Low Coronary Wall Shear Stress Is Associated With Severe Endothelial Dysfunction in Patients With Nonobstructive Coronary Artery DiseaseFree Access

Subjects and study design

We selected all patients who were enrolled in our prospective Emory intravascular registry between August 2012 and October 2014. The Emory intravascular registry consists of patients who had an abnormal noninvasive stress test or presented with refractory anginal syndromes, underwent cardiac catheterization, were found to have nonobstructive epicardial coronary atherosclerotic lesions (fractional flow reserve >0.80), and underwent further coronary reactivity testing with intracoronary Ach infusion. Exclusion criteria included acute myocardial infarction with cardiogenic shock or hemodynamic instability, ejection fraction <30%, coronary artery bypass surgery, severe valvular heart disease, lesions requiring percutaneous or surgical revascularization, presence of visual coronary collateral vessels, inability to provide informed consent, or significant kidney (creatinine >1.5 mg/dl) or liver disease (aspartate aminotransferase, alanine aminotransferase, or alkaline phosphatase higher than 3 times the normal limit). We also excluded patients with hemoglobin <7 g/dl, platelet count <100 K/μl, or international normalized ratio >1.5. For the present analysis, in addition to the clinical exclusion criteria, patients who did not have sufficiently high-quality baseline and post-Ach angiograms for accurate 3-dimensional (3D) reconstruction for purposes of WSS calculations and endothelial function assessment were also excluded. All patients provided written informed consent and the Emory University Institutional Review Board approved the study.

Coronary angiography, coronary velocity measurement with endothelial function assessment

As described previously (1,2,5,14), all vasoactive medications were held for at least 24 to 48 h before cardiac catheterization. Systolic blood pressure, diastolic blood pressure, and heart rate were measured before the start of the procedure. Patients underwent angiography in a biplane cardiac catheterization system (Philips Medical Systems, Andover, Massachusetts; or Toshiba America Medical Systems, Tustin, California). Pressure and velocity measurements were obtained using a 0.014-inch pressure and Doppler flow velocity monitoring guidewire (ComboWire, Volcano Corporation, Rancho Cordova, California). The ComboWire was advanced to the guide catheter tip where the aortic pressure and guidewire pressures were equalized. The guidewire was advanced into the proximal, nontortuous portion of the left anterior descending coronary artery or left circumflex artery (based on operator discretion) at least 5 mm from major angiographic side branches (>2-mm diameter) and the inlet velocity was recorded. The ComboWire was then advanced into the mid or distal artery and the outlet velocity and pressure was recorded. Patients were first evaluated with 10⁻⁸ M Ach for safety before proceeding with 10⁻⁶ M Ach (18.2 μg/ml at the rate of 90 ml/h infused intracoronary through a 3-F microcatheter over 3 min). Biplane coronary angiography was performed after each infusion to assess vascular response. Patients were not moved or repositioned in between serial coronary angiograms. The same radiographic system, magnification, and views were used for the serial angiograms. Intravenous nitroglycerin (200 μg) was then given to reverse residual vasoconstrictive effects and angiography repeated. Subsequently, 140 μg/kg/min adenosine was infused through a ≥20-gauge intravenous catheter for 2 min to induce maximal coronary hyperemia. Continuous recordings of aortic pressure, distal pressure, and velocity were obtained during baseline and drug infusion periods. During maximal hyperemia, fractional flow reserve was defined as the ratio of mean distal to aortic pressure and coronary flow velocity reserve was defined as the ratio of hyperemic to basal average peak velocity. Velocity measurements demonstrated good reproducibility with a concordance correlation coefficient of 0.979 (95% confidence interval: 0.966 to 0.988) (1).

Endothelial function, computational fluid dynamics, and WSS methodology and calculations

Three-dimensional geometric reconstructions of each patient’s target vessel were performed offline on pre- and post-Ach cineangiograms using QAngio XA 3D RE (Medis Medical Imaging Systems, Leiden, the Netherlands). All visible branching vessels were included in the 3D reconstruction. The pre- and post-Ach reconstructions were coregistered using branching points as fiduciary anatomical landmarks. The resulting pre- and post-Ach 3D models were analyzed by means of fully automated software tools for the extraction of centerlines, computation of pointwise radius map, and the discretization of each vessel in longitudinal segments of 0.5-mm length (Figure 1A) (see Online Appendix) (15,16). Percent change in diameter (%ΔD) was calculated for each 0.5-mm segment as 100% × (D_post − D_pre) / D_pre. Based on prior studies, each coronary segment demonstrating %ΔD >−10% was defined as having severe EDFx (17–21). Similarly, segments with %ΔD between −10% to +10% in response to Ach were defined as having mild EDFx, whereas those with %ΔD ≥+10% were defined as having normal endothelial function. After the reconstructed baseline 3D surface was meshed, computational fluid dynamics was carried out using the finite element library LifeV (EPFL, Lausanne, Switzerland; Politecnico di Milano, Milan, Italy; French Institute for Research in Computer Science and Automation, Rocquencourt, France; and Emory University, Atlanta, Georgia). Post-processing was carried out by the open source software Paraview 5.4 (Kitware, Clifton Park, New York). The fluid (blood) was assumed to be a homogeneous pulsatile isothermal incompressible Newtonian fluid (viscosity constant with respect to shear rate) (22), allowing for the use of the incompressible Navier-Stokes equations that can be numerically solved by applying the finite element method (23). Three cardiac cycles were simulated to reach the desired pulsatile regime with satisfactory accuracy in the velocity field. Boundary conditions of the target vessel and its branches were specified as a series of velocity and pressure profiles measured by the ComboWire in conjunction with an application of Murray’s law (24,25) that weighs the flux of each arterial branch, including the distal portion of the target vessel, with respect to its neighbor. This has been demonstrated to be a reasonable mathematical assumption for mildly stenotic coronaries, as in the present study (26). Spatial velocity profiles were assumed to be flat in the 80% inner part of the proximal vessel, with a linear decrease in the profile in the outermost part of the proximal vessel (to satisfy the no-slip boundary condition on the wall), and parabolic in the arterial branch outlets. After computing the pulsatile flow field in the region of interest, the WSS map was determined on the vessel surface as a function of time in the cardiac cycle (27,28), then averaged over time and circumference at each 0.5-mm segment for quantitative analysis (Figure 1B). Based on previous experimental and human data, low WSS was categorized as <1 Pa (1,2,5,29–31). Intermediate WSS was categorized as ≥1 Pa but <2.5 Pa, whereas high WSS was defined as ≥2.5 Pa.

Statistical analysis

Continuous variables are summarized as mean ± SD or median (interquartile range [IQR]), as appropriate, and categorical variables as count and proportion. The association between WSS and severe EDFx was investigated using logistic regression models. To account for the correlations within patient due to repeated measurements (segments), a robust “sandwich” variance estimator was used to estimate the variance-covariance matrix of the regression parameter coefficients by generalized estimating equations. An independence working correlation structure was chosen based on the QIC (quasi-likelihood under the independence model criterions) statistic. The covariates included age, sex, race (African American vs. non-African American), diabetes mellitus (glycosylated hemoglobin >7%), hypertension (systolic blood pressure >140 mm Hg or diastolic blood pressure >90 mm Hg), hyperlipidemia (low-density lipoprotein >100 mg/dl), and active smoking (yes vs. no). Subsequently, a linear regression model was used to investigate the relationship between WSS and %ΔD, again with generalized estimating equations to account for within-subject correlations, and adjusted for the aforementioned covariates. In these models, WSS was treated as a binary variable (<1 Pa vs. ≥1 Pa) and then a continuous variable (measured in Pascals). In addition, a generalized additive mixed model was used to explore the relationship between WSS and severe EDFx (with logit link). The model was fit by a smooth term of WSS using a cubic spline smoothing with a basis dimension of 3. A p value < 0.05 was considered statistically significant. Analyses were performed using SAS 9.4 (SAS Institute, Cary, North Carolina), R version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria), and SPSS 24.0 (IBM Corporation, Armonk, New York).

Segment level relationship between endothelial function and WSS

After accounting for within subject correlation, segments with low WSS demonstrated greater vasoconstriction in response to Ach than did segments with intermediate or high WSS (−10.7% [IQR: −12.1% to −8.8%] vs. −2.5% [IQR: −8.3% to +5.5%] vs. +1.3% [IQR: −8.6% to +11.3%], respectively; p < 0.001) (Figure 2). Furthermore, low WSS was associated with a mean −0.11 %ΔD in response to Ach (p < 0.001) after adjusting for various demographic and cardiovascular risk factors (Table 2).

A higher proportion of segments with low WSS had severe EDFx (71%), compared with intermediate WSS (22%) or high WSS (23%) (p < 0.001) (Figure 3). Interestingly, no segment with normal endothelial function had low WSS. In a univariable logistic regression analysis, low WSS was associated with severe EDFx (odds ratio [OR]: 8.42; p < 0.001]. In a multivariable logistic regression model, female sex (OR: 2.44; p = 0.04), diabetes (OR: 5.01; p = 0.007), and low WSS (OR: 9.14; p < 0.001) were independent predictors of severe EDFx (Table 3). Figure 4A shows a representative coronary vessel with numerous low-WSS segments and Figure 4B shows the corresponding distribution of endothelial function in the same vessel. Notice the high number of segments with severe or mild EDFx colocalized with low-WSS areas. Figure 5 shows the estimated probability of severe EDFx as a function of WSS between 0 and 10 Pa. The estimated probability of severe EDFx increases from 0.50 (50%) to 0.95 (95%) as WSS decreases from 4 to 1 Pa.

Mechanisms of WSS and EDFx: From bench to bedside

There is a growing body of molecular data linking low WSS to the development of EDFx. Experimental ex vivo studies have demonstrated multiple mechanisms by which low WSS induces changes in endothelial cell morphology with loss of alignment and increased turnover (32–34). Sustained low WSS affects the structure and function of endothelial cells, resulting in transformation to a proatherogenic phenotype with polygonal morphology (32,34,35). Moreover, low WSS down-regulates the expression of endothelial nitric oxide synthase (36–38), which generates nitric oxide and plays a crucial role in maintaining coronary arterial vasodilation. The loss of endothelial nitric oxide synthase, and decreased NO production, leads to graded impairment of normal dilatory response of the coronary arteries to endothelium-dependent vasodilators and ultimately vasospasm. Furthermore, in animal models it has been shown that low WSS promotes atherosclerotic plaque progression by increasing endothelial low-density lipoprotein uptake and synthesis (39), up-regulating the expression of growth factors and endothelin, down-regulating the expression of growth inhibitors (e.g., transforming growth factor-β) (30,40), and enhancing the proinflammatory gene expression through the mitogen-activated protein kinase and nuclear factor-kappa B pathway (41,42).

Whether these cellular and experimental observations relating low WSS to numerous proatherosclerotic pathways are relevant in human coronary atherosclerosis was previously unknown. We demonstrate that coronary segments with low WSS demonstrate greater absolute vasoconstriction in response to Ach compared with segments with intermediate and high WSS. Our finding that, even after adjusting for demographic and clinical risk factors, low WSS is a strong predictor of severe EDFx, with an odds ratio of 9.14, higher than that of diabetes (5.01) and female sex (2.44), provides a strong underpinning of the association between low WSS with severe EDFx in patients with nonobstructive CAD. Moreover, we show that the probability of severe EDFx increases with decrease in WSS and a value of <1 Pa is could be used to identify segments with severe EDFx.

Colocalization of EDFx and low WSS

Our study demonstrates that a significantly greater number of segments with severe EDFx reside in low WSS areas. Furthermore, none of the segments with normal endothelial function colocalize with low WSS. Although it is known that WSS patterns are largely driven by regional vascular geometry (low WSS in inner curvature of vessels, outer hips of bifurcations, and upstream and downstream from stenoses), the vascular response to Ach infusion is likely not primarily driven by geometry. Indeed, endothelial responses to Ach are heterogeneous in the same patient despite exposure of all vessel segments to the same systemic risk factors (9,11,12). These data demonstrate the complex interplay between WSS and endothelial function in the presence of nonobstructive CAD. Prognostic data demonstrate that low WSS is associated with coronary plaque progression (1,3,4) and that patients with EDFx have adverse cardiovascular events. Future studies evaluating the relative predictive value of regional EDFx and WSS with respect to plaque progression and clinical events are warranted.

Clinical implications

This study demonstrates that an important regional precursor of coronary plaque progression namely, low WSS, is independently associated with severe EDFx, the final common pathway through which cardiovascular risk factors contribute to atherosclerosis. Interestingly, majority of coronary segments with low WSS had severe EDFx. Although it is known the patients with severe coronary EDFx have adverse cardiovascular outcomes and that coronary segments with low WSS are associated with plaque progression, the long-term outcomes of coronary segments that display both EDFx and low WSS are not known (1,43). Future studies are warranted to investigate the prognostic value of such coronary segments. Another opportunity with such detailed physiologic phenotyping of patients is to refine our diagnosis of patients with ongoing chest pain, ischemia, and nonobstructive CAD. Indeed, almost two-thirds of patients with nonobstructive CAD and persistent chest pain with or without a positive stress tests have evidence of coronary EDFx (44). Whether patients with severe EDFx and evidence of segmental low WSS should undergo more intense risk factor modification, antiatherosclerotic or anti-inflammatory therapies remains to be investigated.

Study strengths and limitations

A strength of our analysis is the methodology of WSS calculation using 3D reconstruction, patient specific pulsatile flow, and boundary conditions imposed on the computational fluid dynamics simulations. The 3D quantitative coronary angiography (QCA) methodology used has been shown to be more accurate than 2-dimensional QCA (conventional single-vessel QCA) measurements (45–48). Although accurate coregistration of 3D WSS values and 2-dimensional QCA-derived diameter measurements for endothelial function assessment is usually challenging, in the present study both these variables were derived from the same 3D-reconstructed geometry, allowing for seamless integration and analysis of the data. Furthermore, once the 3D models were reconstructed, all subsequent operations on surfaces and meshes (centerlines computation, longitudinal discretization) were fully automated, strengthening the reproducibility of the proposed investigation.

A number of limitations require consideration when interpreting the results of the present study. First, the described relationship between WSS and EDFx does not imply cause and effect, but rather an association. However, the vascular biological underpinnings of both disturbed WSS and EDFx are well established and elucidating the mechanistic links between these 2 variables was not the goal of this investigation. Second, correlated error is introduced by the clustering of numerous arterial segments within patients; however, appropriate statistical methods were used to adjust for correlated error. Third, although both WSS and endothelial function data were derived from the same 3D-reconstructed geometries, inherent inaccuracies due to vessel foreshortening can occur during post-processing of the cineangiograms. We have attempted to minimize these errors through stringent quality control, including comparison of vessel lengths at each clearly demarcated branch point, reconstruction using different views and angles, and standard operating protocols. Furthermore, as per our prespecified protocol, majority of the vessels included in the study were left anterior descending artery, future studies could investigate the relationship between low WSS and severe EDFx in the right coronary artery. In addition, prospective multicentric studies investigating the impact of low WSS and EDFx on clinical outcomes can add provide further insights to the results of our study. Last, although we observed that diabetes associated with severe EDFx even after accounting for data clustering, with only 1 patient with diabetes in our dataset, we had limited ability to test this association.