Endothelial Dysfunction and the Risk of Hypertension
The Multi-Ethnic Study of Atherosclerosis
Hypertension is associated with impaired endothelial function in cross-sectional studies. However, few longitudinal data exist on whether endothelial dysfunction precedes the development of hypertension. We examined the cross-sectional and longitudinal relationships between endothelial-dependent brachial artery flow-mediated dilation (FMD) and hypertension prevalence and incidence in 3500 participants from the Multi-Ethnic Study of Atherosclerosis, an ethnically diverse, community-based cohort study. At baseline, the prevalence ratios (95% CI) of hypertension from the highest to the lowest quartile of FMD were 1.00 (referent), 1.26 (1.12 to 1.40), 1.35 (1.21 to 1.52), and 1.68 (1.50 to 1.87; linear trend P<0.001). This association remained (P=0.017) after adjustment for demographics (age, sex, and ethnicity), Multi-Ethnic Study of Atherosclerosis site, and other risk factors. Of the 1869 participants without hypertension at baseline, 584 (31.3%) developed hypertension over a median follow-up of 4.8 years. The unadjusted relative risks (95% CI) of incident hypertension from the highest to the lowest quartile of FMD were 1.00 (referent), 1.38 (1.14 to 1.67), 1.44 (1.19 to 1.74), and 1.64 (1.36 to 1.97; linear trend P<0.001). However, after adjustment for demographics and Multi-Ethnic Study of Atherosclerosis site, the relationship between FMD and incident hypertension was attenuated and not statistically significant: 1.00 (referent), 1.26 (1.04 to 1.52), 1.19 (0.98 to 1.44), and 1.18 (0.97 to 1.44). The longitudinal results also did not appreciably change after adjustment for additional risk factors and baseline blood pressure levels. In this sample, reduced FMD was not an independent predictor of hypertension incidence, suggesting that impaired endothelial function does not play a major role in the development of hypertension.
The normal endothelium senses hemodynamic forces and biochemical signals from the blood and, in turn, responds by synthesizing and releasing vasoactive substances.1 Endothelial-dependent flow-mediated vasodilation (FMD) is predominantly modulated by endothelium-derived NO, which stimulates soluble guanylyl cyclase activity in vascular smooth muscle cells.2 In addition to inducing vasodilation, NO inhibits leukocyte adhesion, thrombosis, and cellular proliferation in the vessel wall.3 Endothelial dysfunction is considered to be an early process in the development of atherosclerosis.4
Hypertension is an established risk factor for incident cardiovascular disease (CVD), including coronary artery disease, peripheral arterial disease, stroke, and heart failure.5 The etiology of essential hypertension has been investigated, and the underlying dysregulations are complex and likely multifactorial. Because several studies have shown that hypertension is associated with impaired brachial artery FMD,6–9 some investigators have proposed that endothelial dysfunction is mechanistically implicated in a sustained increase in blood pressure.10 However, virtually all of the published studies linking hypertension to impaired FMD have been cross-sectional, making it difficult to ascertain whether endothelial dysfunction precedes the onset of hypertension.
Prospective data on the relation between FMD and the subsequent risk of incident hypertension are limited. Rossi et al11 found that lower levels of FMD were associated with an increased risk of incident hypertension in an outpatient cohort of 952 apparently healthy postmenopausal women. Few longitudinal data are available on the relationship of FMD with incident hypertension in men, in minorities, or in population-based epidemiological samples. Also, it is not known whether FMD is associated with hypertension incidence independent of other risk factors that are associated with both impaired FMD and hypertension. We, therefore, sought to evaluate whether FMD is an independent predictor of new-onset hypertension in a multiethnic, population-based cohort of middle-aged and elderly men and women.
The current analysis included participants from the Multi-Ethnic Study of Atherosclerosis (MESA), a population-based study of 6814 community dwelling adults aged 45 to 84 years at baseline. Details of the MESA design have been described elsewhere.12 Participants from 4 ethnic groups (white, black, Hispanic, and Asian primarily of Chinese descent) were recruited from 6 US communities, including Baltimore, Maryland; Chicago, Illinois; Forsyth County, North Carolina; Los Angeles County, California; northern Manhattan, New York; and St Paul, Minnesota. Participants were excluded if they had a history of clinically evident CVD; were under treatment for cancer; were pregnant; weighed >300 lb; had any serious medical condition that would prevent long-term participation in MESA; had significant cognitive deficits; were living in a nursing home or were on the waiting list for a nursing home; had plans to leave the community within 5 years; spoke a language other than English, Spanish, Cantonese, or Mandarin; and/or had a chest computed tomography scan in the previous year. Written informed consent was obtained from all of the participants, and the study was approved by the institutional review boards of all of the participating sites.
FMD assessment was performed at baseline (examination 1). Of the 6814 MESA participants, 6489 (95.2%) successfully underwent FMD testing. Participants were excluded from FMD testing if they had a history of Raynaud phenomenon (n=55), a congenital abnormality of the arm or hand (n=12), or a radical mastectomy on either side (n=100). Participants (n=158) were also excluded at the time of the FMD examination if they had blood pressures in the left and right arms that differed by >15 mm Hg or had uncontrolled hypertension (systolic blood pressure >180 mm Hg or diastolic blood pressure >110 mm Hg). FMD studies were analyzed by the Wake Forest University Cardiology Image Processing Laboratory. Images from 5731 participants were of sufficient quality for reading. Because of financial constraints, images were read for only a subset of participants (n=3501). For our analyses, a single participant with a sex-specific FMD >6 SDs above the mean was excluded, resulting in available data from 3500 participants.
Baseline Risk Factor Measures
Information on demographics, current smoking, education level, current alcohol intake, physical activity, and medical history was obtained using standardized questionnaires.12 Education level was assessed by determining the highest level achieved. Physical activity was defined as the total of all light, moderate, and vigorous activities multiplied by individual metabolic equivalent values for these activities. Anthropometric measurements of height and weight were determined by the use of calibrated scales. Body mass index was calculated as weight in kilograms divided by height in meters squared. Levels of total cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, and glucose were measured from blood samples obtained after a 12-hour overnight fast and under standardized conditions. The Friedewald equation was used to calculate low-density lipoprotein (LDL) cholesterol. Diabetes mellitus was defined as a fasting serum glucose ≥126 mg/dL or use of hypoglycemic drugs or insulin. Serum creatinine was measured, and estimated glomerular filtration rate (eGFR) was calculated by the Modification of Diet in Renal Disease equation.13 High-sensitivity C-reactive protein was measured using a particle-enhanced immunonepholometric assay on a BNII nephelometer.
Flow-Mediated Dilation Assessment
FMD was determined using high-resolution ultrasonography of the brachial artery.7,8,14,15 In brief, participants were asked to abstain from food, consumption of vitamin E or C, and smoking for ≥6 hours before the scan. A standard blood pressure cuff was positioned around the right arm, 2 inches below the antecubital fossa, and the brachial artery of the right arm was imaged 5 to 9 cm above the antecubital fossa using a 9-MHz linear array transducer (M12L transducer, GE Healthcare) at rest and during a 2-minute period beginning immediately before cuff deflation. To induce reactive hyperemia, the brachial artery was occluded for 5 minutes at an occlusion cuff pressure of ≥50 mm Hg above the participant’s systolic blood pressure. Images were digitized, and data were analyzed using a validated semiautomated system.15,16 FMD was expressed as the percentage of increase in the brachial artery diameter (media-adventitial interface to the media-adventitial interface) with reactive hyperemia: FMD=[(peak brachial artery diameter after cuff deflation−diameter at rest)/diameter at rest]×100. A more detailed description of the scanning and reading protocol can be found at the MESA Web site (http://www.mesa-nhlbi.org).
To evaluate intrareader reproducibility for resting brachial artery diameter, peak diameter, and FMD, ultrasound studies from 40 MESA participants (30 men, 18 white, 2 Chinese-American, 10 black, and 10 Hispanic) were re-examined.15 The intraclass-correlation coefficients were 0.99, 0.99, and 0.93, respectively.
Blood Pressure Measurements and Hypertension Ascertainment
Blood pressure measurements were performed at each MESA examination, which were conducted at 18-month intervals. Data from exams 1 to 4 were available for our analyses. After resting for 5 minutes in the seated position, participant blood pressure was measured 3 times at 2-minute intervals using an automated oscillometric device (Dinamap Monitor Pro 100, GE Healthcare). Appropriate sized cuffs were used for blood pressure assessment. Blood pressure was defined as the average of the second and third readings. Participants were asked about their previous diagnoses of hypertension, and the use of antihypertensive medications was also assessed.
Hypertension at baseline (ie, prevalent hypertension) was defined by the presence of any of the following criteria: (1) self-reported history of hypertension; (2) systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg; and/or (3) the use of antihypertensive medication.17 Among participants without hypertension at baseline, the incidence of hypertension was defined as the first follow-up study visit with the presence of systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg and/or the use of antihypertensive medication.
We first examined the cross-sectional relation between FMD and prevalent hypertension in the 3500 MESA participants with an FMD measurement at baseline. We then examined the longitudinal relation between baseline FMD and incident hypertension at follow-up (exams 2 to 4). After excluding participants who had hypertension at baseline (n=1624) and those who did not attend ≥1 follow-up visit (n=7), a total of 1869 MESA participants were available for the analysis of incident hypertension. To account for the sampling of participants for the reading of FMD images, following the MESA analytic guidelines, we used probability sampling weights for all of the analyses. In the current analysis, the unweighted estimates did not differ markedly from the weighted estimates (data not shown). Therefore, we only report the weighted estimates.
For the analysis of FMD and prevalent hypertension, we divided the study population into quartiles on the basis of the distribution of FMD in the MESA cohort. Characteristics of participants and the prevalence of hypertension were calculated by quartile of FMD. Next, the unadjusted prevalence ratios of hypertension and the prevalence ratios adjusted for age, sex, ethnicity, and MESA site associated with quartile of FMD were calculated using log-binomial regression models. An additional model with multivariable adjustment for covariates that might be related to FMD or hypertension was also fitted. In addition to demographics (age, sex, and ethnicity) and MESA site, the following covariates (all chosen a priori) were included: body mass index, diabetes mellitus, LDL and HDL cholesterol levels, cigarette smoking, current alcohol intake, education level (defined as high school education or higher), physical activity, eGFR, and C-reactive protein. Linear trends across quartiles were assessed by including quartile-specific median FMD values as a continuous variable in the regression models. The association between FMD, expressed as a continuous variable, and hypertension at baseline was also evaluated using prevalence ratios, which were modeled to reflect an SD decrease in FMD.
Next, the population eligible for the incident hypertension analyses was divided into quartiles on the basis of their distribution of FMD. Characteristics of the population included in the analysis of incident hypertension were calculated by quartile of FMD. The incidence of hypertension was calculated for each FMD quartile. Unadjusted hypertension incidence rates were calculated as the number of events in each FMD quartile divided by the sum of person-years at risk. The time at risk was calculated as the number of days between the baseline examination and examination 4, unless a participant developed hypertension at an earlier visit (ie, exam 2 or 3) or did not attend examination 4, in which case we calculated the risk time as the elapsed time from baseline to the last examination the participant attended (ie, examination 2 or 3 for n=6 and n=64 participants, respectively). For the analysis of incident hypertension, Poisson regression was used to calculate the adjusted relative risk and 95% CI of hypertension associated with FMD. In addition to the covariates described above, baseline blood pressure was included as an additional covariate in the multivariable adjusted models. Because the distribution of FMD may vary by sex, all of the analyses were repeated using sex-specific FMD quartiles. Results were markedly similar (data not shown), and, therefore, we only report on nonsex-specific FMD quartiles.
We also examined the relationship between resting brachial artery diameter and incident hypertension. However, the results were not appreciably different compared with FMD (data not shown). Therefore, herein, we only report on results using FMD as a main predictor variable.
In an additional secondary analysis, we examined an outcome of incident sustained hypertension, defined as having hypertension at exams 2, 3, and 4 or exams 3 and 4. We also examined an outcome of a clinically meaningful increase in blood pressure, defined as an increase in systolic blood pressure ≥10 mm Hg, an increase in diastolic blood pressure ≥5 mm Hg, and/or the introduction of an antihypertensive medication during follow-up.17 Finally, secondary analyses were performed after incorporating prehypertension in the outcome: incidence of crossing the threshold for prehypertension (≥120/80 mm Hg) and/or the use of antihypertensive medication. However, because the results were similar (data not shown), we do not report on these latter analyses. Statistical analyses were conducted with SAS 9.2 (SAS Institute).
Table 1 shows the baseline characteristics of the samples used for the cross-sectional and longitudinal analyses, according to FMD quartiles. For the cross-sectional analyses, lower FMD quartile levels were associated with older age, black ethnicity, diabetes mellitus, higher levels of systolic and diastolic blood pressure, and lower levels of eGFR. In contrast, higher FMD levels were associated with female sex, white or Chinese ethnicity, a high school education, alcohol intake, and small but significantly higher LDL cholesterol levels. This latter relation was no longer significant after excluding participants on statins (data not shown). For the longitudinal analyses of incident hypertension, lower FMD levels were associated with older age, black ethnicity, diabetes mellitus, and higher baseline blood pressure levels, whereas higher FMD levels were associated with female sex and white or Chinese ethnicity.
FMD and Hypertension Prevalence
At baseline, lower FMD levels were significantly associated with a higher prevalence of hypertension (Table 2). After adjustment for age, sex, ethnicity, and MESA site, the relation between FMD and hypertension was somewhat attenuated, but the trend across FMD quartiles remained statistically significant. This association did not change appreciably after inclusion of additional covariates. Each SD decrease in FMD was associated with a 32% (95% CI: 25% to 39%) greater unadjusted prevalence of hypertension (Table 3). In the fully adjusted model, each SD decrease in FMD was associated with a 10% (95% CI: 3% to 16%) higher prevalence of hypertension.
FMD and Hypertension Incidence
Over a median follow-up of 4.8 years (25th to 75th percentiles: 4.6 to 5.0 years), 584 (31.3%) of the 1869 participants without hypertension at baseline developed hypertension. Lower FMD levels were significantly associated with an increased unadjusted risk of incident hypertension (Table 4). However, after adjustment for age, sex, ethnicity, and MESA site, these associations were substantially attenuated and no longer statistically significant. Adjustment for additional covariates did not alter these findings. A similar pattern was observed when FMD was expressed as a continuous variable (Table 5). Each SD decrease in FMD was associated with a 21% (95% CI: 8% to 35%) higher unadjusted risk of incident hypertension. However, after further covariate adjustment, the relation was attenuated and not statistically significant. Exploratory analyses revealed that, among age, sex, and ethnicity, age had the greatest attenuating effect on the association between lower FMD levels and incident hypertension (data not shown). Table S1, available in an online Data Supplement (please see http://hyper.ahajournals.org), shows the relative risk of incident hypertension for the covariates in the multivariable models.
FMD and Incident Sustained Hypertension
The incidence rates of incident sustained hypertension (per 1000 person-years) were 36.1 for the highest FMD quartile (≥6.5%), and 48.2, 43.2, and 59.5 for decreasing FMD quartiles (4.3% to 6.4%, 2.6% to 4.2%, and <2.6%, respectively; P=0.002 for trend). However, these relations were attenuated, and the trend was not statistically significant after adjustment for demographics, MESA site, and also other covariates (P for trend across quartiles=0.673 after adjusting for age, sex, ethnicity, and MESA site; P for trend across quartiles=0.907 after additional adjustment for body mass index, diabetes, LDL and HDL cholesterol levels, cigarette smoking, alcohol intake, education level, physical activity, eGFR, and C-reactive protein; and P-trend across quartiles=0.922 after additional adjustment for baseline systolic and diastolic blood pressure levels).
FMD and a Clinically Meaningful Increase in Blood Pressure
There was no relationship between baseline FMD and a subsequent increase in systolic blood pressure ≥10 mm Hg, diastolic blood pressure ≥5 mm Hg, and/or the initiation of antihypertensive medication. The unadjusted relative risks (95% CI) from the highest to the lowest quartile of FMD were as follows: 1.00 (referent), 1.01 (0.89 to 1.15), 1.04 (0.92 to 1.18), and 1.09 (0.96 to 1.24; P for trend=0.181). After adjustment for demographics and MESA site, the relationship between FMD and incident hypertension became even weaker: 1.00 (referent), 1.00 (0.88 to 1.13), 0.99 (0.87 to 1.13), and 0.99 (0.87 to 1.14; P for trend=0.906). The results were unchanged after further covariate adjustment (data not shown).
A number of prospective studies including MESA8,15,18 have shown that impaired FMD, a measure of NO bioavailability, predicts CVD events, independent of traditional risk factors. Given the important role of NO in blood pressure regulation and vascular function, a reduction in NO bioavailability, mediated by an increased generation of reactive oxygen species,19,20 may be a mechanism underlying chronic blood pressure elevation and hypertension onset. Several studies have demonstrated a consistent cross-sectional relation between impaired FMD and hypertension.6–10 However, it remains unclear whether endothelial dysfunction predicts progression to hypertension.
In the current study, reduced FMD was associated with an increased prevalence of hypertension at baseline, and this relation was independent of several other possible explanatory factors. This finding is consistent with previous cross-sectional studies6–10 that have examined the relation between FMD and hypertension. Although reduced FMD was associated with incident hypertension in unadjusted analyses, this relationship was attenuated and not statistically significant in adjusted models. A European of Society Working Group21 proposed in 2005 that impaired endothelial function is unlikely to be a direct causal mechanism of hypertension initiation and maintenance. Our results provide longitudinal evidence in support of this hypothesis.
At least 1 previous longitudinal study has examined the relation between impaired FMD and incident hypertension.11 A total of 952 normotensive postmenopausal women, recruited from an outpatient center in Italy, had FMD assessed and were followed for hypertension incidence over a mean period of 3.6 years. Participants were excluded if they had a history of overt CVD or a history of traditional risk factors, such as hyperlipidemia, smoking, diabetes mellitus, and obesity. Lower FMD levels predicted incident hypertension, even after adjusting for age, family history of hypertension, baseline blood pressure levels, body mass index, waist circumference, duration of menopausal period, years of education, alcohol use, and physical activity. The association between reduced FMD and incident hypertension was large in magnitude. Compared with the highest quartile of FMD (≥5.5%), the multivariable-adjusted relative risks (95% CI) associated with decreasing FMD quartiles (4.3% to 5.4%, 3.6% to 4.2%, and ≤3.5%) were 1.92 (1.62 to 3.55), 3.00 (2.43 to 4.29), and 5.77 (4.34 to 8.10), respectively.
The contrary findings of our study may be explained by differences in the characteristics of the study population. Our study used a large multiethnic, community-based sample that included an equivalent number of men and women drawn from several geographically diverse communities. Participants in MESA were also not excluded on the basis of traditional CVD risk factors. In addition, the quartile-stratified incident hypertension rates (per 1000 person-years), as shown in Table 4, were substantially greater than those observed in the study conducted by Rossi et al (10.4 for Q4, 20.9 for Q3, 40.8 for Q3, and 61.1 for Q1).11 However, the cumulative incidence of hypertension in the current study (31.3% over a median of 4.8 years) is comparable to the proportions observed in other population-based studies, including the Framingham Heart Study.22–24
Overall, the results of our study strongly suggest that endothelial dysfunction is not an independent predictor of hypertension in the general population. One explanation for this finding is that endothelial dysfunction may have influenced other risk factors for hypertension that were included in the multivariable-adjusted models. However, the relationship between lower FMD levels and incident hypertension became weaker and nonsignificant after adjustment for demographics and MESA site, which are factors that FMD could not have directly influenced. A more likely explanation is that endothelial dysfunction is a consequence of hypertension. In the Cardiovascular Risk in Young Finns Study,25 FMD was assessed in adults aged 24 to 39 years, and the relation of blood pressure levels measured in childhood and adolescence with subsequent FMD levels was examined. In male participants, systolic blood pressure levels in adolescence predicted lower FMD levels in adulthood, independent of traditional risk factors. These findings suggest that blood pressure elevations, at least in adolescent boys, may adversely affect the biological processes underlying endothelial function, such as NO bioavailability, in early adulthood. Therefore, a chronic increase in blood pressure may induce endothelial damage over time, thereby contributing to atherosclerosis development and CVD event onset.4
There are several limitations to our study. The follow-up period was relatively short. However, hypertension incidence during follow-up was relatively common, even among those in the highest FMD quartile. It is, therefore, possible that the relation between reduced FMD and hypertension onset may become even weaker over a longer follow-up period as the incident hypertension rates between the highest and lowest FMD quartiles narrowed. Although nonsignificant, the adjusted relative risks of incident hypertension associated with lower FMD quartiles were not entirely negligible. Thus, despite our study being the largest to date (N=1869) to examine the relation between FMD and incident hypertension, it is possible that an even larger sample size might have led to a statistically significant association between lower FMD quartile levels and incident hypertension. However, no clear association with incident hypertension was present when examining FMD as a continuous variable (adjusted relative risk: 1.00; P=0.935; model 3; Table 5). In addition, even if the results in the fully adjusted model (model 3; Table 4) were to become statistically significant in a larger sample, the relative risk (95% CI) of 1.14 (0.93 to 1.38), comparing Q1 versus Q4, is substantially weaker in magnitude than the relative risk of 5.77 (4.34 to 8.10) observed in the smaller study (N=952) by Rossi et al.11 Endothelium-independent vasodilation, typically assessed by exogenous nitrate administration, was not assessed in this study. Thus, we cannot definitely exclude the possibility that reduced FMD was additionally explained by smooth muscle dysfunction. Similar to other large population-based studies,8 we did not include this measure because of limited feasibility in conducting this measure in a large number of participants. Blood pressures measured in the clinic environment are known to be variable. Therefore, it is possible that we included participants with hypertension at baseline in the analysis of FMD and incident hypertension. However, the results were not different when excluding those participants with baseline blood pressures in the prehypertensive range (systolic/diastolic blood pressure: 120 to 139/80 to 89 mm Hg26; data not shown). The classification of hypertension at follow-up could also be affected by misclassification, but analyses defining the outcome as incident hypertension across consecutive visits (ie, incident sustained hypertension) did not produce different results.
Strengths of the current study include the use of a large multiethnic cohort that was drawn from several communities in the United States, the longitudinal study design, and the careful and standardized assessment of cardiovascular risk factors, including blood pressure readings across time. Because limited data exist on the longitudinal relation between FMD and the subsequent risk of hypertension, the current study provides valuable new information.
Because evidence from cross-sectional studies has shown that impaired NO-mediated endothelial-dependent vasodilation is associated with hypertension, it has been proposed that endothelial dysfunction may be an important underlying causal factor in hypertension onset. In an ethnically diverse, community-based population sample, no independent association was present between reduced FMD and the incidence of hypertension. These findings do not support the contention that endothelial dysfunction plays a major role in hypertension onset. Interventions designed to improve NO bioavailability may not reduce the incidence of hypertension in initially nonhypertensive individuals.
We thank the other investigators, the staff, and the participants of the MESA for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org.
Sources of Funding
MESA was supported by contracts NO1-HC-95159 through NO1-HC-95166 and NO1-HC-95167, as well as R01-HL077612, from the National Heart, Lung, and Blood Institute.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
- Received September 17, 2009.
- Revision received October 11, 2009.
- Accepted February 27, 2010.
Joannides R, Haefeli WE, Linder L, Richard V, Bakkali EH, Thuillez C, Luscher TF. Nitric oxide is responsible for flow-dependent dilatation of human peripheral conduit arteries in vivo. Circulation. 1995; 91: 1314–1319.
Benjamin EJ, Larson MG, Keyes MJ, Mitchell GF, Vasan RS, Keaney JF Jr, Lehman BT, Fan S, Osypiuk E, Vita JA. Clinical correlates and heritability of flow-mediated dilation in the community: the Framingham Heart Study. Circulation. 2004; 109: 613–619.
Yeboah J, Crouse JR, Hsu FC, Burke GL, Herrington DM. Brachial flow-mediated dilation predicts incident cardiovascular events in older adults: the Cardiovascular Health Study. Circulation. 2007; 115: 2390–2397.
Wallace SM, Yasmin, McEniery CM, Maki-Petaja KM, Booth AD, Cockcroft JR, Wilkinson IB. Isolated systolic hypertension is characterized by increased aortic stiffness and endothelial dysfunction. Hypertension. 2007; 50: 228–233.
Bild DE, Bluemke DA, Burke GL, Detrano R, Diez Roux AV, Folsom AR, Greenland P, Jacob DR Jr, Kronmal R, Liu K, Nelson JC, O'Leary D, Saad MF, Shea S, Szklo M, Tracy RP. Multi-ethnic study of atherosclerosis: objectives and design. Am J Epidemiol. 2002; 156: 871–881.
Corretti MC, Anderson TJ, Benjamin EJ, Celermajer D, Charbonneau F, Creager MA, Deanfield J, Drexler H, Gerhard-Herman M, Herrington D, Vallance P, Vita J, Vogel R. Guidelines for the ultrasound assessment of endothelial-dependent flow-mediated vasodilation of the brachial artery: a report of the International Brachial Artery Reactivity Task Force. J Am Coll Cardiol. 2002; 39: 257–265.
Yeboah J, Folsom AR, Burke GL, Johnson C, Polak JF, Post W, Lima JA, Crouse JR, Herrington DM. Predictive value of brachial flow-mediated dilation for incident cardiovascular events in a population-based study: the Multi-Ethnic Study of Atherosclerosis. Circulation. 2009; 120: 502–509.
Herrington DM, Fan L, Drum M, Riley WA, Pusser BE, Crouse JR, Burke GL, McBurnie MA, Morgan TM, Espeland MA. Brachial flow-mediated vasodilator responses in population-based research: methods, reproducibility and effects of age, gender and baseline diameter. J Cardiovasc Risk. 2001; 8: 319–328.
Gokce N, Keaney JF Jr, Hunter LM, Watkins MT, Menzoian JO, Vita JA. Risk stratification for postoperative cardiovascular events via noninvasive assessment of endothelial function: a prospective study. Circulation. 2002; 105: 1567–1572.
Wolin MS. Interactions of oxidants with vascular signaling systems. Arterioscler Thromb Vasc Biol. 2000; 20: 1430–1442.
Brunner H, Cockcroft JR, Deanfield J, Donald A, Ferrannini E, Halcox J, Kiowski W, Luscher TF, Mancia G, Natali A, Oliver JJ, Pessina AC, Rizzoni D, Rossi GP, Salvetti A, Spieker LE, Taddei S, Webb DJ. Endothelial function and dysfunction: part II–association with cardiovascular risk factors and diseases: a statement by the Working Group on Endothelins and Endothelial Factors of the European Society of Hypertension. J Hypertens. 2005; 23: 233–246.
Juonala M, Viikari JS, Ronnemaa T, Helenius H, Taittonen L, Raitakari OT. Elevated blood pressure in adolescent boys predicts endothelial dysfunction: the Cardiovascular Risk in Young Finns Study. Hypertension. 2006; 48: 424–430.
Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr, Jones DW, Materson BJ, Oparil S, Wright JT Jr, Roccella EJ. Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. Hypertension. 2003; 42: 1206–1252.