Although there appears to be a role for statins in reducing cerebrovascular events, the exact role of different lipid fractions in the etiopathogenesis of cerebrovascular disease (CVD) is not well understood. A secondary analysis of data collected for the placebo arm (n = 2,078) of the Cholesterol and Recurrent Events (CARE) trial was performed. The CARE trial was a placebo-controlled trial aimed at testing the effect of pravastatin on patients after myocardial infarction. Patients with histories of CVD were excluded from the study. A Cox proportional-hazards model was used to evaluate the association between plausible risk factors (including lipid fractions) and risk for first incident CVD in patients after myocardial infarction. At the end of 5 years, 123 patients (6%) had incident CVD after myocardial infarction (76 with stroke and 47 with transient ischemic attack). Baseline non–high-density lipoprotein (HDL) cholesterol level emerged as the only significant lipid risk factor that predicted CVD; low-density lipoprotein cholesterol and HDL cholesterol were not significant. The adjusted hazard ratios (adjusted for age, gender, hypertension, diabetes mellitus, and smoking) for CVD were 1.28 (95% confidence interval [CI] 1.06 to 1.53) for non-HDL cholesterol, 1.14 (95% CI 0.96 to 1.37) for low-density lipoprotein cholesterol, and 0.90 (95% CI 0.75 to 1.09) for HDL cholesterol (per unit SD change of lipid fractions). This relation held true regardless of the level of triglycerides. After adjustment for age and gender, the hazard ratio for the highest natural quartile of non-HDL was 1.76 (95% CI 1.05 to 2.54), compared to 1.36 (95% CI 0.89 to 1.90) for low-density lipoprotein cholesterol. In conclusion, non-HDL cholesterol is the strongest predictor among the lipid risk factors of incident CVD in patients with established coronary heart disease.
Recent studies have suggested a possible role of apolipoprotein B–containing lipoproteins in the prediction of cerebrovascular disease (CVD) in the general population. Current national guidelines for stroke endorse the Adult Treatment Panel III guidelines in patients at risk for CVD, with a primary focus on the management of low-density lipoprotein (LDL) cholesterol, but do not specifically address non–high-density lipoprotein (non-HDL) (a surrogate measure of apolipoprotein B lipoproteins) as a possible primary therapeutic target. In this study, we evaluated the associations between different lipid fractions and cerebrovascular events in a post–myocardial infarction (MI) population, with an emphasis on non-HDL cholesterol (as it contains all atherogenic apolipoprotein B–containing lipoproteins and is accepted as an inexpensive surrogate marker for apolipoprotein B).
Methods
The participants randomized to the placebo arm of the Cholesterol and Recurrent Events (CARE) study formed the study population (n = 2,078) for this post hoc analysis. The design, methods, and results of the CARE trial have been described, Briefly, the CARE trial was a randomized, double-blind, placebo-controlled clinical trial that enrolled 4,159 men and women across 80 centers (67 in the United States and 13 in Canada) to test the effect of pravastatin on patients with average cholesterol (total cholesterol <240 mg/dl) within 3 to 18 months of index MI. For inclusion, subjects were required to have fasting glucose levels ≤220 mg/dl, left ventricular ejection fractions ≥25%, and absence of symptomatic congestive heart failure.
Total cholesterol, high density lipoprotein (HDL) cholesterol, and triglycerides were measured directly in fasting plasma, and LDL cholesterol was calculated using Friedewald’s formula. Non-HDL cholesterol was calculated by subtracting HDL cholesterol from total cholesterol. Patients were followed up for an average period of 5 years to detect fatal and nonfatal coronary and cerebrovascular events. Baseline information on a wide range of variables collected from the study population at the time of enrollment in this study was used for this study.
Neurologic events were included in the adverse event reporting protocol. Data on reported cases of cerebrovascular events were provided to the CARE end points committee. Cases were distributed to the 8 committee members for event categorization, and disagreements were resolved by consensus. Patients with history of CVD were excluded from the analysis. The possible predictors were selected for univariate and multivariate analyses for the first incident cerebrovascular event.
Cerebrovascular events included stroke and transient ischemic attack. Stroke was defined as the acute onset of a focal neurologic or monocular symptom resulting in either death or signs of presumed vascular origin that persisted for ≥24 hours. Stroke categorization is provided in the Appendix . Transient ischemic attacks were defined as focal neurologic deficits lasting <24 hours. Both stroke and transient ischemic attack were included in this analysis. Only a single event was counted per study patient (stroke took precedence over transient ischemic attack).
Demographic and baseline laboratory values are presented as mean ± SD or proportions as appropriate, unless otherwise specified. The baseline categorical variables for patients with and without CVD during follow-up were compared using chi-square tests and continuous variables using Wilcoxon’s rank-sum tests.
We used Cox proportional-hazards model to evaluate the associations between baseline lipids and other risk factors and risk for CVD. Triglycerides were log e transformed to improve normality. Regression coefficients were calculated to estimate the hazard ratios (HRs) associated with 1-SD higher baseline values of each baseline lipid fraction: 0.39log e triglyceride, 17.9 mg/dl non-HDL cholesterol, 14.6 mg/dl LDL cholesterol and 9.2 mg/dl HDL cholesterol. Multivariate models were constructed for each lipid risk factor adjusting for significant univariate predictors to calculate the multiple-adjusted HRs and their 95% confidence intervals (CIs). Lipids were grouped into quartiles to assess the magnitudes of the strength of the associations of different lipid fractions and cerebrovascular events. Statistical analysis was performed using Stata version 10 (StataCorp LP, College Station, Texas). Two-tailed p values <0.05 were considered significant.
Results
During a mean follow-up period of 5 years, 123 patients (6%) had cerebrovascular events after MI, amounting to 12 CVD events per 1,000 person-years. Stroke occurred in 61% (n = 76) and transient ischemic attacks in 39% (n = 47). Most of the CVD events occurred in patients without previous percutaneous coronary intervention (n = 91), coronary artery bypass grafting (n = 81), or thrombolysis (n = 120).
Patients with CVD during follow-up had a higher prevalence of co-morbidities: hypertension (57% vs 42%), diabetes (24% vs 14%), congestive heart failure (18% vs 7%), peripheral vascular disease (9% vs 3%), and angina (31% vs 16%). They also had older age (65 vs 58 years), higher non-HDL cholesterol (172.9 vs 169.3 mg/dl), and lower HDL cholesterol (37.2 vs 39.1 mg/dl) compared to patients without CVD ( Table 1 ). Additionally, patients who experienced cerebrovascular events had a higher incidence of recurrent MI (24% vs 13%, p <0.001), unstable angina (31% vs 16%, p <0.001), and intermittent claudication (3% vs 1%, p = 0.04).
| Variable | CVD | p Value | |
|---|---|---|---|
| Yes (n = 123) | No (n = 1954) | ||
| Demographic profile | |||
| Age (years) | 65 ± 7 | 58 ± 9 | <0.001 |
| Men | 104 (84%) | 1,684 (86%) | 0.53 |
| Hypertension | 71 (57%) | 828 (42%) | <0.01 |
| Angina | 39 (31%) | 320 (16%) | <0.001 |
| Congestive heart failure | 23 (18%) | 137 (7%) | <0.001 |
| Diabetes mellitus | 30 (24%) | 274 (14%) | <0.01 |
| Peripheral vascular disease | 11 (9%) | 53 (3%) | <0.001 |
| Continued smoking | 24 (19%) | 310 (16%) | 0.38 |
| Nonwhite race | 13 (10%) | 148 (8%) | 0.43 |
| Body mass index (kg/m 2 ) | 27.4 ± 3.6 | 27.7 ± 7.9 | 0.67 |
| Lipid profile | |||
| LDL cholesterol (mg/dl) | 138.9 ± 14.2 | 138.5 ± 14.6 | 0.77 |
| HDL cholesterol (mg/dl) | 37.2 ± 8.1 | 39.1 ± 9.3 | 0.03 |
| Non-HDL cholesterol (mg/dl) | 172.9 ± 17.4 | 169.3 ± 17.8 | 0.03 |
| Triglycerides (mg/dl) | 165 ± 58 | 154.6 ± 61 | 0.07 |
| Medications | |||
| Aspirin | 96 (77%) | 1,629 (83%) | 0.09 |
| Antiplatelet agents | 103 (83%) | 1,657 (85%) | 0.55 |
| Anticoagulants | 5 (4%) | 64 (3%) | 0.53 |
| β blockers | 38 (31%) | 761 (39%) | 0.08 |
| Angiotensin-converting enzyme inhibitors | 25 (20%) | 260 (13%) | 0.03 |
| Calcium channel blockers | 45 (36%) | 749 (38%) | 0.66 |
| Antiarrhythmic agents | 63 (51%) | 921 (47%) | 0.39 |
| Low left ventricular ejection fraction (≤40%) | 98 (79%) | 1,627 (83%) | 0.25 |
The following variables were associated with cerebrovascular events in the univariate model: age, diabetes mellitus, hypertension, the left ventricular ejection fraction (≤40%), peripheral vascular disease, congestive heart failure, angina, baseline non-HDL cholesterol, triglycerides, and baseline use of aspirin or angiotensin-converting enzymes or digoxin or antiplatelet agents. No association was observed between CVD and race, body mass index, use of alcohol, continued smoking, atrial fibrillation, heart rate, and the use of β blockers or nitrates or anticoagulants. Significant independent predictors of CVD risk (p <0.05) on multivariate analysis included increasing age, baseline non-HDL cholesterol, hypertension, diabetes, and continued smoking after MI ( Table 2 ).
| Variable | Adjusted HR (95% CI) | p Value |
|---|---|---|
| Hypertension | 1.54 (1.08–2.18) | 0.016 |
| Age | 1.12 (1.08–1.14) | <0.001 |
| Non-HDL cholesterol ⁎ | 1.28 (1.06–1.53) | 0.014 |
| Diabetes mellitus | 1.84 (1.22–2.77) | 0.004 |
| Continued smoking | 2.07 (1.31–3.26) | 0.002 |
⁎ For non-HDL cholesterol, the adjusted HR is per SD change (17.9 mg/dl change).
Using stepwise regression (forward selection and confirmed with backward selection method) among the lipid risk factors, we found that non-HDL cholesterol, when analyzed as a continuous variable (per SD change), had the strongest association with CVD ( Table 3 ). In model 1, the HR for baseline non-HDL cholesterol (per SD change) adjusted for age and gender was 1.28 (95% CI 1.07 to 1.54). With the addition of hypertension, diabetes, smoking, and body mass index (model 2), the adjusted HR was 1.28 (95% CI 1.06 to 1.53), and for model 3, the HR adjusted for variables in model 2 plus baseline HDL (per SD change) was 1.27 (95% CI 1.05 to 1.54). LDL and log-transformed triglycerides were not included in model 3, because of significant collinearity with non-HDL cholesterol. Similar analyses were repeated for baseline LDL and HDL cholesterol (per SD change) ( Table 3 ).
| Variable | Model 1 ⁎ | Model 2 † | Model 3 ‡ |
|---|---|---|---|
| Non-HDL cholesterol | 1.28 (1.07–1.54) | 1.28 (1.06–1.53) | 1.27 (1.05–1.54) |
| LDL cholesterol | 1.10 (0.92–1.32) | 1.14 (0.96–1.37) | 1.16 (0.97–1.39) |
| HDL cholesterol | 0.86 (0.72–1.04) | 0.90 (0.75–1.09) | 0.98 (0.80–1.20) |
† With age, gender, hypertension, diabetes, smoking, and body mass index.
‡ With age, gender, hypertension, diabetes, smoking, body mass index, and lipid fractions (adding baseline HDL [per SD change] to model with non-HDL, adding baseline HDL [per SD change] and log-transformed triglyceride [per SD change] to model with LDL, adding baseline non-HDL [per SD change] and log-transformed triglycerides [per SD change] to model with HDL cholesterol).
In summary, non-HDL cholesterol was the only lipid variable that significantly predicted CVD across all models. Even after adjusting for triglyceride levels, non-HDL cholesterol remained a stronger predictor of CVD compared to other lipid risk factors, although the p value was 0.05 (because of collinearity of triglycerides with non-HDL in our data set). When the analyses was restricted to patients with baseline triglycerides ≤200 mg/dl, adjusted HR for non-HDL cholesterol was 1.30 (95% CI 1.05 to 1.61) compared to 1.12 (95% CI 0.95 to 1.42) for LDL cholesterol, supporting the superiority of non-HDL cholesterol even when triglyceride levels are not elevated.
When we categorized lipid values into quartiles, the adjusted HR (adjusted for age and gender) of CVD for baseline non-HDL was 1.76 (95% CI 1.05 to 2.54) for levels ≥185 mg/dl (quartile 4) compared to non-HDL ≤156 mg/dl (quartile 1). Similar analyses for baseline LDL and HDL cholesterol yielded a lower magnitude of association ( Table 4 ). There was a linear increase in the log HR of first incident CVD with increasing non-HDL cholesterol quartiles suggesting a linear association between non-HDL cholesterol and cerebrovascular events. This relation was similar in both genders and in patients with and without diabetes and metabolic syndrome. Although not statistically significant, baseline HDL cholesterol expectedly demonstrated a negative association with CVD.
| Variable | Patients | Events | Relative Risk (95% CI) in Multivariate Model |
|---|---|---|---|
| Non-HDL cholesterol (mg/dl) | |||
| Quartile 1 (126–156) | 533 | 26 | 1.00 (referent) |
| Quartile 2 (157–170) | 504 | 27 | 1.18 (1.00–2.03) |
| Quartile 3 (171–184) | 524 | 37 | 1.56 (1.03–2.14) |
| Quartile 4 (≥185) | 499 | 33 | 1.76 (1.05–2.54) |
| LDL cholesterol (mg/dl) | |||
| Quartile 1 (107–127) | 524 | 28 | 1.00 (referent) |
| Quartile 2 (128–138) | 512 | 34 | 1.14 (0.75–1.42) |
| Quartile 3 (139–149) | 526 | 26 | 1.24 (0.80–1.61) |
| Quartile 4 (≥150) | 498 | 35 | 1.36 (0.89–1.90) |
| HDL cholesterol | |||
| Quartile 1 (20–33) | 544 | 32 | 1.00 (referent) |
| Quartile 2 (34–38) | 517 | 29 | 0.88 (0.53–1.46) |
| Quartile 3 (39–44) | 487 | 29 | 0.85 (0.51–1.41) |
| Quartile 4 (≥45) | 512 | 33 | 0.81 (0.49–1.32) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
