Value of HDL Cholesterol, Apolipoprotein A-I, Lipoprotein A-I, and Lipoprotein A-I/A-II in Prediction of Coronary Heart Disease
The PRIME Study
Objective— We have examined the association between the incidence of coronary heart disease (CHD) and plasma high density lipoprotein (HDL) cholesterol, apolipoprotein A-I (apoA-I), and 2 HDL fractions, lipoprotein A-I and lipoprotein A-I:A-II.
Methods and Results— These parameters were measured in subjects recruited in France and in Northern Ireland in the Prospective Epidemiological Study of Myocardial Infarction (PRIME) Study, a prospective cohort study. Among the subjects free of CHD on entry, 176 in France and 113 in Northern Ireland suffered an ischemic attack (CHD patients) during the 5-year follow-up, whereas 6612 French and 2172 Northern Irish men showed no CHD symptoms (CHD-free subjects). All 4 HDL parameter levels were lower in CHD patients than in CHD-free subjects. After the cohort was divided into quintiles based on the distribution of HDL parameter levels, a significant (P<0.0001) linear increase in relative risk was observed for each HDL parameter level. However, regression logistic analyses showed that apoA-I was the strongest predictor (more powerful than HDL cholesterol) and that lipoprotein A-I and lipoprotein A-I:A-II did not supplement apoA-I in predicting CHD.
Conclusions— Among the parameters related to HDL, apoA-I appears to be the strongest independent risk factor.
Coronary heart disease (CHD) is one of the main causes of premature death in industrialized countries,1,2⇓ and its prevalence has been increasing in developing countries. However, reported CHD mortality rates differ from one industrialized country to another.3 In particular, the attack rates of myocardial infarction (MI) and CHD death present a north-south geographical gradient in Western Europe; for instance, the risk in Northern Ireland is twice as high as the risk in France. Differences in conventional risk factors (total cholesterol, blood pressure, and cigarette smoking) do not fully explain this difference in the incidence between northern and southern countries.4 A multicenter case-control study, Etude Cas-Témoins sur l’Infarctus du Myocarde (ECTIM), comparing MI survivors and CHD-free subjects in Northern Ireland and in France investigated the hypothesis that besides the classic lipid risk factors, such as LDL cholesterol (LDL-C) and HDL cholesterol (HDL-C), other factors, particularly lipoprotein particles, might contribute to an explanation for the geographical contrast in the disease.5
Lipoprotein particles can be characterized by their apolipoprotein content and constitute separate metabolic pools.6 Plasma lipoproteins are composed of a mixture of such particles, the plasma levels of which can be measured by using a combination of polyclonal and monoclonal antibodies.7 Concentrations of LpA-I, which contain apoA-I but not apoA-II, proved to be of interest in comparing MI incident CHD cases with CHD-free control subjects in the ECTIM Study. A high-risk profile, partially characterized by a low LpA-I level, was more frequent in the population of Northern Ireland.5 However, even though case-control studies have shown that low LpA-I is associated with CHD,5,8–12⇓⇓⇓⇓⇓ no prospective study has directly analyzed LpA-I and LpA-I:A-II as risk factors for CHD.
Prospective cohort studies are needed to analyze CHD risk according to etiologic factors (especially biological) on firm methodological bases.13 The Prospective Epidemiological Study of Myocardial Infarction (PRIME study) was set up to investigate the differences in CHD incidence between Northern Ireland and France and their possible determinants.14 Lipid and lipoprotein parameters related to HDL are candidates. A number of prospective cohort studies have already investigated the value of plasma apoA-I levels in predicting CHD,15–21⇓⇓⇓⇓⇓⇓ but only 4 of these have simultaneously determined HDL-C levels15,17,20,21⇓⇓⇓ and, thus, could directly compare the predictive value of HDL-C and apoA-I. All studies concluded that apoA-I was no more predictive of CHD than was HDL-C. Finally, cohort studies have generally analyzed the predictive value of parameters for MI and coronary death, the so-called hard coronary end points, but none has made an evaluation of hard incident CHD cases on the one hand and angina pectoris on the other.
In the present study, therefore, we have studied different issues: (1) HDL parameters (such as HDL-C), apoA-I, and HDL fractions (such as LpA-I and LpA-I:A-II) as risk factors for CHD; (2) their predictive value for hard CHD events or angina; and (3) the proportion of relative risk of CHD incidence between Northern Ireland and France, which could be explained by differences in HDL parameters.
Details of the protocol and conduct of the PRIME study have previously been described.14 For more details, please see the online data supplement, available at http://atvb.ahajournals.org.
A blood sample was drawn after a 12-hour fast into tubes containing EDTA. Plasma was separated by centrifugation at 4°C within 15 minutes at each clinic, kept at 4°C, and sent weekly by air (with the exception of those from the Lille Center) to the Central Laboratory at the Pasteur Institute of Lille. All samples were treated in exactly the same way (regarding delay and temperature) whatever the center. Cholesterol and triglyceride levels were determined by automated enzymatic procedures (Boehringer-Mannheim) adapted to a Hitachi 705 analyzer. Cholesterol was measured in HDL-containing supernatant after phosphotungstate/magnesium chloride precipitation of apoB-containing lipoproteins (Boehringer-Mannheim). ApoA-I was quantified by using commercial reagent by immunonephelometry (Behringwerke). LpA-I was measured by using differential immunoelectrophoresis, as previously described.22 The LpA-I:A-II level was calculated as its apoA-I content by subtracting LpA-I from total apoA-I. The coefficients of variation were 2% and 4% for apoA-I and LpA-I, respectively. LDL-C was calculated according to the Friedewald formula.23
Statistical analysis was carried out by using an SAS statistical package (SAS Institute). Differences between centers were evaluated by ANOVA. The strength of association between baseline parameters and CHD event was assessed by means of logistic regression analysis comparing CHD subjects (incident CHD cases) and control subjects. Theoretically, proportional hazard regression may be considered as more suited to incidence data analysis than logistic regression. However, the annual incidence of CHD events was low (<1%), and the follow-up duration was fixed (5 years). Under these conditions, the 2 methods yield virtually identical results, which were systematically verified in the present study. Standardized β coefficients in the regression analysis were computed as the product of β coefficients by the SD of the corresponding explaining variable. Two models were mainly used for analysis: (1) a simple model with adjustment for age and center only and (2) a multivariate model that included, in addition to age and center, risk factors other than HDL parameters (diabetes, high blood pressure, smoking, and LDL-C and triglyceride levels). Statistical significance was set at P<0.05.
Characteristics of Incident CHD Cases and CHD-Free Subjects
The general characteristics of the incident CHD cases and those free of CHD at the end of follow-up (CHD-free subjects) are shown in Table 1. There were 176 incident CHD cases in France and 113 in Northern Ireland, corresponding to 2.6% and 5.2% of the French and the Northern Irish cohorts, respectively. In all centers, the percentage of smokers was higher in CHD patients than in subjects without CHD. The frequency of diabetes mellitus was ∼2- to 3-fold higher in CHD patients than in CHD-free subjects in both countries. Systolic and diastolic blood pressure levels were significantly higher in incident CHD cases than in CHD-free subjects in both countries, but no difference was noted between Northern Ireland and France. Treatment for high blood pressure was less frequent in Northern Ireland than in France in CHD patients and in CHD-free subjects.
HDL Parameters in Incident CHD Cases and CHD-Free Subjects
HDL-C, apoA-I, LpA-I, and LpA-I:A-II concentrations measured on entry are presented in Table 2. HDL parameters revealed differences between incident CHD cases and CHD-free subjects in each country. Indeed, ANOVA showed that every HDL parameter was higher in CHD-free subjects than in incident CHD cases, irrespective of country. However, HDL-C, apoA-I, and LpA-I levels in Northern Irish incident CHD cases and CHD-free subjects were significantly lower than the levels in the respective French groups, whereas LpA-I:A-II levels were similar in both countries.
The coefficients of correlation between HDL variables were similar in CHD patients and CHD-free subjects. So, in the whole cohort, HDL-C was highly correlated with apoA-I (r=0.76, P<0.0001) and LpA-I (r=0.64, P<0.0001), and apoA-I was correlated with LpA-I (r=0.64, P<0.0001) and LpA-I:A-II (r=0.88, P<0.0001). But the LpA-I/apoA-I ratio was not correlated with apoA-I (r=−0.02) and only moderately with HDL-C (r=0.18, P<0.0001).
HDL Parameters and CHD Risk
Subjects were subdivided into quintiles based on the distribution of levels of HDL-C, apoA-I, LpA-I, and LpA-I:A-II of the CHD-free population. Relative risks were adjusted for age, center, smoking, hypertension, diabetes, LDL-C, and triglycerides. The results are shown in Table 3. Relative risks decreased steadily as each HDL parameter level increased, and the linear trend for each variable was significant (P<0.0001) for each of the 4 comparisons. Moreover, Table 3 shows that the relative risk associated with apoA-I appears lower than that associated with HDL-C, LpA-I, or LpA-I:A-II in each quintile and that the relative risk associated with LpA-I:A-II was lower than that associated with LpA-I.
To assess the predictive value of the various HDL parameters, individual HDL parameters were included in logistic regression analysis after controlling for the lipid (LDL-C and triglyceride) and nonlipid covariates noted above. The standardized β coefficients were −0.230, −0.403, −0.211, and −0.352 for HDL-C, apoA-I, LpA-I, and LpA-I:A-II, respectively. Several multivariate models were then used. These included different HDL parameters in a logistic regression analysis after controlling for the same covariates: model 1, described in Table 4, included HDL-C and apoA-I; model 2 included HDL-C, apoA-I, and LpA-I; and model 3 included HDL-C, apoA-I, and the LpA-I/apoA-I ratio. When apoA-I was included in the models, HDL-C no longer appeared as a risk factor. Models 2 and 3 showed that low LpA-I or a low LpA-I/apoA-I ratio was not significantly related to CHD when apoA-I was included in the model.
HDL Parameters, MI, and Angina
The PRIME Medical Committee classified initial coronary events into 2 categories: (1) MI and coronary death (n=150) and (2) angina (n=139). This distinction gave us the opportunity to assess whether the relationship between HDL parameters and the 2 CHD categories was similar by using logistic regression analyses, including lipid (LDL-C and triglycerides) and other specific covariates, as mentioned above. Separate analyses of individuals with MI and coronary death and individuals with angina showed similar estimates (Table 5), inasmuch as both were related to a lower apoA-I level. ApoA-I was more strongly related to any CHD event than were other HDL parameters, especially HDL-C. Furthermore, neither HDL-C nor LpA-I nor LpA-I:A-II was significantly related to MI or angina when apoA-I was included in the model (data not shown).
Association of HDL Parameters With CHD in Northern Ireland and France
Because CHD incidence in France and in Northern Ireland is clearly different, we went on to analyze separately the relation of HDL parameters to CHD in the 2 countries to evaluate whether HDL parameters contribute to the between-country incidence gradient. Table 6 presents the results of logistic regression analyses performed separately in Northern Irish and French subjects by using the model including HDL-C, apoA-I, and LpA-I. The results were similar in the 2 countries, demonstrating that apoA-I was the only parameter significantly related to the occurrence of CHD in each country. Results of analyses including either HDL-C, apoA-I, and LpA-I or HDL-C, apoA-I, and the LpA-I/apoA-I ratio, after controlling for other lipid and non lipid parameters, were similar in both countries; ie, only apoA-I appeared to be significantly associated with CHD.
The 5-year relative CHD risk between Northern Ireland and France was 1.91.24 To evaluate the proportion of this relative risk explained by the apoA-I difference between the 2 populations, we simply used the logistic regression model that included (apart from apoA-I) LDL-C, triglycerides, and other nonlipid covariates. We found that the mean apoA-I difference (4.5 mg/dL) between countries was associated with an approximate estimate of relative risk of 1.071 (95% CI 1.037 to 1.106), which represented ∼7% of the excess relative risk between the 2 countries. However, this rough estimate is probably underestimated because the effect of the intraindividual variability of apoA-I on the coefficient in the logistic function (attenuation effect) was not taken into account.
Many studies have demonstrated that low levels of HDL-C are associated with an increased risk of CHD.25–27⇓⇓ In the PRIME study, in which baseline blood samples were available, HDL-C was confirmed as a powerful predictor of CHD risk. However, this inverse association was not limited to the cholesterol content of HDL. Indeed, lower levels of apoA-I or of the HDL fractions defined by their apolipoprotein content (ie, LpA-I and LpA-I:A-II) were also associated with CHD risk when they were considered as separate variables in the analysis. Therefore, low apoA-I levels appeared to be the most powerful HDL parameter for predicting CHD in the PRIME study when various HDL parameters were included simultaneously in regression models.
The potential limitation of the statistical analysis is due to high correlations between HDL parameters. Collinearity may create spurious associations when independent variables are included simultaneously in a multivariate model. However, all analyses (ie, those with quintiles, standardized β of logistic regression analyses, which included each HDL parameter, and more complex models involving various subsets of these HDL parameters) were consistent in indicating that low apoA-I levels proved to be the strongest HDL parameter for predicting CHD in the PRIME study. Four prospective studies15–17,21⇓⇓⇓ have compared the predictive value of apoA-I with that of HDL-C levels and concluded that apoA-I was not more predictive of CHD than was HDL-C. The reason for the difference between results in the PRIME study and other prospective studies remains unclear. The first potential explanation is a difference in the laboratory methods even if all methods have been standardized. HDL-C was determined by different apoB-containing lipoprotein precipitants that give similar results,28 but apoA-I was measured by using different methods. Nephelometry using automatic measurement was used in the PRIME study and in the British United Provident Association Study,20 whereas slightly lesser reproducible methods (radial immunodiffusion and electroimmunodiffusion) were used in the 2 other large studies (Physicians’ Health Study17 and Quebec Cardiovascular Study21). That could explain the results in these last 2 studies because the greater the variability in a measure, the lower its discriminating power is likely to be. Differences in methodology are also to be considered. In the Physicians’ Health Study, only 15% of blood samples were drawn after a 12-hour fast. Because the structure of HDL is strongly related to metabolic modifications after a meal, it is possible that low HDL-C levels observed in incident CHD cases were emphasized in the fed state, whereas apoA-I levels were only slightly or not at all modified, therefore leading to a greater discriminating power for HDL-C than for apoA-I. Finally, differences between populations are to be considered. With the exception of the Tromso Study,15 the average age was similar in these prospective studies even if the age range was narrower in the PRIME study than in others. Furthermore, there was a lower percentage of current smokers in the Physicians’ Health Study and a higher percentage of diabetics in the Quebec Cardiovascular Study than in the present study. These differences could modulate HDL-C and apoA-I levels, as we have demonstrated for environmental factors such as alcohol consumption and cigarette smoking.29 Because participants in the PRIME study had lower triglyceride levels than participants included in the 2 other studies,17,21⇓ the finding that apoA-I was a stronger risk marker than was HDL-C could be partly due to the fact that triglyceride levels are more negatively correlated with HDL-C than with apoA-I.
Differences in usual risk factor levels were also noticed between Northern Ireland and France: higher frequency of smokers, lower rate of diabetic and hypertensive subjects, and higher levels of triglycerides in Northern Ireland than in France. Some of these differences may partly explain the higher CHD risk in Northern Ireland than in France,24 but we were able to calculate that this geographic difference could partly be taken into account by the difference in apoA-I concentrations between the 2 populations even though the estimate of 7% explained relative risk is probably underestimated because of the regression-dilution effect.30 Data from the PRIME study29 enabled us to evaluate that 50% of the between-country difference of apoA-I was due to differences in alcohol intake, cigarette consumption, and physical activity; 20% of apoA-I variability was independent of HDL-C. Genetic background was also probably different in Northern Ireland and France. We have already observed between-country differences of apoE genotype frequency and Lp(a) levels,5 which are strongly related to genetic polymorphism.31 So, in contrast to previously published studies, the PRIME study finding (ie, that there is a stronger association between apoA-I and CHD than between HDL-C and CHD) was probably indicative of several causes: differences in methodological, environmental, and genetic factors.
The PRIME study is the first prospective cohort study analyzing the association between incident CHD and directly measured LpA-I and LpA-I:A-II levels and showing that LpA-I and LpA-I:A-II are inversely related to the incidence of CHD. Because apoA-I concentration was the sum of apoA-I content in LpA-I and LpA-I:A-II and because both fractions were inversely associated with CHD risk on the other hand, it is logical that apoA-I should be more strongly associated with CHD risk that any apoA-I–containing lipoprotein. In another prospective study,17 LpA-I but not LpA-I:A-II was significantly associated with CHD risk, but LpA-I and LpA-I:A-II were not measured but estimated from apoA-I and apoA-II plasma concentrations and an arbitrarily fixed apoA-I/apoA-II ratio. As in the Physicians’ Health Study and all case-control studies,5,9–12⇓⇓⇓⇓ we also noted that concentrations of LpA-I were lower in CHD subjects than in CHD-free subjects. Most cross-sectional studies also showed a low concentration of LpA-I:A-II particles in CHD patients compared with control subjects,5,10–12⇓⇓⇓ whereas no difference was noted in only 1 study.9 The finding that LpA-I and LpA-I:A-II appear to have a similar quantitative relation with CHD does not mean that both particles are metabolically equivalent. There is evidence that LpA-I and LpA-I:A-II particles have different metabolic functions, but relatively little is known about their respective roles. Interactions between apoA-I–containing particles and proteins involved in lipoprotein metabolism, such as enzymes, lipid transfer proteins, and receptors, suggest that LpA-I:A-II is less effective than LpA-I in the transportation of cellular cholesterol from peripheral cells to hepatocytes.32 However, there is no direct evidence that cholesterol transport from peripheral tissues to the liver is less effective through LpA-I:A-II than through LpA-I, and the increased risk associated with a decrease of 1 mg/dL in LpA-I or LpA-I:A-II was similar (1.7%) in the PRIME study.
Angina has rarely been analyzed as an end point in prospective studies, because event identification is currently considered less reliable than MI and/or CHD death. The definition of angina as described in the present study (ie, the association of a chest pain with 1 positive finding from a cardiovascular examination, such as angiography, scintigraphy, exercise test, or ischemic ECG change at rest) increases the specificity of the classic clinical diagnosis of angina, which is considered unreliable in epidemiology. In the present study, apoA-I was the strongest HDL parameter associated with both CHD categories (MI/coronary death and angina), and so the same HDL parameters appear to be associated whatever the clinical pattern of the CHD initial event. This implies that even if the pathophysiology of acute coronary event and angina is different, we may hypothesize that low HDL levels appear to be a common contributor to the appearance of arterial lesions at the origin of both types of CHD syndrome.
In conclusion, the PRIME study is the first prospective cohort study in which HDL-C, apoA-I, and HDL subfractions defined as their apolipoprotein content (ie, LpA-I and LpA-I:A-II) have been measured in all participating subjects. Each of these HDL parameters proved to be associated with CHD. However, apoA-I appears to be the strongest independent risk factor whatever the population (French or Northern Irish) and the clinical pattern (MI/coronary death or angina). Until the 1990s, apolipoprotein measurement had not reached its full potential because the lack of standardization and reference methods for apoA-I assay was a limitation to the general application of apoA-I quantification in clinical practice. International standardization of apoA-I33 has made apoA-I concentration a reliable measurement for inclusion in the determination of CHD risk profile. Finally, if all Northern Irish and French subjects with low levels of LpA-I and/or LpA-I:A-II, leading to a low apoA-I level, are prone to coronary events, the geographic comparison shows that the higher risk in Northern Ireland than in France is partly due to a lower LpA-I level in the population.
PRIME Study Group
The PRIME study is organized under an agreement between INSERM and the Merck Sharp & Dohme-Chibret Laboratory, with the following participating Laboratories: the Strasbourg Multinational Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) Project, Department of Epidemiology and Public Health, Faculty of Medicine, Strasbourg, France (D. Arveiler and B. Haas); the Toulouse MONICA Project, INSERM U558, Department of Epidemiology, Paul Sabatier-Toulouse Purpan University, Toulouse, France (J. Ferrières and J.B. Ruidavets); the Lille MONICA Project, INSERM U508, Pasteur Institute, Lille, France (P. Amouyel and M. Montaye); the Belfast MONICA Project, Department of Epidemiology and Public Health, Queen’s University of Belfast, Belfast, Northern Ireland (A. Evans and J. Yarnell); the Department of Atherosclerosis, INSERM U325, Lille, France (G. Luc, J.M. Bard, and J.-C. Fruchart); the Laboratory of Hematology, La Timone Hospital, Marseille, France (I. Juhan-Vague); the Laboratory of Endocrinology, INSERM U326, Toulouse, France (B. Perret); the Vitamin Research Unit, The University of Bern, Bern, Switzerland (F. Gey); the Trace Element Laboratory, Department of Medicine, The Queen’s University, Belfast, Northern Ireland (D. McMaster); the DNA Bank, INSERM U525/SC7, Paris, France (F. Cambien); and the Coordinating Center, INSERM U258, Paris-Villejuif, France (P. Ducimetière, P.Y. Scarabin, and A. Bingham).
The PRIME study is supported by grants from Merck Sharp & Dohme-Chibret (France) and from the Department of Health and Social Services (Northern Ireland). We thank the following organizations, which allowed the recruitment of the PRIME subjects: the Health Screening Centers organized by the Social Security of Lille (Institut Pasteur), Strasbourg, Toulouse and Tourcoing; Occupational Medicine Services of Haute-Garonne; the Urban Community of Strasbourg; the Association Inter-entreprises des Services Médicaux du Travail de Lille et Environs; Comité pour le Développement de la Médecine du Travail; Mutuelle Générale des PTT du Bas-Rhin; and Laboratoire d’Analyses de l’Institut de Chimie Biologique de la Faculté de Médecine de Strasbourg.
↵*Members of the Prospective Epidemiological Study of Myocardial Infarction (PRIME) Study group appear in the Appendix.
Received January 29, 2002; revision accepted May 7, 2002.
- ↵Sans S, Kesteloot H, Kromhout D. The burden of cardiovascular diseases mortality in Europe: Task Force of the European Society of Cardiology on Cardiovascular Mortality and Morbidity Statistics in Europe. Eur Heart J. 1997; 18: 1231–1248.
- ↵Tunstall-Pedoe H, Kuulasma K, Amouyel P, Arveiler D, Rajakangas AM, Pajak A. Myocardial infarction and coronary deaths in the World Health Organization MONICA Project. Circulation. 1994; 90: 583–612.
- ↵Parra HJ, Arveiler D, Evans AE, Cambou JP, Amouyel P, Bingham A, McMaster D, Schaffer P, Douste Blazy P, Luc G, Ducimetière P, Fruchart J-C, Cambien F. A case-control study of lipoprotein particles in two populations at contrasting risk for coronary heart disease: the ECTIM Study. Arterioscler Thromb. 1992; 12: 701–707.
- ↵Obrien T, Nguyen TT, Hallaway BJ, Hodge D, Bailey K, Holmes D, Kottke BA. The role of lipoprotein A-I and lipoprotein A-I/A-II in predicting coronary artery disease. Arterioscler Thromb Vasc Biol. 1995; 15: 228–231.
- ↵Coste Burel M, Mainard F, Chivot L, Auget JL, Madec Y. Study of lipoprotein particles LpAI and LpAI: AII in patients before coronary bypass surgery. Clin Chem. 1990; 36: 1889–1891.
- ↵Syvanne M, Kahri J, Virtanen KS, Taskinen MR. HDLs containing apolipoproteins A-I and A-II (LpA-I: A-II) as markers of coronary artery disease in men with non-insulin- dependent diabetes mellitus. Circulation. 1995; 92: 364–370.
- ↵PRIME Study Group. The PRIME study: classical risk factors do not explain the severalfold differences in risk of coronary heart disease between France and Northern Ireland. Q J Med. 1998; 91: 667–676.
- ↵Lamarche B, Moorjani S, Lupien PJ, Cantin B, Bernard PM, Dagenais GR, Despres JP. Apolipoprotein A-I and B levels and the risk of ischemic heart disease during a five-year follow-up of men in the Quebec cardiovascular study. Circulation. 1996; 94: 273–278.
- ↵Parra HJ, Mezdour H, Ghalim N, Bard JM, Fruchart JC. Differential electroimmunoassay of human LpA-I lipoprotein particles on ready-to-use plates. Clin Chem. 1990; 36: 1431–1435.
- ↵Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem. 1972; 18: 499–502.
- ↵Ducimetiere P, Ruidavets JB, Montaye M, Haas B, Yarnell J. Five-year incidence of angina pectoris and other forms of coronary heart disease in healthy men aged 50–59 in France and Northern Ireland: the Prospective Epidemiological Study of Myocardial Infarction (PRIME) Study. Int J Epidemiol. 2001; 30: 1057–1062.
- ↵Goldbourt U, Medalie JH. High density lipoprotein cholesterol and incidence of coronary heart disease: the Israeli Ischemic Heart Disease Study. Am J Epidemiol. 1979; 109: 296–308.
- ↵Gordon T, Castelli WP, Hjortland MC, Kannel WB, Dawber TR. Diabetes, blood lipids, and the role of obesity in coronary heart disease risk for women: the Framingham Study. Ann Intern Med. 1977; 87: 393–397.
- ↵Warnick GR, Nguyen T, Albers AA. Comparison of improved precipitation methods for quantification of high-density lipoprotein cholesterol. Clin Chem. 1985; 31: 217–222.
- ↵Luc G, Bard J, Evans A, Arveiler D, Ruidavets J, Amouyel P, Ducimetiere P. The relationship between apolipoprotein AI-containing lipoprotein fractions and environmental factors: the prospective epidemiological study of myocardial infarction (PRIME study). Atherosclerosis. 2000; 152: 399–405.
- ↵Clarke R, Shipley M, Lewington S, Youngman L, Collins R, Marmot M, Peto R. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol. 1999; 150: 341–353.
- ↵Utermann G, Menzel HJ, Kraft HG, Duba HC, Kemmler HG, Seitz C. Lp(a) glycoprotein phenotypes: inheritance and relation to Lp(a)-lipoprotein concentrations in plasma. J Clin Invest. 1987; 80: 458–465.
- ↵Escola-Gil JC, Julve J, Marzal-Casacuberta A, Ordonez-Llanos J, Gonzalez-Sastre F, Blanco-Vaca F. ApoA-II expression in CETP transgenic mice increases VLDL production and impairs VLDL clearance. J Lipid Res. 2001; 42: 241–248.
- ↵Marcovina SM, Albers JJ, Henderson LO, Hannon WH. International Federation of Clinical Chemistry standardization project for measurements of apolipoproteins A-I and B, III: comparability of apolipoprotein A-I values by use of international reference material. Clin Chem. 1993; 39: 773–781.