenotypes in arterial grafts were equally telling. Unadjusted comparisons showed that grafts in the competitive-flow group had lower mean wall shear stress and lower wall shear stress gradients along the graft wall, together with higher relative residence time, a parameter that quantifies how long blood particles linger in a region and is associated with thrombogenic potential. After the researchers applied Holm correction for multiple comparisons, the differences in mean wall shear stress and relative residence time remained statistically significant, while the wall shear stress gradient difference did not. Biologically, this pattern makes sense: a graft that is being quietly outcompeted carries less flow, and lower flow means lower shear on the endothelium and slower, longer-residing blood, both of which are conditions that in the long run can promote graft failure even when the original anastomosis was technically perfect.

Saphenous vein grafts told a completely different story. In the strict binary analysis of 80 vein grafts drawn from 59 patients, analyzed with a patient-clustered Firth-type penalized generalized estimating equation model to account for the fact that some patients contributed multiple grafts, the preoperative cFFR lost its statistical significance. Instead, two intraoperative TTFM waveform characteristics took center stage: higher pulsatility index and higher diastolic filling fraction were both associated with lower odds of the competitive-flow phenotype, with odds ratios of 0.725 and 0.919 respectively, both statistically significant. The pulsatility index reflects how pulsatile the graft flow waveform is, and the diastolic filling fraction captures how much of the graft’s flow arrives during diastole, the phase in which coronary perfusion naturally dominates. Vein grafts with more coronary-like, diastole-dominant waveforms were therefore less likely to be competing with a vigorous native vessel.

Perhaps the most sobering finding for the computational side of the field was what the CFD parameters failed to show in the vein grafts. After Holm correction, no CFD-derived parameter differed significantly between the two vein graft phenotypes, a null result that contrasts sharply with the significant wall shear stress and residence time differences seen in the arterial grafts. The authors also report the discriminative performance of their models honestly: optimism-corrected areas under the curve were 0.653 for the LIMA analysis and 0.703 for the vein graft analysis, with a cluster-bootstrap fitting success rate of 67.85 percent for the more complex vein model. These are modest discrimination values, useful for framing hypotheses rather than for making bed-side decisions, and the team is explicit that the findings are exploratory and require prospective external validation before any clinical application.

The methodological care underlying these conclusions deserves attention, because low-flow graft studies are notoriously vulnerable to statistical fragility. The gray-zone exclusion of six LIMA and four SVG grafts with cFFR values between 0.75 and 0.80 sharpened the binary contrast but reduced the sample, and the supplementary analyses that treated cFFR as a continuous outcome served as a check that the binary categorization was not driving the associations. Internal validation through 2,000 bootstrap resamples corrected the apparent performance of the logistic and penalized models for overfitting, and the clustering of multiple grafts within the same patient was handled with penalized generalized estimating equations rather than naive pooling. Each of these choices trades statistical power for robustness, and the modest AUCs reflect that trade. The work was supported by the National Natural Science Foundation of China, the Natural Science Foundation of Beijing and an undergraduate research fund at Beijing University of Chinese Medicine, and the ethics approval came from the Medical Ethics Committee of Beijing University of Chinese Medicine.

What does this mean for the surgeon staring at a TTFM console at two in the morning? The study suggests that the answer depends on which conduit is in hand. A sluggish arterial graft attached to a target with good preoperative physiology is more likely to be a victim of its own success, competing with a native vessel that never became severely stenotic, whereas a sluggish vein graft with a low pulsatility index and a low diastolic filling fraction fits the competitive pattern instead. Neither signal is definitive on its own, and the authors are careful not to claim that intraoperative numbers alone can replace angiographic or physiologic confirmation. But by grafting together three layers of evidence—the preoperative coronary physiology, the intraoperative waveform characteristics and the postoperative computational hemodynamics—the study sketches a graft-specific framework for interpreting one of cardiac surgery’s most common intraoperative anxieties. If prospective validation confirms these associations, the low-flow graft of the future may be read not as a single alarming number but as a pattern, with arterial and venous conduits each telling their own hemodynamic story.

Subject of Research: Hemodynamic differentiation of competitive flow from anastomotic stenosis in low-flow coronary artery bypass grafts

Article Title: Competitive flow identification and hemodynamic analysis in coronary artery bypass grafting

Article References: Kai, C., Gaoyang, L., Zhou, Z., Yanxi, C., Bao, L., Yue, F., Shaohui, Y., & Boyan, M. (2026). Competitive flow identification and hemodynamic analysis in coronary artery bypass grafting. BioMedical Engineering OnLine. https://doi.org/10.1186/s12938-026-01622-6

Image Credits: AI Generated

DOI: 10.1186/s12938-026-01622-6

Keywords: coronary artery bypass grafting, competitive flow, transit-time flow measurement, computed fractional flow reserve, computational fluid dynamics, wall shear stress, left internal mammary artery, saphenous vein graft, hemodynamics, coronary computed tomography angiography, pulsatility index, graft failure