The Correlation and Predictive Power of LDL Cholesterol with Coronary Calcification as Measured by Electron Beam Computed Tomography
PDF Version

ABSTRACT | INTRODUCTION | METHODS | RESULTS
DISCUSSION | TABLES | FIGURES | ACKNOWLEDGEMENTS

ABSTRACT
Aims: The purpose of this study was to examine the correlation and predictive power of LDL cholesterol for calcified atheromatous disease as measured by electron beam computed tomography (EBCT).

Methods & Results: Six-thousand one-hundred and ninty-nine subjects underwent EBCT of their coronary arteries, serum lipid testing, body fat determination and assessment of health status by questionnaire. Associations between coronary calcification and predictor variables were assessed by Spearman rank correlation and analysis of variance. Predictive power of LDL cholesterol for calcified atherosclerotic plaque was determined by multivariate logistic regression.

The correlation between LDL cholesterol and plaque score was very modest (r = 0.055, p < 0.001). Men, tobacco smokers or hypertensive subjects had nearly twice as much calcified plaque. Diabetic subjects had nearly 3 times the amount of plaque as those who were not diabetic. Results of the multivariate logistic regression revealed that LDL cholesterol is a modest but significant predictor of coronary plaque. After adjusting for age, gender and HDL cholesterol, the odds of plaque formation was 1.05 times higher for each 10 mg/dL increase in LDL cholesterol (p < 0.001).

Conclusions: LDL cholesterol is weakly correlated with and predictive of calcified atherosclerotic plaque burden as measured by EBCT.

Top of Page

INTRODUCTION

LDL cholesterol is an established component of atherosclerotic plaque. Biochemical research has provided evidence that low-density lipoprotein (LDL) cholesterol can promote atherosclerotic calcification of vascular cells. This effect was found to be due to products of lipid oxidation and not a function of native LDL or its concentration in serum. Overall, 45% of patients with calcified plaque as measured by electron beam computed tomography (EBCT) have an LDL < 130 mg/dL. Such patients are at increased risk for future cardiac events but would not be treated based on the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III guidelines. Furthermore, the Heart Protection Study found a consistent reduction in risk for future events irregardless of the LDL cholesterol level. Some reports have shown that the use of cholesterol indices as a screening technique for cardiac events is neither highly sensitive or specific.

In addition to cholesterol debris, inflammatory cells and fibrotic tissue, calcium accumulates in plaque from the early stages of atheroma development. The use of coronary calcification as measured by EBCT as a means of stratifying patients for risk of future cardiac events has been studied. The results of these studies indicate that coronary calcification may be a useful tool in predicting future risk for individual patients.
Another potential use of this assessment tool is as a surrogate marker for coronary atherosclerotic disease. Histopathologic research has shown a high correlation between coronary calcification and total atherosclerotic plaque burden, or total amount of plaque present in the coronary arteries. Due to positive remodeling in the coronary vessel wall, the risk for acute coronary events is directly related to the plaque burden and not to percent stenosis of the artery. In this context, coronary calcification can be studied in the framework of traditional cardiac risk factors such as LDL cholesterol using the calcium score as the outcome of interest in the place of cardiac events. Review of the literature reveals that studies using this perspective are limited.

The purpose of this study was to examine the relationship between LDL cholesterol and coronary calcification in an ambulatory setting. Specifically, this study assessed the correlation and predictive power of LDL cholesterol for calcified atheromatous disease as measured by EBCT. Based on the preceding discussion, we hypothesized that LDL cholesterol would be minimally predictive of coronary calcification.

Top of Page

METHODS

Subjects
From October 1999 to February 2002, 8,101 consecutive patients who presented for preventive medicine services at a private, university affiliated disease prevention center in San Diego, CA, were eligible for initial enrollment in the study. Patients evaluated at the center more than once were included with their original study only. All patients who were taking lipid altering medications were eliminated from the study. These medications included HMG CoA reductase inhibitors, niacin, oral chelation therapies, fibrates and hormonal therapies. Subjects with a history of coronary heart disease related surgeries (i.e. stent placement, coronary artery bypass graft) were also excluded. Individuals with triglyceride values greater than 400 mg/dL were unable to be included in the study due to the inability to calculate LDL values using the Friedewald formula. Of the initial sample, a total of 6,199 subjects were available for analysis in the study. Most patients were self referred or referred from their local doctors and were seeking preventive health information as a supplement to their routine medical care.

Imaging
All patients underwent imaging with an Imatron C-150 scanner. Images were obtained with 100-ms scan time. Using 3 mm slices starting at the level of the carina and proceeding to the level of the diaphragm, approximately 40 to 45 slices of each subject's heart were obtained. Tomographic imaging was electrocardiographically triggered at 40 or 65% of the R-R interval, depending on the subject's heart rate. Coronary calcification was defined as a plaque of >= 2 pixels (area = 1.37 mm2) with a density of greater than or equal to 130 Hounsfield unites (HU). Quantitative calcium scores were calculated according to the method described by Agatston et al. Coronary calcium scoring was performed by either a physician or computed tomography technician with specific training for the methodology described above. In addition to a calcium score, this methodology produces a total plaque volume and calculates the total number of lesions present in the coronary arteries.

Laboratory
All patients underwent random serum lipid analysis using the Cholestex LDXÒ system. In brief, capillary whole blood specimens were obtained by fingerstick with the subject in the seated position using a 35 ml lithium heparin-coated capillary tube. Body mass index was calculated with the patient clothed without shoes. Body fat measurement was conducted using the OmronÔ HBF-300 body fat analyser.

Statistical Analysis
The outcome variables for this study include coronary calcium score, coronary plaque volume and total number of coronary lesions. The primary exposure variable was LDL cholesterol. The covariates included HDL and total cholesterols, triglycerides, age, gender, body mass index, percent body fat, diagnosis of hypertension or diabetes mellitus, current and past tobacco use, stress level and history of premature coronary heart disease in parent or sibling. LDL, HDL, total cholesterol, triglycerides, age, body mass index and percent body fat were analyzed as continuous variables. The remaining predictor variables were dichotomized except for the stress variable which was categorized into 4 levels; none, mild, moderate, severe. Historical variables were obtained via patient self-report. Premature coronary heart disease was defined as cardiac event before the age of 55 for
men and 65 for women.

Univariate associations for the continuous variables were calculated using the Spearman rank correlation. Comparison of group means for categorical variables was conducted using a one-way ANOVA. Tukey's test was used for multiple comparisons of the stress variable. Transformation of coronary calcium score failed to normalize the distribution of this variable. Coronary calcium scores were therefore dichotomized for use in logistic regression with the 2 categories being a score of 0 and a score greater than 0 (i.e. presence or absence of plaque). Univariate logistic regression was conducted for all predictor variables. Variables that were significantly associated with the outcome at a p-value of less than or equal to 0.10 were included for multivariate logistic regression. Stepwise regression was performed to construct the most parsimonious multivariate model. Predictor variables that changed the odds ratio for LDL cholesterol by more than 5% were retained in the final model. A significance level of 0.05 was used for all analyses. All statistical analyses were conducted using SAS version 8.0Ò statistical package. The study protocol complies with the Declaration of Helsinki and was approved by the committee for protection of human subjects (CPHS) at San Diego State University.

Top of Page

RESULTS

The characteristics of the subjects analyzed are presented in Table 1. The mean and median plaque scores, volumes and number of lesions differed significantly due to the non-normal distribution of these variables. For skewed data such as these, the median value tends to be a better measure of central tendency. The range of values for plaque scores, volume and lesions was from 0 to 6,523, 0 to 5,246 and 0 to 96, respectively. The sample has a higher percentage of men than women (62 vs. 38%). Only 2.7% of the sample related a diagnosis of diabetes mellitus which is lower than the current national average of 5.9%.

Table 2 provides the correlations between the continuous predictor variables and the outcomes. Total cholesterol was the only variable that was not significantly associated with any of the outcomes. The largest linear correlation was found for age and total plaque score (r = 0.4143). Age was also significantly correlated with plaque volume (r = 0.412) and total number of coronary lesions (r = 0.399). The next highest correlation was between HDL cholesterol and total lesions, which showed an inverse relationship (r = -0.198). The relationship between HDL and plaque score and volume was similar in magnitude and direction. The correlation between LDL cholesterol and plaque score was very modest (r = 0.055). Figure 1 shows the nature of this relationship.

Univariate associations between plaque score, volume and lesions and categorical predictor variables are shown in Table 3. Significant associations were found between all predictor variables except for being a current smoker and having a family history of premature coronary heart disease in a parent or sibling. For men, the average score, volume and number of lesions was double that of women. Individuals with a diagnosis of hypertension had nearly twice the score, volume and number of lesions as those who were not hypertensive. Former smokers were found to have a similar relationship. Diabetic subjects had nearly 3 times the amount of plaque as those who were not diabetic.
Current smokers were not found to have a significantly different amount of plaque to those who were nonsmokers. On subanalysis, current smokers were found to be significantly younger. There were no differences between nonsmokers and smokers with respect to gender, diagnosis of hypertension or diabetes, LDL or HDL cholesterol levels, BMI or percent body fat.

There was a significant inverse relationship between stress level and coronary plaque. Further analysis of this relationship found a significantly higher percentage of women in the higher stress categories. Lower stress categories were also significantly older than higher stress levels. Only the severe versus moderate stress level comparison was not statistically significant with respect to age. There were no differences found for LDL and HDL cholesterol levels and stress category.

The univariate predictive ability of all of the independent variables is shown in Table 4. All of these variables were significant predictors of the plaque score except for being a current smoker and having a positive family history for premature coronary heart disease in a parent. The largest odds ratio was found for being diagnosed with diabetes. These subjects had an odds that was more than 4 times higher for having any plaque compared to nondiabetics. Being male or having a diagnosis of hypertension equated to more than 2 times the odds of having any plaque compared to women or not being hypertensive, respectively. Subjects whose LDL cholesterol was 10 mg/dL higher had an odds that was 1.04 times higher for having any plaque. Conversely, having an HDL cholesterol that was 5 mg/dL higher reduced the risk of plaque development by 10%.

Results of the multivariate logistic regression revealed that LDL cholesterol is a modest but statistically significant predictor of coronary plaque development. After adjusting for age, gender and HDL cholesterol, the odds of plaque formation was 1.05 times higher for each 10 mg/dL increase in LDL cholesterol. The results of the multivariate logistic regression are shown in Table 5. The final multivariate model was found to be a good fit using the Hosmer-Lemeshow goodness-of-fit test.

Top of Page

DISCUSSION

Epidemiologic studies have repeatedly shown an association between LDL cholesterol levels and an increased risk for subsequent coronary heart disease events. Odds ratios for LDL cholesterol and the prediction of coronary disease events typically cluster around a value of 2. Data such as this and the results of clinical trials that have shown consistent benefit from lowering cholesterol indices have caused the NCEP to recommend primary and secondary prevention strategies based ultimately on LDL cholesterol levels. Under the ATP III guidelines, the percent increase in patients being treated compared to ATP II would be 140% overall, 157% among males, 122% among females, 131% among those 65 years old, and 201% among those <45 years old.

Recent clinical trials have shown that reduction in mortality by HMG CoA reductase inhibitors may be independent of LDL cholesterol level. The LIPID trial had an average LDL concentration of 150.3 and the WOSCOPS trial had an average LDL of 189.2, yet they both had total mortality reduction rates of 22%. There was a difference with respect to coronary heart disease mortality with the WOSCOPS trial having a rate reduction of 34% while the LIPID trial showed a reduction of 24%. On the other hand, the LIPID trial had a greater reduction in other cardiovascular mortality as compared to LIPID (28 vs. 21% respectively). The recently completed Heart Protection Study found similar results with a risk reduction of approximately 25% regardless of serum LDL concentration.

There is evidence that the reduction in mortality in patients treated with statin class medications may be due to modification of other risk factors such as plaque stabilization, improvement in endothelial function and anti-inflammatory properties. A recent report found that atorvastatin reduced oxidative stress in vitro and in vivo. The authors concluded that statins may contribute to the vasoprotective effects. Simvastatin therapy has also been shown to result in a significant decrease in median C-reactive protein levels at 6 weeks. Homocysteine has recently been found to induce 3-hydoxy-3-methyglutaryl coenzyme A reductase and intracellular cholesterol accumulation in vascular endothelial cells in a redox-dependent fashion. In this study, simvastatin prevented the intracellular accumulation of cholesterol and reversed the homocysteine-induced suppression of nitric oxide production.

The results presented from the current study support the notion that the absolute LDL cholesterol concentration is not the determining factor for plaque development and potential for subsequent cardiac events. This is further supported by biochemical research that has shown that the native form of LDL is not associated with atherosclerotic calcification.2 Furthermore, Agmon et al demonstrated a lack of predictive association between LDL cholesterol and aortic atherosclerosis using transephageal echocardiography. Our results have also been reproduced in another independent center.

Since the oxidized form of LDL appears to be one of the culprit components of atheroma development, the decrease in events seen in lipid lowering clinical trials may be due to anti-inflammatory effects in addition to lowering the levels of both oxidized and non-oxidized LDL cholesterol. A more cost-effective form of treatment would target the injurious form of LDL or those individuals with increased susceptibility to oxidative and inflammatory stress. Coronary calcification as measured by EBCT can be used to identify such a subset of individuals. High dose folic acid therapy has been shown to improve endothelial function independent of serum homocysteine reduction. Research into other substances such as flavinoids that reduce the oxidation of LDL cholesterol are underway.

An interesting yet presumably paradoxical finding of this study is that of patients with a family history of premature coronary heart disease having a lower calcium score than those who did not have such a history. Although this difference was not statistically significant, these results suggest that when patients are aware of their risk for disease development, changes may occur that can lead to reduction in the extent of disease compared to those who have no such information. This would be a fascinating area of exploration from a health behavior perspective.

Limitations of this study include the cross-sectional design and using random serum lipid measurements. The former reduces the ability to assess true casuality between the predictor variables and the outcomes. The latter will result in values that are lower than fasting levels. The use of random lipid levels has been associated with a 7 and 3% decrease in LDL cholesterol at 3 and 5 hours postprandially, respectively . However, since the LDL variable was used in the continuous form and patients were not categorized based on LDL level, misclassification bias will not occur. In effect, even if the LDL values were 3 - 7% higher, the correlation would remain modest and nonlinear.

It is recommended that longitudinal studies comparing serum LDL levels and coronary calcification development be conducted in order to test the validity of our findings. Interventional trials are currently being conducted to examine the effect of medications on plaque development. It is also recommended that determining the length of time a patient has smoked and how much they have smoked during that time would improve the assessment of the smoking variable with respect to plaque development.

Top of Page

TABLES

Table 1 – Descriptive Statistics

¹CHD:  cardiac event prior to age 55 for men, 65 for women

Table 2 – Univariate Associations for Continuous Variables