The test in question uses a device called EndoPAT, which measures the reactive hyperemia index, or RHI. During the exam, a blood pressure cuff briefly cuts off circulation to one arm, and sensors on the fingertips record how dramatically the pulse amplitude surges when the cuff is released. That surge depends heavily on nitric oxide, a molecule produced by the endothelium, the thin layer of cells lining all blood vessels. Because endothelial dysfunction is a central mechanism in atherosclerosis, RHI has become a validated surrogate for systemic vascular health and has been shown to predict cardiovascular events. The question the researchers asked was whether it could also forecast late lumen loss, the gradual re-narrowing of the stented segment measured in millimeters by quantitative coronary angiography at follow-up.

The biological logic seemed sound. After a stent is implanted, the arterial injury triggers neointimal hyperplasia, an overgrowth of tissue that remains clinically relevant even in the drug-eluting stent era. Endothelial re-coverage of the stent struts is a key determinant of healing, and impaired nitric oxide bioavailability promotes vascular smooth-muscle proliferation, the same pathway that drives restenosis. Two earlier studies had linked RHI to restenosis, but with a crucial caveat: in both, the predictive signal came from measurements taken after the procedure, not before. Komura and colleagues found that follow-up RHI independently predicted binary in-stent restenosis in 249 patients, while Yamamoto’s group showed that an eight-month, but not an early three-week, RHI was associated with restenosis after primary PCI. No prior study had related pre-procedural RHI to continuous late lumen loss or asked whether any relationship depended on the caliber of the implanted device.

The new study analyzed 132 consecutive patients who underwent EndoPAT assessment before PCI between February 2009 and June 2016, with follow-up angiography available in 91 of them at a median interval of 9.4 months. The cohort was typical of a PCI population: mean age 64.1 years, 72 percent male, 42.4 percent with diabetes, and near-universal secondary prevention therapy, including statins in 97 percent. Mean RHI was 1.636, and 63.6 percent of patients fell below the 1.67 threshold conventionally used to define endothelial dysfunction. Most implanted devices, 90 percent, were second-generation drug-eluting stents, with a mean diameter of 3.06 millimeters.

Because late lumen loss was severely skewed, with a skewness of 4.6 and a kurtosis of 29.4, the team relied on rank-based correlation and a robust regression technique with HC3 standard errors rather than ordinary least squares. The result was essentially null. RHI showed only a borderline inverse correlation with in-stent late lumen loss, a Spearman coefficient of minus 0.19 with a p-value of 0.07, and it was not independently associated with the outcome in the adjusted model, with a beta of minus 0.11 and a 95 percent confidence interval spanning minus 0.39 to plus 0.18. Multiple imputation for the 31 percent of patients missing follow-up, outlier removal, logarithmic transformation of RHI, and lesion-level models using generalized estimating equations all returned concordant null estimates. No correlation survived false-discovery-rate correction across the three angiographic outcomes examined.

The exploratory part of the analysis is where things got more interesting. When the researchers stratified patients by the diameter of the stent actually implanted, a correlation emerged only in the small-caliber group. Among the 33 patients who received stents of 2.75 millimeters or less, the RHI–late lumen loss correlation was minus 0.375, nominally significant at p equals 0.032, while among the 58 recipients of larger stents it was a negligible minus 0.055. Notably, the same stratification performed using the native reference vessel diameter, the size of the artery before treatment, showed no association at any cut-off. The authors argue this asymmetry makes biological sense: the endothelium heals the implanted stent lumen, not the pre-procedural reference segment, so the achieved caliber is the relevant substrate. Reference vessel diameter is also interpolated from adjacent segments and carries more measurement error, which would dilute any true effect modification toward the null.

Yet the researchers are careful to label this finding hypothesis-generating rather than confirmatory, and for good reason. The formal stent-diameter-by-RHI interaction was not statistically significant, with a p-value of 0.48, and a Fisher r-to-z comparison of the two subgroup correlations also fell short at p equals 0.13. A statistically significant correlation in one subgroup alongside a non-significant one in another does not itself prove the difference between them is real. Moreover, the study was powered to detect only fairly large effects: with 91 patients, it had 80 percent power for a Spearman coefficient of 0.29 or greater, and within the 33-patient small-caliber subgroup only 0.47 or greater. The nominally significant small-caliber correlation could therefore equally be an overestimate produced by low statistical power.

The composition of the small-stent group adds a further complication. Recipients of stents 2.75 millimeters or smaller had virtually identical high-sensitivity C-reactive protein levels to those with larger stents, but they carried a heavier metabolic and renal burden: 57.6 percent had diabetes compared with 27.6 percent, their estimated glomerular filtration rate was lower at 71.6 versus 83.1 milliliters per minute per 1.73 square meters, and they had more diseased vessels. They also received beta-blockers and ACE inhibitors or ARBs less often. A small implanted stent, in other words, marks a higher-risk patient phenotype as well as a smaller treated caliber. Nineteen of the 33 small-caliber patients also had diabetes, so the caliber and diabetes subgroups overlapped substantially and cannot be treated as independent signals. In a parallel diabetes-stratified analysis, the RHI–late lumen loss correlation was stronger among patients with diabetes, minus 0.366, than without, minus 0.020, and the diabetes-by-RHI interaction was nominally significant in the primary model but lost significance after additional adjustment for lesion length and smoking.

Inflammation, at least as indexed by a single pre-procedural hs-CRP measurement, did not explain any of the results. hs-CRP was uncorrelated with RHI, with late lumen loss, and with implanted stent diameter, and adding it to the primary model left the RHI coefficient essentially unchanged. Neurohormonal blockade was too uniform across the cohort to be informative, since more than four in five patients were taking a beta-blocker and nearly nine in ten an ACE inhibitor or ARB, and no direct neurohormonal biomarker was recorded. Lesion length, which was associated with late lumen loss at a coefficient of plus 0.27, behaved as a competing predictor rather than a confounder, since it was independent of RHI and its inclusion barely moved the RHI estimate.

The authors conclude that pre-procedural EndoPAT RHI is unlikely to serve as a stand-alone predictor of angiographic restenosis in unselected patients, and their data do not support routine pre-procedural testing for this purpose. The pattern across studies, in which baseline measurements fail while post-procedural ones predict, suggests that endothelial function assessed after vascular injury and healing carries more restenosis-relevant information than a single pre-procedural value. Any future confirmation of the small-stent signal, they argue, should come from prospective studies pre-stratified by intended device diameter, pairing baseline with post-procedural RHI, and adjusting for diabetes and renal function. The limitations are worth keeping in view: this was a single-center retrospective series with selective EndoPAT referral, follow-up angiography was partly clinically driven and missing in nearly a third of patients in a way related to smoking status, and the confidence interval for the primary correlation, minus 0.38 to plus 0.02, remains compatible with a small-to-moderate inverse association that a larger study might yet detect.

Subject of Research: Association between peripheral microvascular endothelial function and in-stent late lumen loss after percutaneous coronary intervention

Article Title: Peripheral microvascular function and in-stent late lumen loss after percutaneous coronary intervention: a retrospective cohort study with an exploratory analysis by implanted stent diameter

Article References: Kim, H., Joh, H. S., Lim, W.-H., Kim, H.-L., Kim, S.-H., & Seo, J.-B. (2026). Peripheral microvascular function and in-stent late lumen loss after percutaneous coronary intervention: a retrospective cohort study with an exploratory analysis by implanted stent diameter. Frontiers in Cardiovascular Medicine, 13, 1940897. https://doi.org/10.3389/fcvm.2026.1940897

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1940897

Keywords: endothelial function, reactive hyperemia index, EndoPAT, in-stent restenosis, late lumen loss, percutaneous coronary intervention, quantitative coronary angiography, stent diameter, drug-eluting stents, neointimal hyperplasia, retrospective cohort study, coronary artery disease