Prognostic value of peripheral endothelial function on major adverse cardiovascular events above traditional risk factors

Abstract

Aims

This study aims to identify whether adding peripheral microvascular dysfunction (PMED), a marker of atherosclerosis to established risk scores has an incremental prognostic value for major adverse cardiovascular events (MACE).

Methods and results

This is a retrospective study of patients who underwent measuring peripheral arterial tonometry from 2006 to 2020. The optimal cut-off value of the reactive hyperaemia index (RHI) that had maximal prognostic value associated with MACE was calculated. Peripheral microvascular endothelial dysfunction was defined as the RHI lower than the cut-off. Traditional cardiovascular risk factors such as age, sex, congestive heart failure, hypertension, diabetes, stroke, and vascular disease were determined to calculate the CHA₂DS₂-Vasc score. The outcome was MACE defined as myocardial infarction, heart failure hospitalization, cerebrovascular events, and all-cause mortality. A total of 1460 patients were enrolled (average age 51.4 ± 13.6, 64.1% female). The optimal cut-off value of the RHI was 1.83 in the overall population and in females and males was 1.61 and 1.8, respectively. The risk of MACE during 7 [interquartile range (IQR): 5,11] years of follow-up was 11.2%. Kaplan–Meier analysis showed that lower RHI is associated with worse MACE-free survival (P < 0.001). Multivariate Cox proportional hazard analysis, controlling for classic cardiovascular risk factors or risk scores such as CHA₂DS₂-Vasc and Framingham risk score revealed that PMED is an independent predictor of MACE.

Conclusion

Peripheral microvascular dysfunction predicts cardiovascular events. Non-invasive assessment of peripheral endothelial function may be useful in early detection and improving the stratification of high-risk patients for cardiovascular events.

Lay summary

• We conducted a retrospective study of patients who underwent peripheral arterial tonometry to, first, identify the optimal cut-off value in predicting major adverse cardiovascular events and, second, evaluate the incremental prognostic value of peripheral microvascular dysfunction over conventional risk assessment models.

Key findings

• The optimum cut-off value of RHI in predicting MACE was 1.83 in the overall population and males, while it was lower in females (1.6).

• Integrating PMED status into previously established risk scores (CHA₂DS₂-Vasc score and Framingham risk score) significantly improves the prognostic stratification of patients at risk of MACE.

Introduction

Endothelial dysfunction plays a key role in atherogenesis and is an important prognostic factor in developing cardiovascular diseases (CVDs).^1,2 Previous studies showed that endothelial dysfunction independently predicts the incidence of major adverse cardiovascular events (MACE) even in patients with minimal risk factors.^3,4 Different methods have been suggested to accurately evaluate endothelial dysfunction including both invasively and non-invasively. Peripheral arterial tonometry (PAT) has been gaining increasing interest in recent years.^5,6 Unlike the other methods of measuring endothelial dysfunction, PAT is a non-invasive, non-operative-dependent method that makes it a useful tool for large cohort studies.⁷ The PAT technology is based on reactive nitric oxide-mediated vasodilation, which is impaired in endothelial dysfunction. Previous studies showed the PAT result has been correlated well with the invasive evaluation of coronary microvascular function.⁶ Even though the association of the PAT and CVDs,⁸ stroke,⁹ and atherosclerotic plaque progression¹⁰ has been extensively investigated, yet, there is no agreement upon the optimal threshold in peripheral microvascular dysfunction (PMED) in identifying high-risk patients for MACE. Moreover, it is not clear whether adding PMED to the established risk stratification method has an incremental prognostic value in predicting MACE.

The current study sought to identify the optimal cut-off value for reactive hyperaemia induced by the PAT test [reactive hyperaemia index (RHI)] that could most accurately identify the patients with high risk for MACE [myocardial infarction (MI), heart failure (HF) hospitalization, cerebrovascular events]. We also tended to evaluate the incremental prognostic value of PMED over two established risk assessment models, the CHA₂DS₂-Vasc score and the 10-year Framingham risk score, to predict the risk for MACE.

Method

This is a retrospective cohort study enrolling participants who underwent peripheral endothelial function assessment using PAT testing at the Center for Coronary Physiology from 2006 to 2020. The indication for the test included patients with angina symptoms who were found to have no new obstructive lesion by coronary catheterization and had a low-risk profile in cardiac stress. The exclusion criteria included inconclusive test results, lack of follow-up after the PAT test, and those who did not provide research authorizations (Figure 1 shows the study flow diagram). The study protocol was approved by Mayo Clinic institutional review board, and written informed consent was obtained from all the participants.

Assessment of peripheral endothelial function

Peripheral microvascular endothelial function was assessed by the reactive hyperaemia response using the digital plethysmography (EndoPAT2000 device, Itamar Medical Inc., Ltd, Caesarea, Israel), a validated method in measuring peripheral endothelial function. The detailed study protocol was described elsewhere.^8,9,11 In brief, after placing the finger probe on the middle fingers of both hands, baseline recording and equilibration were done for 5 min, and then the blood pressure arm cuff of the test arm was inflated to 60 mmHg above the patient’s baseline systolic blood pressure. After 5 min, the cuff was deflated, and reactive hyperaemia was recorded for the next 5 min.

The reactive hyperaemia ratio was calculated by dividing the average pulse wave amplitude after deflation by the baseline average in the same arm. The final measure (RHI) was then computed automatically by normalizing the ratio to the baseline and indexing it to the control arm. All vasoactive medications such as calcium channel blockers, beta-blockers, and nitrates were discontinued 24 h before the procedure.

Data collection and follow-up

Demographic and clinical data were collected through a detailed chart review by an investigator blinded to the PAT data. The collected data included age, sex, and comorbidities such as hyperlipidaemia (HLP; serum total cholesterol ≥ 200 mg/dL), smoking status (current exposure vs. previous or non-smoker), body mass index (BMI, kg/m²), prior history of hypertension (HTN; systolic blood pressure ≥ 140 and/or diastolic blood pressure ≥90 mmHg or being treated for HTN), diabetes mellitus (DM), congestive HF, cerebrovascular events [ischaemic or haemorrhagic stroke and transient ischaemic attack (TIA)], MI, atrial fibrillation (AF), and peripheral vascular disease (PAD) including intermittent claudication, abdominal or thoracic aortic surgery, and arterial or venous thrombosis.

The CHA₂DS₂-Vasc score was calculated by summation of congestive HF (1 point), HTN (1 point), DM (1 point), age (65–74 years, 1 point; ≥ 75 years, 2 points), female sex (1 point), vascular disease (1 point), and previous cerebrovascular event (2 points).¹²

Framingham risk score in patients without prior history of CVDs was calculated using the sex-specific prediction equitation based on conventional cardiovascular risk factors including age, total cholesterol, HDL cholesterol, systolic blood pressure, DM, and smoking status after excluding the patients with a previous history of CVDs.¹³

All patients were followed up by reviewing the medical records looking for the incidence and date of HF hospitalization, newly diagnosed MI, cerebrovascular events, and all-cause mortality. The time and cause of mortality were then verified by reviewing the death certificate.

Statistical analysis

Continuous variables were reported as mean ± SD for normally distributed or median (IQR) for skewed variables. Discrete variables were reported as frequency (percentage). Between groups’ comparison was analysed using paired t-test and Mann–Whitney U test for normally distributed and skewed variables, respectively. Pearson or Kendall correlations were used to measure the strength and direction of association for continuous and ordinal variables, respectively. A chi-squared test was performed for comparing categorical variables.

To determine the optimal cut-off value of RHI in predicting MACE, we performed the surv_cutpoint function in the survminer R package. The subgroup analysis was then performed to obtain an optimal cut-off value for males and females separately. The PMED was then defined as RHI ≤ cut-off and non-PMED as RHI > cut-off.

Kaplan–Meier analysis was performed to estimate the event-free survival with log-rank and P-value reported. Cox proportional hazard analysis was then applied to ascertain the prognostic value of the RHI in MACE in multivariate analysis. Results were reported as hazard ratio (HR) with a corresponding 95% confidence interval (CI). We tested three different Cox proportional hazard models: (i) PMED and traditional cardiovascular risk factors such as age, sex (female as a reference), smoking, BMI, HLP, diabetes, HTN, AF, and cerebrovascular diseases. The selection of the variables was based on clinical relevance and literature,¹⁴ (ii) PMED and CHA₂DS₂-Vasc score, and (iii) PMED with Framingham risk score.

The incremental value of PMED over the CHA₂DS₂-Vasc score and Framingham risk score in predicting MACE was assessed by testing the improvement in fit (Δchi-square) after adding the PMED to the model using the hierarchical strategy. Furthermore, the concordance index (C-index) was measured to compare the prognostic value of different models in predicting MACE. In addition, the time-dependent receiver operating characteristic (time-dependent ROC) analysis was conducted, and the area under the curve (AUC) over the 15 years follow-up time was compared before and after adding the PMED to the risk scores, to evaluate the discriminative prognostic value.

For all tests, a two-tailed P-value < 0.05 was considered statistically significant. All statistical analyses were performed using SPSS software, version 28 (IBM Corp., Armonk, NY, USA), and R studio software (version 4.2.2, http://www.rstudio.org).

Results

Baseline characteristics

During the follow-up period, there were 50 patients (58% female) who were excluded due to loss to following-up for MACE, including 13%, 38%, and 56% for the first 3, 5, and 10 years, respectively. Most of the excluded patients were residing outside the USA (70%). There were no significant differences in RHI between those who were lost to follow-up and the included patients.

A total of 1460 patients were included. The average age at the time of the PAT test was 51.4 ± 13.6 years (range 18–87 years old), and 64.1% were women. The median and IQR were 2.18 (1.75, 2.65) for RHI and1 (1, 2) for CHA₂DS₂-Vasc score. The mean 10-year risk Framingham risk score was 8.6 ± 7.5. Detailed baseline clinical characteristics of participants are presented in Table 1.

Estimating the cut-off value of RHI

Based on the maximum standardized log-rank statistics, the RHI of 1.83 was identified as the best threshold to predict the MACE in total population (log-rank = 6.68). Peripheral microvascular dysfunction and non-PMED were then defined as RHI ≤1.83 and RHI >1.83, respectively (Figure 2).

Sex-specific analysis revealed a cut-off point of 1.8 in males (log-rank = 4.6) and 1.61 (log-rank = 5.13) in females. The association between PMED and clinical characteristics is presented in Table 1.

Cardiovascular risk factors and comorbidities were more significant in PMED compared to non-PMED, including diabetes (14.4% vs. 9.5%, P = 0.008) and higher BMI (29.2 ± 6.35 vs. 28 ± 6.26 Kg/m², P < 0.001) (Table 1). The median CHA₂DS₂-Vasc did not differ significantly between patients with and without PMED [1 (1, 2) in both) P = 0.37] (Table 1).

Long-term follow-up for major cardiovascular adverse events

During 7 (5, 11) years of follow-up, 154 patients (11.2%) developed MACE including HF hospitalization (n = 66, 4.5%), MI (n = 56, 3.8%), cerebrovascular event (n = 46, 3.2%), and all-cause mortality (n = 72, 4.2%).

Patients who developed MACE were significantly older (59.9 ± 13.6 vs. 50.4 ± 13.3, P < 0.001), more likely to be male (65.7% vs. 51.2%, P < 0.001), more likely to have PMED (51.8% vs. 27%, P < 0.001), as well as other cardiovascular risk factors such as HTN (71.3% vs. 35.7%, P < 0.001), DM (23.8% vs. 9.3%, P < 0.001), and AF (25% vs. 4.2%, P < 0.001). The median CHA₂DS₂-Vasc score and Framingham risk score were significantly higher in patients who developed MACE compared to those who did not [3 (IQR, 2) vs. 1 (IQR, 1), P < 0.001] and (11 (IQR, 15) vs. 6 (IQR, 7), P < 0.001), respectively.

Kaplan–Meier analysis showed significant differences in incident MACE between PMED and non-PMED groups (log-rank chi-square = 48.9, P < 0.001) (Figure 3).

Cox proportional hazard analysis was performed on three different models to estimate event-free survival. Model 1 incorporated the PMED, traditional risk factors were statistically significant (likelihood ratio chi-square, 268.04, P < 0.001), and PMED was an independent predictor of poor prognosis of MACE [HR, 2.4 (95% CI, 1.74–3.3, P < 0.001)]. The sequential Cox regression analysis revealed that adding PMED significantly improves the overall model (ΔChi-square = 27.9, P < 0.001) (Table 2).

The second model incorporated the PMED and CHA₂DS₂-Vasc scores and was statistically significant for predicting MACE (Likelihood ratio chi-square, 202.8, P < 0.001). Peripheral microvascular dysfunction was an independent prognostic factor (HR, 2.69, 95% CI, 1.97–3.68, P < 0.001). In addition, PMED significantly improved the prognostic value of the model after being added to the CHA₂DS₂-Vasc score (Δchi-square = 37.6, P < 0.001) with an improvement of C-index from 0.74 (95% CI, 0.7–0.78) to 0.76 (95% CI, 0.72–0.8) (P = 0.006). Furthermore, comparing the AUC under the time-dependent ROC curve at 1, 3, 5, 10, and 15 years of follow-up revealed the incremental prognostic value of PMED to the CHA2DS2-Vasc score, especially for the long-term MACE prediction (Table 3 shows the statistically significant value at 5 years and after) (Figure 4).

Peripheral microvascular dysfunction had also a prognostic significance in Model 3 which incorporated the Framingham risk score and PMED (PMED: HR, 2.83, 1.79–4.46, P < 0.001). The overall model was also statistically significant (chi-square, 92.6, P < 0.001), and PMED had incremental prognostic value in sequential forward modelling of Cox regression (Δchi-square = 19.8, P < 0.001) with an improvement of C-index from 0.75 (95% CI, 0.69–0.81) to 0.76 (95% CI, 0.7–0.82) (P = 0.003) (Table 3). The time-dependent ROC curve demonstrated a slight increase in AUC after integrating the PMED to Framingham risk score, particularly during the short-term follow-up (Figure 5 and Table 4).

Discussion

The current study demonstrated that PMED is associated with a higher incidence of MACE and was an important predictor of poor prognosis independent of traditional cardiovascular risk factors. The optimum cut-off value of RHI in predicting MACE was 1.83 in the overall population and males, while it was lower in females (1.6). The results suggest improvement of the prognostic stratification of patients at-risk of MACE by integrating the PMED status to cardiovascular risk factors or previously established risk scores.

Endothelial cells play a crucial role in maintaining homeostasis, affecting the entire circulatory system. The endothelium regulates the vascular tone and the blood flow through several mechanisms including releasing of vasoactive substances [such as nitric oxide, prostacyclin, and endothelin (ET)] or remodelling of the vessel wall.^15,16 Cardiovascular risk factors exert adverse effects on vascular homeostasis by impairing the endothelial function that subsequently causes blood flow reduction and accumulation of reactive oxygen species.^2,15 The process account for an initial reversible phase of atherogenesis.¹⁶ Previous studies showed that endothelial dysfunction reflects the early manifestation of traditional risk factors. In addition, endothelial dysfunction can be useful to monitor the response to interventions such as lifestyle modification and medication.¹⁷ For instance, in a study of subjects with coronary artery disease, the 12-week intensive lifestyle modifications (plant-based diet, strenuous physical activity, and 1 h/day stress management) were associated with significant improvement of endothelial function.¹⁸ A meta-analysis consisting of 46 trials that investigated the effect of statin therapy on endothelial function revealed that treatment with a statin for the median duration of 8 months was associated with significant improvement of both central and peripheral endothelial function.¹⁹ Peripheral microvascular dysfunction is independently associated with the extent of atherosclerosis detected by coronary CT angiography (CCTA) and 1-year MACE in patients presenting without a history of coronary artery diseases.²⁰ Another study in patients with asymptomatic systolic dysfunction and low risk of CVD revealed that PMED is associated with a four times higher risk of progression to the overt symptomatic stage (Stage C).⁸ Those previous findings together with the result of our work suggest the potential application of endothelial function in the primary and secondary prevention of cardiovascular events.^8,21 In addition, for the first time in this study, we attempted to identify the sex-specific optimal cut-off value of RHI with maximum prognostic value for cardiovascular events in a large cohort of patients during the long-term follow-up. The results demonstrated that a lower cut-off of RHI indicated PMED in females compared to males. One potential reason might be due to the fact that the study population was specific to patients with chest pain and normal coronary artery. It has been shown previously that females in this population are more likely to have microvascular dysfunction; hence, the range of RHI is expected to be lower compared with males. In addition, the difference can also be attributed to the sex-specific microvascular endothelial factors, such as the sex difference of ET-1 expression and receptor activation, the sensitivity of the autonomic nervous system, and the ratio of vasoconstrictors to vasodilators.²² Consistent with our findings, the Framingham Heart Study on third-generation offspring showed that males had higher average baseline pulse wave amplitude, while the reactivity response was significantly lower compared with females.²³

There are a variety of methods for assessing peripheral endothelial function including both invasive and non-invasive techniques. Flow-mediated dilation (FMD) of the brachial artery is one of the most well-established methods that have been shown to provide additive prognostic value to traditional risk factors in predicting the incidence and progression of coronary artery disease.²⁴ However, FMD mostly evaluates conduit vessels; hence, the measurement is influenced by the brachial artery and body size,^25,26 while the PAT testing assesses the endothelial function at both macro- and microvascular levels. Another advantage of the PAT method is the ability to control for any systemic changes by using the contralateral arm as a control.¹⁵

Another highlight of the present study was demonstrating the incremental prognostic value of PMED by integrating it with previously established risk assessment models for cardiovascular events. There are several risk prediction models based on conventional cardiovascular risk factors which have been validated to identify individuals’ risk of long-term cardiovascular events. The CHA₂DS₂-Vasc score is a widely used risk assessment model with the primary aim of predicting stroke in patients with AF.¹² However, later studies demonstrated its application in estimating the risk of CVD and death in non-AF populations as well.^27,28 In a study on patients presenting with acute MI (8.3% had AF), the CHA₂DS₂-VASc score was an independent predictor of increased incidence of 1-year MACE (recurrent MI, admission for HF, cerebrovascular accident, or death) with a score ≥7 associated with 100% risk of MACE.²⁹ In line with previous investigations, our study showed that the CHA₂DS₂-Vasc score modestly predicted the risk of MACE, with a significant increase in prognostic value after incorporating PMED status. In addition, we observed that the model including all traditional risk factors had better overall performance in correctly classifying the MACE occurrence compared to the CHA₂DS₂-Vasc score. This might be due to the limitation of the CHA₂DS₂-Vasc score in incorporating other important CVD risk factors such as BMI, previous history of MI, and AF. This is also a concern with the Framingham risk score which can only be applied to asymptomatic patients with no previous history of CVD.³⁰

This study underscores the potential benefit of assessing peripheral microvascular endothelial as a marker of CVD using the PAT test. Given its significant prognostic value, together with its characteristics as an inexpensive, non-invasive tool, peripheral endothelial dysfunction using the PAT test may be potentially useful in optimizing individualized therapeutic strategies.

However, future investigations are warranted to develop a more comprehensive risk estimation model by incorporating novel risk factors such as PMED, as well as to evaluate the role of PMED in the long-term management and treatment efficacy.

There are several limitations to this study that need to be considered in designing future research. The first limitation is due to the design of the study as an observational cohort; thus, the associations do not necessarily reflect a causal relationship. Also, the data collection in a retrospective manner as in this study is always prone to the risk of selection bias. Of note, since the current study was performed in patients with chest pain and non-obstructive coronary artery disease, it is potentially affected by selection bias; therefore extrapolating the results to the general populations or any specific patient groups requires careful consideration. Second, the optimal cut-off value that was identified for PMED in the current study population needs to be confirmed before applying it to other specific populations. Finally, due to the limited number of patients in our database, we were not able to perform the external validation, which needs to be done in further studies.

Conclusion

This study demonstrated the best cut-off value of RHI to predict MACE is 1.83. Peripheral microvascular dysfunction defined as RHI ≤1.83 was a significant prognostic factor of MACE-free survival even after controlling for traditional cardiovascular risk factors. Moreover, integrating the PMED into CHA₂DS₂-VASc and the 10-year Framingham risk score may improve the risk stratification of patients at risk of cardiovascular events. This study supports the role of non-invasive assessment of endothelial function for risk prediction in our clinical practice.

Funding

None.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Authors’ contribution

N.R., T.T., and A.L. contributed to the conception or design of the work. N.R., T.T., and A.R. contributed to the analysis and interpretation of data for the work. N.R. drafted the manuscript. J.D.S., F.L.J., L.O.L., and A.L. critically revised the manuscript. All gave final approval and agreed to be accountable for all aspects of work ensuring integrity and accuracy.

References

Author notes