Author + information
- Received November 9, 1999
- Revision received October 20, 2000
- Accepted November 22, 2000
- Published online March 1, 2001.
- William F Penny, MD, FACC∗,
- Ori Ben-Yehuda, MD†,
- Kenji Kuroe, MA†,
- John Long, PhD∗,†,
- Alan Bond, PhD∗,†,
- Valmik Bhargava, PhD, FACC∗,
- Joseph F Peterson, MD∗,†,
- Martin McDaniel, BA†,
- Joseph Juliano, BS†,
- Joseph L Witztum, MD†,
- John Ross Jr., MD, FACC† and
- Kirk L Peterson, MD, FACC†
This study assessed coronary artery endothelial function in patients with hypercholesterolemia before and after lipid lowering, using quantitative angiography to examine the acetylcholine (Ach) response along the entire analyzable vessel.
Lipid lowering reverses endothelial dysfunction, but whether improvement occurs only in some segments and not others has not been established. Statistical correlation of improvement with specific lipid moieties remains undefined.
Quantitative angiography was performed after Ach (10−6, 10−5, 10−4M) in 29 patients with coronary atherosclerosis before and 18 ± 5.2 months after lipid-lowering treatment (statins, bile sequestrant resins). Standard lipid moieties and markers of oxidized low density lipoprotein (LDL) (immunoglobulin G and M autoantibody titers to malondialdehyde-LDL, E06 epitope) were measured serially.
Pre-treatment of the vessel diameters at control and with 10−6M, 10−5M and 10−4M Ach were 2.108 ± 0.085, 2.086 ± 0.087, 2.069 ± 0.084 and 1.963 ± 0.097 mm (M ± SE), respectively, and increased at follow-up to 2.139 ± 0.094, 2.119 ± 0.086, 2.127 ± 0.084 and 2.080 ± 0.085 mm (p < 0.0001). Improvement in the most constricted and modest declination in the more dilated segments were observed. Change in the E06 and Apolipoprotein A-1 titers correlated with improved vasomotion (p = 0.027 and 0.005, respectively). The pre- and post-treatment levels of the E06 epitope, as well as the post-treatment IgM autoantibody titer to MDA-low density lipoprotein, also correlated (p < 0.028, < 0.001 and p < 0.004, respectively).
Drug treatment reverses endothelial dysfunction, but the effect is heterogeneous. Most coronary segments show enhancement, while others show declination of dilation, underscoring the importance of assessing the entire analyzable artery. Improvement in vasomotion correlates most significantly with markers of plasma-oxidized low-density lipoprotein.
After lipid lowering, a reversal of coronary artery endothelial dysfunction in patients with hypercholesterolemia has been reported (1–5), suggesting that amelioration of constrictive vasomotion may help reduce coronary events. These studies focused on limited segments of the coronary artery and did not account for the heterogeneity of response to acetylcholine (Ach), such that normal vasodilatation and paradoxical vasoconstriction may occur in different segments of the same artery (6,7). Furthermore, it has not been defined whether improvement in coronary reactivity is predicted by absolute pre- or postlipid-lowering plasma levels of lipid moieties or the magnitude of their change during drug treatment.
In this study we examined the changes with lipid lowering in coronary vasomotor response to Ach using serial quantitative angiography with comprehensive analysis of a number of segments of the same artery. A correlation was then made between the change in vasomotor response and pre- and post-treatment serum lipid moieties.
Patient selection, demographics and characteristics
Patients with stable angina, a mildly positive exercise treadmill test and hypercholesterolemia (serum total cholesterol >200 mg/dl) were enrolled in a protocol approved by the Investigational Review Board of the University of California, San Diego, School of Medicine and the Veterans Affairs Medical Center, San Diego. Individuals with a recent myocardial infarction, unstable or variant angina, congestive heart failure or severe peripheral vascular disease were excluded. All patients had mild atherosclerotic luminal irregularities (<50% obstructive) of the artery studied.
After baseline lipid moieties and coronary arteriography had been performed, enrollees were instructed in the Step I diet of the National Cholesterol Education Program and prescribed a graduated dose of an HMG coenzyme A inhibitor (lovastatin) aimed at reducing their total and low density lipoprotein (LDL) cholesterol to below 200 mg/dl and 100 mg/dl, respectively. Seven of 29 patients were given a bile-acid sequestrant subsequently. Follow-up angiography was performed once the patient’s total and LDL cholesterol had each reached a plateau (three successive determinations) on lovastatin of 40 mg/day or greater. If a patient smoked cigarettes, he/she was encouraged to stop although cessation was not required for a patient to remain in the study.
Blood samples were collected monthly. Lipid measurements were as described in the Lipid Research Clinic Manual of Laboratory Operations (8). The inter-assay coefficient of variation was 0.8% for total cholesterol, 1.7% for total triglycerides and 2.4% for high density lipoproteins (HDL) cholesterol. Apolipoprotein (Apo)-A1, Apo-B as well as lipoprotein (a) were determined on a Roche-COBAS BIO instrument using appropriate INCSTAR kits and standards.
Immunoglobulin G and M (IgG, IgM) titers of autoantibodies to malondialdehyde-LDL (MDA-LDL) in human plasma were determined using modifications of a previous assay (9). Malondialdehyde-LDL was produced by sequential ultracentrifugation of healthy human donor plasma in the presence of high concentrations of antioxidants (10). The degree of modification of the lysine residues of Apo-B was determined by the trinitrobenzenesulfonic acid assay (11,12)and was consistently above 75%. Data were expressed in relative luminescence units (RLU) per 100 ms. Samples were run in a single assay; the intra-assay coefficient of variation varies from 8% to 10%.
The murine monoclonal antibody E06 recognizes an oxidized phospholipid epitope that is on LDL as it undergoes oxidation. Thus, it is a sensitive index of early stages of oxidized LDL (ox-LDL) (13,14). The E06 epitope concentration in Apo-B 100 containing particles was measured by a chemiluminescence enzyme immunoassay (14). This assay was based on two monoclonal antibodies, an antihuman Apo B-100 antibody, MB47, and a biotin-labeled antioxidized phospholipid antibody, E06. The amount of Apo-B-100 containing particles bound to the MB47 coated plates was measured by using another biotin-labeled antihuman Apo-B-100 monoclonal antibody, MB24, to normalize the bound LDL of each sample. These two anti-Apo-B-100 antibodies bind to distinct Apo-B-100 epitopes. Data were given for absolute levels of E06 bound per well (RLU/100 ms) or as a ratio of E06/MB24. Because each MB24 antibody binds once to each LDL, the E06/MB24 ratio yields an estimate of E06 epitopes/LDL. All samples were run in a single chemiluminescence enzyme immunoassay; the intra-assay coefficient of variation was 6%.
All vasodilator therapy was discontinued 48 h before the angiographic procedure (6). After diagnostic angiography, an end-hole 3F-infusion catheter was placed into the proximal segment of the coronary artery to be studied. Atrial pacing at 80 beats/min was initiated from the right atrium. Biplane coronary angiography was performed at control and during serial, 3-min infusions of Ach at 10−6M, 10−5M and 10−4M; assuming a coronary flow of 80 ml/min, this would yield intracoronary Ach concentrations of 10−8M, 10−7M and 10−6M, respectively. For each injection, 9 ml of Omnipaque 350 (Winthrop-Breon Laboratories, New York) at 6 ml/s were injected using a Medrad Mark IV injector (Medrad Inc., Pittsburgh, Pennsylvania).
Quantitative coronary angiography
In any given patient, all images were acquired at identical X-ray gantry, table height and source-to-image intensifier distances. Projections chosen displayed the maximum length of the artery and minimized foreshortening and vessel overlap. The guiding catheter was imaged unopacified, and the catheter tip was later measured using electronic calipers to calibrate the magnification for each injection. End-diastolic cine frames were digitized at a resolution of 512 × 512 pixels, 8 bits, at end-diastole using a 35 mm SONY (Tokyo, Japan) SME 3500 Cine Video System interfaced to an Imaging Technology, Inc. (Bedford, Massachusetts) 150 Series video digitizer and a Silicon Graphics, Inc. (Mountainview, California) 4D/380/VGX computer. Vessel edges were determined by an automatic edge detection routine (15)with potential for manual override. The vessel centerline and diameters perpendicular to this centerline were calculated and interpolated as previously described (6). The arteriogram was partitioned into successive 3 mm segments, and the diameters in each segment were then averaged, Figure 1. Our routines ensured the geometric identity, or very near correspondence, of segments that were compared as part of the vasomotion analysis.
Average of segment data per patient
Data were analyzed as a two-level, repeated measures analysis of variance (SuperANOVA 1.11, Abacus Concepts, Inc., Berkeley, California) with the first level comparing responses at each dose of Ach and the second level comparing results at baseline to those after cholesterol lowering. A statistical contrast was constructed before data analysis in which the single comparison was the combined change in the vessel diameter values between baseline and follow-up studies (cell weights ± 0.25 for control, Ach 10−6M, 10−5M and 10−4M for absolute vessel diameter analysis; cell weights ± 0.33 for Ach 10−6M, 10−5M and 10−4M for absolute change [Δ] and percentage change [%Δ] diameter analysis). Because a single comparison was constructed a priori in each contrast analysis, a p value <0.05 was considered significant.
Analysis of segment data, using a standard repeated measures analysis of variance, is confounded by an unequal number of segments nested within each patient and the potential for intra-class correlation of these segments in any given patient. Some authors have approached this type of problem by use of a general linear model of intra-class correlation (16). This approach was not suitable here where repeated measures of coronary vasomotion, before and after lipid lowering, were made in each patient. Thus, we applied a hierarchical linear model (HLM) that adjusts variances for patient differences, segment differences and, finally, repeated measures (17). For hierarchical analysis, we used HLM v. 4.01 (Bryk, Raudenbush & Congdon, Scientific Software International, Chicago, Illinois).
Correlation with lipid moieties
The HLM indicated that 58% of the high-level variance in this study was between patients and 42% was between coronary artery segments within patients. Both variance components were statistically significant. Because significant variance was at the segmental level, the effects of lipid moieties on change in coronary vasomotion were developed in level-3 HLM models using nine measured lipid moieties (total cholesterol, LDL, HDL, lipoprotein (a), Apo-A-1, Apo-B, IgM autoantibody, IgG autoantibody and the E06 epitope). Three separate models were used to analyze the influence of the absolute pre-, the absolute post-, and the change (Δ) in lipid moieties. All possible covariates were included so as to obtain a value for μ0, the estimated residual level-3 error variance. Additional analyses were then run, each one successively excluding one of the set of covariates. The proportion of variance accounted for by a particular covariate was taken as: Only covariates that accounted for at least 5% of the residual error variance were accepted back into the model until the excluded covariates in aggregate accounted for no more than 10% of the residual variance.
Patient attributes and lipid moieties
Forty patients were initially enrolled in the study; seven patients refused follow-up angiography, and in three patients Ach was not given at follow-up, because of progression of coronary artery disease. One patient in whom one dose of Ach was infused was also excluded from final analysis.
The final cohort consisted of 27 men and 2 women, with a mean age of 59.5 ± 8.3 years (M ± SD, range 44 to 73 years). Seven patients were current cigarette smokers, and 17 were former smokers; five had never smoked. Fifteen patients had hypertension, and two had diabetes mellitus. Seven patients were on antioxidant vitamins (Vitamin E, C or both) throughout the protocol. The left anterior descending artery was studied in 24 patients, and the left circumflex artery in five patients. The lipid moiety values, before and after lipid lowering, are given in Table 1.
Quantitative coronary dimensions
Average segment responses per patient
The mean vessel diameters during the series of infusions were significantly greater at follow-up than they were at baseline (n = 29, p < 0.0001, Fig. 2, panel A, Table 2). The response to Ach was calculated as the change in vessel diameter (Δ diam) from control infusion for each segment (negative numbers indicate constriction), then averaged for each patient (Fig. 2, panel B, Table 2). The response to Ach was also calculated as the percent change in vessel diameter from control infusion for each segment (negative numbers indicate constriction), then averaged for each patient (Fig. 2, panel C, Table 2). Vasoconstriction induced by the Ach infusions was significantly decreased after lipid lowering (n = 29, p < 0.05 vs. before lipid lowering).
Hierarchical linear model
In the hierarchical model, the lipid reduction effect (LRfx) is taken as the arithmetical difference between the baseline and follow-up vasomotion response (Δ diam) to each dose of Ach. The effect of lipid lowering is then assessed in terms of the linear relationship between LRfxand the drug dosage of Ach.
In a two-level analysis of covariance, patients were the grouping variable, the mean LRfxfor each patient was the dependent variable and dosage of Ach the covariate. The intercept, which is the mean LRfxadjusted across subjects for Ach 10−5M, was not significantly different from zero (coefficient = 0.0364; t = 0.98; p > 0.3). The common slope estimate, however, was significantly greater than zero (coefficient = 0.0438; t = 2.05; p < 0.05), indicating a significant dosage effect and, therefore, a significant impact of lipid reduction on arterial vasomotion. The variance in the intercept, which indicates the variability among patients in their responses to treatment, was also significant (m0j= 0.0317; c2 = 128.8; p < 0.001). These findings confirm that, after hypolipidemic therapy, coronary dimensions, on average, manifest less vasoconstriction in response to progressive doses of Ach.
In a three-level analysis (analysis of covariance), both patient-level and segment-level effects are incorporated into a single model: level 1 is the assay level, with dosage of Ach as a covariate; level 2 is the segment level, and level 3 is the patient level. The dependent variable is LRfx, and the dose- response slope is constrained to a homogeneous, fixed estimate by setting the random effects in the slope equations (i.e., r1jkand u10k) to zero. Although the slope coefficient was smaller than that estimated from the patient level, it was much more reliable (coefficient = 0.0287; t = 3.90; p < 0.001). The variance in the intercept, which indicates the variability among segments and among patients in their responses to treatment, was also significant (for segments, variance = 0.0270; c2 = 827.8; p < 0.001; for patients variance = 0.0361; c2 = 434.1; p < 0.001). Thus, the incorporation of segmental, as well as average patient, information into the statistical model uncovered more fully the significant effect of lipid lowering on coronary vasomotion.
Heterogeneity of change in vasomotor response
A heterogeneous pattern of response (improvement in endothelial function in at least one segment and declination in endothelial function in at least one segment) was noted in 25 of the 29 patients at the maximum dose of Ach. Moreover, 20 of the 29 patients had vessel segments that both improved and worsened by 10 or more percentage points (Fig. 3). Also, we ranked the segments from worst to best for each patient (as previously mentioned) and adjusted for the disparate number of segments per patient by normalization on a scale from 0 (most constricted) to 100 (most dilated), then analyzed by decile (Fig. 4). At baseline, the most constricted decile demonstrated a mean change of −18% from control, and the most dilated decile demonstrated a change of +6%. At follow-up, with baseline decile rankings maintained, the range had narrowed, and the slope had flattened due to improved function in the most constricted segment deciles and some declination of function in the segment deciles, which dilated at baseline. These changes in response to Ach after cholesterol lowering did not predominate in any given region of the artery.
Correlation of improvement in vasomotion with lipid moieties
The lipid-lowering E06 autoantibody (as a difference and also as an absolute before or after lipid-lowering variable) showed in aggregate the most powerful influence on improvement in vasomotion. The most potent variable was the postlipid-lowering E06 autoantibody level, which accounted for 22.1% of the variance in vasomotion (p < 0.007). The absolute postlipid-lowering IgM autoantibody to oxidized LDL and the change with Apo-A-1 also correlated significantly with the change in coronary vasomotion.
By simple linear regression, no statistically significant correlation (p < 0.05) was found between the pre–lipid lowering Ach response and any one of the pre-treatment lipid moieties. No significant correlation was found between coronary vasomotion improvement and the nominal variables of smoking, hypertension, diabetes mellitus or use of antioxidant vitamins.
We report here the results of a comprehensive coronary artery segmental analysis in patients with hypercholesterolemia treated with cholesterol-lowering agents. While the significant improvement in endothelial function corroborates previous studies (1–5), assessment of the entire analyzable artery suggests that the phenomenon may be more complex than indicated by investigations that focused on selected coronary segments.
Our quantitative angiographic methodology differs from previous lipid-lowering studies in these respects: 1) a number of contiguous segments are analyzed in the same patient; 2) an interpolation routine ensures correspondence of segments analyzed before and after Ach and before and after lipid lowering; (6)and 3) our statistical analysis incorporates a calculation of data variance at both the patient and segmental levels (17). This latter approach uses the statistical power inherent in analyzing several segments in each patient while accounting for nesting of segments within patients, the repeated measures for analysis of coronary vasomotion and the use of several doses of Ach in testing endothelial integrity.
Heterogeneity of vasomotor response to cholesterol-lowering
The response to cholesterol lowering was dependent upon the initial response of the segment to Ach, with a pattern of improved vasomotion in the most constricted segments from the baseline study and a declination of the most dilated segments. This pattern was incremental throughout the artery when baseline responses to Ach were ranked by decile (Fig. 4). The finding of most marked improvement in the most constricted of selected segments of the control study was also described in the work of Anderson et al. (3)and Treasure et al. (2). This pattern also parallels the results of the MARS trial (18), a study of changes in angiographically defined coronary stenoses after cholesterol lowering. Here there was a small degree of regression of stenosis in those patients in the test group (on lovastatin) compared with the placebo group. Subset analysis revealed that lesions >50% stenotic on the baseline study showed regression (improvement), while lesions <50% at baseline demonstrated progression of stenosis severity (worsening) at follow-up.
Sources of heterogeneity
The observed pattern may be due to a regression toward the mean, where extreme outlying points (both the most Ach constricted and dilated segments, arbitrarily chosen for quantification from a normal distribution) fall closer to the central tendency (the average) when resampled and quantified at follow-up. The segmental vasomotor responses also may be constrained by limits of maximal vasoconstriction (i.e., 100%) and vasodilation. Thus, the variability inherent in quantitative analysis may result in less marked mean change at follow-up in segments that exhibit extreme changes upon initial assessment, as there can be more underestimation than overestimation of the subsequent response due to these anatomic constraints.
The heterogeneous response to lipid lowering also may be due to pathophysiologic mechanisms. An obvious source is the segmental distribution of atherosclerosis. Another source could be a nonuniform decrease by lipid lowering in the oxidative milieu of the endothelium and subendothelium. This uneven effect may lead to a variable decrease in sensitivity of the muscarinic receptors of the vascular smooth muscle cells to the effect of Ach. This change could have the most marked effect on the most abnormal vessel segments, resulting in reversal of vasoconstriction, and less effect on the segments that were more normal at the outset and, thus, were not affected predominantly by direct muscarinic stimulation.
Variability in segmental response to lipid lowering may reside in other factors that control vasomotor tone, including shear stress, catecholamines, bradykinin, endothelin-1 and platelet-derived products (19). Isolation of these factors in a clinical investigation is difficult and was not attempted in this protocol. Nevertheless, assuming that coronary blood flow remains constant, one can deduce that basal vasodilation would serve to reduce vessel curvature and diminish shear forces along the wall and in surrounding segments (20,21). This might explain, in part, the reduced dilation with lipid lowering in those segments that initially manifested a more normal response to Ach challenge.
Improvement of the vasomotor response to lipid lowering
The current evidence is compelling that lipid lowering is associated with a net improvement in coronary artery dimensions and the vasomotor response to a drug that stimulates the endothelial-derived relaxation factor, nitric oxide (NO). Several explanations for this improvement appear tenable: 1) reduced potentiation of factors that inactivate endothelial-derived relaxation factor, such as oxidized LDL or its by-products (22–25), 2) an effect of antioxidants (9)or 3) an up-regulation of endothelial nitric oxide synthase (eNOS) as a direct effect of drug treatment (26).
Markers of ox-LDL correlated significantly in this study with the response to Ach. Other investigations support an important role for ox-LDL in endothelial dysfunction. Oxidized LDL adversely affects NO production and activity in vascular ring preparations (27,28)in endothelial cells (22,23)and in platelets (24). Anderson et al. measured the susceptibility of LDL to oxidation using the lag phase of conjugated diene formation induced by Cu++and found that the vasodilator response to Ach related significantly to the susceptibility of LDL to oxidation (25). A reported significant (p = 0.023) correlation between oxidized LDL levels and dipyridamole-induced coronary flow reserve, as measured by positron emission tomography, supports this conclusion (29). Tamai et al. (30)reported improved Ach-induced vasodilatation in the forearm with acute LDL apheresis. In these subjects there was a significant inverse correlation between plasma LDL (as well as oxidized LDL) and the maximal concentration in the venous effluent of the forearm of NOx(a rapidly oxidized, in vivo product of NO) measured before and after Ach administration. Other reports suggest a relationship between autoantibodies of ox-LDL and progression of atherosclerosis (9)as well as risk of myocardial infarction (31).
In the E06 assay utilized here, the murine monoclonal antibody recognizes an oxidized phospholipid epitope on ox-LDL. The number of E06 epitopes per LDL particle was related significantly to the severity of endothelial dysfunction and in the post-therapy study was the single most powerful independent risk factor, accounting for fully 22% of the variance. Because this is a direct assay for an epitope of early stages of ox-LDL, it provides further strong evidence that oxidation of LDL is an important variable influencing abnormal coronary vasomotion in patients with atherosclerosis.
There is also evidence that ascorbic acid, a vitamin antioxidant, reverses endothelial dysfunction in patients with coronary heart disease (32)and probucol, a potent antioxidant as well as lipid-lowering agent, improves endothelial dysfunction over and above that produced by lipid lowering itself (3). Other investigations, however, have shown no improvement in endothelial dysfunction by either a panel of antioxidant vitamins or superoxide (33). We found no significant correlation between improvement in response to Ach and patient use of antioxidants.
Up-regulation of eNOS
Lauf et al. (26)reported that inhibition of endothelial HMG CoA reductase by simvistatin and lovastatin directly upregulates eNOS, predominantly by post-transcriptional mechanisms. This would suggest that lipid lowering might not be integral to the salutary effect of statins. It may also help explain either the absence or the lack of a strong statistical correlation between changes in serum levels of either total cholesterol or LDL and the improvement in vasomotor response to Ach.
This study examined the vasomotor response of a single main artery in the total coronary arterial tree. A different conclusion might be drawn if all branches were examined in the territory of the artery into which Ach was infused or if all coronary arteries were examined. However, the accuracy of quantitative angiography is suboptimal when very small vessels are measured (34). In our own phantom studies, the accuracy and precision of measurement declines significantly when conduits are 0.6 mm or less in diameter (35). Likewise, optimal angiographic projection for all subbranches is technically challenging, even when biplane views are used. Infusion of Ach in all coronary arteries has not been performed due to the need for separate left and right coronary infusions.
We conclude that, in patients with hypercholesterolemia, lipid-lowering therapy improves endothelial function and causes a small, but measurable, dilation of basal coronary artery dimensions. Nevertheless, the response to Ach is complex, with a heterogeneous pattern of increased diameter in some segments and reduction in others. This pattern underscores the importance of inclusion of the entire analyzable artery when studying coronary vasomotion. Investigations that focus on responses from isolated segments of an artery, especially serial studies that include only segments that demonstrate a visually detectable response to Ach at baseline, would be vulnerable to erroneous or exaggerated conclusions. Despite the variability in segmental vasomotion to Ach challenge, the aggregate coronary artery response correlated highly with markers of plasma ox-LDL, reinforcing its pathogenetic role in the endothelial dysfunction associated with atherosclerosis.
The services of Johanna Smith, RN, and Susan Uehling, RN, in the execution of this clinical study are gratefully acknowledged.
☆ Supported, in part, by Ischemic SCOR grant HL-17682-18, the Atherosclerosis SCOR grant HL-56989, the San Diego Foundation for Cardiovascular Research and Education and an unrestricted grant from Merck & Co., Inc., West Point, Pennsylvania.
- endothelial nitric oxide synthase
- high density lipoprotein
- hierarchical linear model
- immunoglobulin G
- immunoglobulin M
- low density lipoprotein
- lipid reduction effect
- nitric oxide
- oxidized low density lipoprotein
- Received November 9, 1999.
- Revision received October 20, 2000.
- Accepted November 22, 2000.
- American College of Cardiology
- Leung W,
- Lau C,
- Wong C
- Egashira K,
- Hirooka Y,
- Kai H,
- et al.
- Seiler C,
- Suter T.M,
- Hess O.M
- Penny W,
- Rockman H,
- Bhargava M,
- et al.
- El-Tamimi H,
- Mansour M,
- Wargovich T.J,
- et al.
- ↵Lipid Research Clinics Program: Manual of Laboratory Operations, Volume 1, Edition 2, Revised 1982: Lipid and Lipoprotein Analysis. U.S. Department of Health, Education and Welfare publication No. (NIH) 76-628. Washington, DC: U.S. Government Printing Office.
- Palinski W.S,
- Ylä-Herttuala S,
- Rosenfeld M.E,
- et al.
- Habeeb A.F.S.A
- Steinbrecher U.P,
- Witztum J.L,
- Parthasarathy S,
- Steinberg D
- Dijkstra E.W
- Bryk A,
- Raudenbush S,
- Congdon R
- Kanai A.J,
- Strauss H.C,
- Truskey G.A,
- Crews A.L,
- Grunfeld S,
- Malinski T
- Niebauer J,
- Cooke J.P
- Liao J.K,
- Shin W.S,
- Lee W.Y,
- Clark S.L
- Chen L.Y,
- Mehta P,
- Mehta J.L
- Anderson T.J,
- Meredith I.T,
- Charbonneau F,
- et al.
- Laufs U,
- La Fata V,
- Plutzky J,
- Liao J.K
- Raitakari O.T,
- Pitkänen O.-P,
- Lehtimäki T,
- et al.
- Tamai O,
- Matsuoka H,
- Itabe H,
- Wada Y,
- Kohno K,
- Imaizumi T
- Levine G.N,
- Frei B,
- Koulouris S.N,
- Gerhard M.D,
- Keaney J.F,
- Vita J.A
- Gilligan D.M,
- Sack M.N,
- Guetta V,
- et al.
- Reiber J.H.C,
- Van Der Zwet P.M.J,
- Von Land C.D,
- et al.
- Peterson K.L