INTRODUCTION

Cardiovascular disease (CVD) is the leading cause of death in older women in developed societies, and the number of deaths is expected to increase by 30% from the 1990s to 2020.1 Ischemic heart disease becomes more common after the menopause, and could therefore be partly related with the associated metabolic changes, in particular increases in concentrations of total cholesterol, low-density lipoprotein (LDL) cholesterol and triglycerides, and reductions in the concentration of high-density lipoprotein (HDL) cholesterol, which can contribute to the increased cardiovascular risk in women aged 60 years or older.2,3

For several years now low concentrations of HDL cholesterol have been known to be associated with an increased risk of developing CVD, especially coronary heart disease.4,5 This increased risk remains in persons treated with statins,6 even when very low concentrations of LDL cholesterol are achieved.7

Thus, the low concentration of HDL cholesterol may be a conditioning factor for the persistence of the cardiovascular risk, at least in certain patients, in spite of treatment with statins. In addition, HDL cholesterol seems to have a greater impact in women. In the Framingham study, the risk of developing myocardial infarction associated with a low HDL cholesterol concentration was far greater in women than in men,8 and in 4 American studies an increase in HDL cholesterol produced a greater cardiovascular benefit in women than in men.9

Only a few series have studied the prevalence of low concentrations of HDL cholesterol in Spanish women and the transverse association between HDL cholesterol and CVD. The aim of this study was to evaluate the prevalence of low concentrations of HDL cholesterol and the association between the HDL cholesterol concentration and the prevalence of CVD in a wide sample of women aged ≥55 years with hypertension in cluded in the RIMHA study (Riesgo de Ictus en las Mujeres con Hipertensión Arterial— Risk of stroke in women with hypertension).10

METHODS

This study is an analysis of the RIMHA study, which is a cross-sectional, multicenter, epidemiological study undertaken in primary care centers in Spain.10 The study was approved by an independent clinical research ethics committee and the data were collected between June and December 2004. Each researcher collected information about the first 8 non-selected hypertensive women who attended their office, after being duly informed and providing signed consent to participation in the study.

The primary endpoint of the study was to estimate the risk of stroke in this particular population, and the results have already been published.10 Amongst the secondary aims of the study was the evaluation of other cardiovascular risk factors. The study included women who had had hypertension for at least 6 months (according to the diagnosis in their clinical histories) and were aged ≥55 years. Patients were excluded if they refused to give informed consent or had a disease or psychological disorder advising against their participation. Information was gathered about demographic and anthropometric data, cardiovascular risk factors and diseases.

The blood pressure was measured with an OMRON® M5 automatic device following the recommendations in force at the time.11 All the patients underwent an electrocardiogram. Left ventricular hypertrophy was considered to be present when the Cornell voltage criteria for women were fulfilled.12 A woman was considered to be diabetic if this diagnosis was reflected in the clinical history or if her baseline fasting glucose level was above 125 mg/dL.

The biochemical data were obtained from a blood test undergone within the 6 months prior to data collection. If this had not been done, or if the patient had started a new treatment since then that could affect the result, a new blood test was carried out. The glomerular filtration rate was calculated with the shortened MDRD (Modification of Diet in Renal Disease) equation.13

The measurement of the lipids was done at the usual laboratory for each office, not centralized. In general, the total cholesterol and the triglyce-rides were measured by enzymatic assays, the HDL cholesterol by a direct in vitro enzymatic method, and the LDL cholesterol was calculated from the Friedewald formula. To define the HDL cholesterol concentration as low, we used a concentration of <46 mg/dL, a concentration that the European Guidelines on Cardiovascular Prevention define as a marker of increased cardiovascular risk in women.14

Statistical Analysis

The sample size was calculated according to the primary endpoint of the study.10 According to the likelihood of the expected risk of stroke taken from the Framingham study, with a 95% confidence interval and a maximum sampling error of 0.41%, we estimated a sample of 9200 women.10

The qualitative variables are presented as their frequency dis tribution. The quantitative variables are summarized as their mean and standard deviation (SD). For the comparison of the means between the groups, we used the Student t test for independent groups, and for the association between categorical variables we used the c2 test or Fisher's exact test. HDL cholesterol was managed as a binary variable ("normal" or "low" HDL cholesterol) to make the comparison between the two groups, and then in quintiles according to the HDL cholesterol concentration.

The variables associated with a low concentration of HDL cholesterol were evaluated with a multivariate logistic regression model that included demographic and metabolic variables and CVD. Another two models were used to study the association between low concentrations of HDL cholesterol and CVD, with CVD being the dependent variable; the variable HDL cholesterol was again binary ("normal" or "low" HDL cholesterol) and, alternatively, as quintiles of HDL cholesterol. The adjustment variables selected were statins and those variables that could affect the presence of CVD. These models were also repeated taking as the dependent variable the presence of heart disease (excluding cerebrovascular disease). The odds ratios (OR) and their 95% confidence intervals (CI) are given. In all the comparisons, a null hypothesis was rejected if the alpha error was less than 0.05. The data were processed using the statistical program SPSS 13.0.

RESULTS

Population Characteristics

In total, 11 042 women (85.8% of the total sample) had a complete lipid profile available (mean age, 67.9 years; 20% with established CVD) (Table 1). No significant differences were found between the women included and those excluded from the analysis.

HDL Cholesterol Concentration and Prevalence of Low HDL Cholesterol

The mean (SD) concentration of HDL cholesterol was 55.9 (14.6) mg/dL. This concentration was lower in the diabetic women (53.5 [14.7] vs 56.9 [14.5] mg/ dL in the non-diabetic women; P<.001), in women with prior CVD (54.3 [15.7] vs 56.3 [14.3] mg/dL in women without CVD; P<.001) and in women treated with statins (55.2 [15] vs 56.5 [14.5] mg/dL in the non-treated women; P<.001).

The prevalence of low concentrations of HDL cholesterol (<46 mg/dL) was 24.3% (95% CI, 23.5-25.1). The prevalence was higher in the diabetic women as compared with the non-diabetic women (29.9% vs 21.9%; P<.001), in the women with CVD (30.1% vs 22.8% in the women without CVD; P<.001) and in the women treated with sta tins as compared with the women not treated with statins (26.9% vs 22.2%; P<.001). For a cut-off point of <50 mg/dL, the prevalence was 36% (95% CI, 35.1-36.9).

The prevalence was greater for the age range of 55-69 years (26.9%, 26.2%, and 25.8%, res pectively, in the women younger than 60, aged 60-64, and 65-69 years), and lower in the older women (22.5%, 20.9%, and 20.6%, respecti vely, in the women aged 70-74, 75-79 years, and 80 years or over; c2 for the linear trend <0.001). In each age range the prevalence was greater in diabetic women and in women with CVD (Figures 1 and 2) (c2 for the linear trend <0.001 in all cases, except in the women with diabetes, P=.004).

Figure 1. Percentage of hypertensive women with high-density lipoprotein (HDL) cholesterol concentrations <46 mg/dL, stratified according to age and the presence or absence of diabetes mellitus.

Figure 2. Percentage of hypertensive women with high-density lipoprotein (HDL) cholesterol concentrations <46 mg/dL, stratified according to age and the presence or absence of cardiovascular disease.

Among the women without CVD, the prevalence of low concentrations of HDL cholesterol was similar in those with and without treatment with diuretics or beta-blockers (22.9% and 22.8%, respectively; P=.919), whereas in the women with CVD the prevalence was numerically superior in the treated women, though the difference was not statistically significant (31.6% vs 27.8%; P=.062). In the non-diabetic women, the prevalences were 21.9% and 22% (without and with treatment with diuretics or beta-blockers; P=.894), and for the diabetic women the figures were 27.8% and 31.6% (without and with treatment with diuretics or beta-blockers, respectively, P=.032).

Prevalence of High Triglyceride Levels

Triglyceride levels were ≥150 mg/dL in 38.2% of the patients, a prevalence that was greater in the women with low concentrations of HDL cholesterol (56.5% vs 32.3% in the women with normal HDL cholesterol; P<.001), in the diabetic women (48.3% vs 34%; P<.001) and in the women with CVD (44.7% vs 36.5%; P<.001).

Characteristics Related With Low HDL Cholesterol Concentrations

Table 1 shows the characteristics of the women with normal and low HDL cholesterol concentrations. The group with low HDL cholesterol concentrations contained a greater percentage of women who were smokers, or who had obesity, a large abdominal circumference and diabetes me llitus, and a worse blood pressure control. In the women with low HDL cholesterol concentrations, the concentration of total cholesterol was slightly lower, that of LDL cholesterol slightly higher and that of triglycerides higher. The prevalence of high concentrations of triglycerides was much greater in the women with low HDL cholesterol concentrations. Finally, the women with low HDL cholesterol concentrations had a greater prevalence of left ventricular hypertrophy, a low glomerular filtration rate, overall CVD and heart disease.

The variables related with low concentrations of HDL cholesterol were evaluated in a multivariate model with the following variables: age, menopause, body mass index, waist circumference, diabetes mellitus, smoking and CVD (Table 2). The presence of low concentrations of HDL cholesterol was associated with excess weight, diabetes, smoking and CVD and, inversely, with age. In the women without CVD, the related variables were the same and the magnitude of the association was si milar (Table 2). The inclusion in the multivariate models of treatment with diuretics or beta- blockers did not modify the associations, and the presence of a low concentration of HDL cholesterol was not related with these treatments.

Relation Between Low HDL Cholesterol Concentrations and Cardiovascular Disease

The prevalence of overall CVD was greater in the women with low HDL cholesterol concentrations as compared with those with HDL cholesterol concentrations in the range of normality (24.7% vs 18.4%; P<.001) (Table 1). After excluding cerebrovascular disease, the prevalences of heart disease were 20.6% (women with low HDL cholesterol concentrations) and 14.5% (women with normal HDL cholesterol concentrations; P<.001). Table 3 shows the OR of CVD and heart disease in the women with low HDL cholesterol concentrations compared with the women with normal HDL cholesterol concentrations obtained from the multivariate models. The first model included the variables age; smoking; diabetes; body mass index, categorized as normal, overweight or obese; blood pressure con trol; LDL cholesterol; treatment with statins and HDL cholesterol concentration (low or normal). There was an independent association between a low HDL cholesterol concentration and CVD (OR of CVD in women with a low HDL cholesterol concentration =1.42; 95% CI, 1.26-1.60, as compared with women with a normal HDL cholesterol concentration; P<.001). The OR of heart disease, excluding cerebrovascular disease, was similar (Table 3). The inclusion in this model of triglycerides (normal or raised) did not modify the relation between low HDL cholesterol concentrations and CVD, although there was a slight relation between triglycerides ≥150 mg/ dL and CVD (OR=1.14; 95% CI, 1.02-1.27).

The second model included the same variables together with left ventricular hypertrophy and glomerular filtration rate (normal or reduced). The associations between low HDL cholesterol concentrations and CVD and between low HDL cholesterol concentrations and heart disease remained significant (Table 3). Nor did the inclusion of triglycerides change the association between low HDL cholesterol concentrations and CVD, but in this second model no asso ciation was seen between high concentrations of triglycerides and CVD.

Relation Between HDL Cholesterol Concentration and Cardiovascular Disease

To evaluate the association between the concentration of HDL cholesterol and CVD, the population was divided into the following quintiles of HDL cholesterol: <45 mg/dL (first), 45-50 mg/dL (second), 51-58 mg/dL (third), 59-66 mg/dL (fourth), and >66 mg/dL (fifth). The prevalence of CVD was highest in the lowest quintile of cholesterol (25.5%) and lowest in the higher quintiles of HDL cholesterol (c2 for the linear trend, <0.001). The prevalence of heart disease showed the same linear trend, as did the prevalence of each individual entity, except for stroke (Table 4).

Finally, the association between the con centration of HDL cholesterol and the prevalence of CVD or heart disease was analyzed using the same multivariate models, introducing the HDL cholesterol in quintiles. Taking as the reference the women with the highest HDL cholesterol concentration (fifth quintile, >66 mg/dL), the OR of CVD and heart disease increased progressively with lower concentrations of HDL cholesterol (Table 5). Compared with the women with the highest concentrations of HDL cholesterol (>66 mg/dL), those with the lowest HDL cholesterol concentration (<45 mg/dL) had an adjusted OR of CVD that was 60% greater (OR=1.59; 95% CI, 1.35-1.88; P<.001). The addition of left ventricular hypertrophy and glomerular filtration rate barely modified the relation (Table 4), which began to be significant for an HDL cholesterol concentration <58 mg/dL.

DISCUSSION

One out of every 4 women with hypertension aged ≥55 years had a low concentration of HDL cholesterol, and this was associated with a greater prevalence of overall CVD, and heart disease in particular. The varia bles associated with a low HDL cholesterol concentration were excess weight, diabetes, smoking and a history of CVD. In spite of the effect of diuretics and beta-blockers on the lipids, we found no relation between treatment with these drugs and the presence of low HDL cholesterol concentrations, probably because, in the context of the effect of multiple variables, their influence is lower.

Most studies that have reported the prevalence of low HDL cholesterol concentrations in our setting have done so in the context of the metabolic syndrome (cut-off point in women of <50 mg/dL). Given this, the studies have reported prevalences of low HDL cholesterol concentrations of 34.4% in patients with CVD (men, 27.8%; women, 52%),15 of 23.3% in hypertensive persons aged over 55 years (men, 13.3%; women, 31.1%)16 and lower in persons of working age, although with a prevalence >30% in women.17 Using these same criteria, the prevalence in patients with dyslipidemia was 35.4% (39% in women) in a European study,18 whereas in the United States, in the NHANES study, it was 37.1% (39.3% in women).19

We found a prevalence of low HDL cholesterol concentrations of 24.3% for a cut-off point of <46 mg/dL and 36% for a cut-off point of <50 mg/ dL, in line with earlier studies. The prevalence was higher in the women who had a greater cardiovascular risk (in diabetics and women with a history of CVD). Although persons living in Mediterranean areas have higher concentrations of HDL cholesterol, probably related with their dietary habits,20 specific groups with a high cardiovascular risk, such as those described, have a considerably higher prevalence of low HDL cholesterol concentrations.

The association between low HDL cholesterol concentrations and CVD or heart disease was independent of other risk factors. Moreover, the association between the HDL cholesterol concentration and CVD or heart disease was also seen for a few HDL cholesterol concentra tions situated within the range of normality (in particular, below 58 mg/ dL) where the OR of CVD increased significantly after adjusting the model for all the risk factors and target organ damage. The inverse continuous relation between the HDL cholesterol concentra tion and the risk of CVD has been described in follow-up studies such as the Framingham study,4 particularly in women. In the women of the Framingham study the adjusted risk of having a myocardial infarction in the lowest quartile of HDL cholesterol (<47 mg/ dL) was 6 times greater than in the top quartile (≥67 mg/dL), with the risk increasing much more than that for the men.8

The cardiovascular risk associated with low concentra tions of HDL cholesterol should be interpreted in the context of the high prevalence of the metabolic syndrome in the group of hypertensive women with low HDL cholesterol concentrations (91.4%) (Table 1). Many studies have shown that this particular group of metabolic anomalies increases the risk of developing CVD21-23 and, in addition, the low HDL cholesterol concentration usually coincides with high concentrations of triglycerides. In our study, the in clusion in the models of triglycerides did not modify the relation between low HDL cholesterol concentrations and CVD; in fact, the relation between triglycerides and CVD was weaker, and even disappeared when target organ damage was added to the model. Whereas a low HDL cholesterol concentration is a well-established cardiovascular risk factor, studies on triglycerides and CVD have found less consistent relations, although postprandial hypertriglyceridemia is emerging as an independent risk factor, especially in women.24,25

The protective properties of HDL cholesterol are well-known. As well as participating in the inverse transport of cholesterol for its hepatic elimination, HDL particles possess antioxidant, anti-inflammatory and antithrombotic properties that protect against the development of atherosclerosis and CVD.26-29 A combined analysis of 4 follow-up studies showed that an increment of 1 mg/dL in levels of HDL cholesterol was associated with a reduction in the incidence of coronary heart disease of 2% in men and 3% in women.9 Even so, the evidence that increasing HDL cholesterol concentrations by the use of drugs can reduce cardiovascular risk is modest. Earlier studies with gemfibrozil30 and with niacin (the most effective drug for increasing HDL cholesterol)31 indicate that the cardiovascular benefits of these drugs are closely related with the increase in HDL cholesterol. A recent meta-analysis of 23 clinical trials found that the trials in which LDL cholesterol decreased at the same time as HDL cholesterol increased were those that achieved the greatest reduction in cardiovascular risk.32 Large clinical trials are currently in progress involving patients with a high cardiovascular risk in order to assess whether the addition of fenofibrate33 or extended-release niacin34,35 to statins reduces the incidence of cardiovascular complications. These clinical trials may determine whether the pharmacologic increase in HDL cholesterol reduces the incidence of CVD in patients already receiving statins, and what is the most suitable concentration of HDL cholesterol to achieve with drugs in high cardiovascular risk patients.

Several limitations should be considered concerning this study. The first is its cross-sectional nature, which prevents the establishment of a cause-effect relation between the associations found. The second limitation concerns the sample selection method, which was consecutive and not random, resulting in the greater likelihood of inclusion of patients with disease processes. Thirdly, the original study did not collect data about treatment with fibrates or physical activity, which may slightly increase HDL cholesterol concentrations, and we were not therefore able to analyze their importance in the multivariate model. However, these limitations should have no impact on the relation seen between HDL cholesterol concentrations and the prevalence of CVD. Finally, the analyses were not performed at a centralized laboratory. We cannot therefore rule out the possibility of a certain degree of variability in the measurements or that this may have influenced the results. Nevertheless, we believe that the sample size analyzed and the analysis methods used strengthen the study, and that the results complement the information from follow-up studies and clinical trials, as they highlight the problems faced by physicians in their daily clinical practice. In spite of having been undertaken in a sample of women with hypertension, the prevalence of hypertension is so high in the age group studied36,37 that the results of this study can be applied to most women of a similar age.

CONCLUSIONS

One out of every 4 hypertensive women aged 55 years or over seen in primary care offices had low HDL cholesterol concentrations. The prevalence of low HDL cholesterol concentrations was greater in women with an increased cardiovascular risk and was independently associated with the presence of CVD and heart disease. Moreover, the inverse association between HDL cholesterol concentrations and the prevalence of CVD was found at figures currently considered to be normal. The results of large ongoing clinical trials with HDL cholesterol-raising drugs should better determine the effect on CVD of pharmacologically increasing HDL cholesterol concentrations, and the attitude to follow in patients with low HDL cholesterol concentrations, particularly in those with a high cardiovascular risk.

ACKNOWLEDGMENTS

The RIMHA study was supported by the Sociedad Española de Hipertensión-Liga Española para la Lucha contra la Hipertensión Arterial (SEH-LELHA) and the Socie dad Española de Medicina Rural y Generalista (SEMER GEN), and was financed by a research grant from Merck Sharp & Dohme de España. The authors of the manuscript wish to thank the study researchers for their participation.

ABBREVIATIONS
CVD: cardiovascular disease
HDL: high-density lipoprotein
LDL: low-density lipoprotein

The present study was financed with a research grant from Merck Sharp & Dohme de España. Two of the authors (Luis Cea-Calvo and Verónica Inaraja) work full time in the Clinical Research Department of Merck Sharp & Dohme de España.

Correspondence: Prof. A. Coca.
Unidad de Hipertensión Arterial. Hospital Clínic. Villarroel, 170. 08036 Barcelona. España.
E-mail: acoca@clinic.ub.es

Received October 7, 2008.
Accepted for publication May 26, 2009.