|Home | About | Journals | Submit | Contact Us | Français|
We sought to determine the relative strength of high-sensitivity C-reactive protein (hs-CRP) and lipid levels as markers for future ischemic stroke as compared with coronary heart disease in women.
While hs-CRP and lipid levels are established risk determinants for vascular disease, the relative strength of these biomarkers for ischemic stroke as compared to coronary disease (CHD) is uncertain.
Among 15,632 initially healthy women who were followed for a 10-year period, we compared hs-CRP, total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), non-high-density lipoprotein cholesterol (non-HDL-C), high-density lipoprotein cholesterol (HDL-C), apolipoproteins A-I and B100, and lipid ratios as determinants of ischemic stroke as compared to coronary heart disease.
After adjustment for age, smoking status, blood pressure, diabetes, and obesity, the hazard ratios (HRs) and 95% confidence intervals for the third versus the first tertile for future ischemic stroke as compared to CHD were 1.91 (1.13–3.21) and 2.26 (1.64–3.12) for TC, 1.29 (0.83–2.02) and 2.09 (1.53–2.85) for LDL-C, 0.57 (0.36–0.92) and 0.38 (0.27–0.52) for HDL-C, 1.72 (1.03–2.86) and 2.93 (2.04–4.21) for non-HDL-C, and 2.76 (1.51–5.05) and 1.66 (1.17–2.34) for hs-CRP, respectively. Of the lipid ratios, the TC to HDL-C ratio had the largest HR for both future ischemic stroke and CHD [HR 1.95 (1.16–3.26) and 4.20 (2.79–6.32), respectively].
In this large, prospective cohort of initially healthy women, lipid levels are significant risk determinants for ischemic stroke, but with a magnitude of effect smaller than that observed for CHD. hs-CRP associates more closely with ischemic stroke than CHD. Concomitant evaluation of lipid levels and hs-CRP may improve risk assessment for stroke as well as CHD.
Among 15,632 initially healthy women followed for a 10-year period, we compared C-reactive protein (hs-CRP) and a full lipid panel as determinants of ischemic stroke as compared to coronary heart disease. The adjusted hazard ratios (HRs) and 95% confidence intervals for the third versus the first tertile for future ischemic stroke as compared to CHD were 1.91 (1.13–3.21) and 2.26 (1.64–3.12) for total cholesterol, 0.57 (0.36–0.92) and 0.38 (0.27–0.52) for HDL-C, and 2.76 (1.51–5.05) and 1.66 (1.17–2.34) for hs-CRP, respectively. Lipid levels associate more closely with CHD than ischemic stroke, while hs-CRP associates more closely with stroke than CHD.
While hyperlipidemia and inflammation play major roles in atherothrombosis, there has been controversy regarding the relative contribution of these processes to ischemic stroke as compared to coronary heart disease (CHD). For example, while both lipids levels and high-sensitivity C-reactive protein (hs-CRP) predict CHD (1–8), several studies have not found lipid levels to predict incident stroke (9–15). By contrast, some prior studies of hs-CRP and other inflammatory biomarkers have suggested that these markers are closely associated with future stroke, although data are somewhat limited in women (2,16–19). While methodological issues, such as considering only fatal stroke (12), combining hemorrhagic and ischemic stroke as a single endpoint (11), or being restricted to somewhat limited sample size (9,10), may have contributed to the lack of a relationship between cholesterol and stroke described by many investigators, it is also possible that inflammation and hyperlipidemia contribute differentially to clinical events in the cerebral and coronary vascular beds.
To address these issues, we analyzed a full lipid panel and hs-CRP in a large-scale cohort of women with the specific aim of evaluating the magnitude of risk for each biomarker for ischemic stroke as compared to CHD.
The Women’s Health Study (WHS) comprises a randomized, double-blind, placebo controlled 2×2 factorial design trial of aspirin and vitamin E in the primary prevention of cardiovascular disease and cancer conducted among initially healthy women aged 45 years or older. Participants were enrolled between November 1992 and July 1995 and were followed prospectively for all cardiovascular events including nonfatal myocardial infarction, nonfatal stroke, coronary revascularization procedures, and cardiovascular-related death. All participants in the WHS provided written informed consent and the study protocol was approved by the institutional review board of the Brigham and Women’s Hospital (Boston, Mass).
The methods of endpoint validation have been described elsewhere in detail (20). Briefly, an endpoints committee comprised of physicians reviewed the medical records of all women who were reported to have suffered a cardiovascular event. CHD was defined as a composite of confirmed fatal or non-fatal myocardial infarction, death from coronary artery disease, and coronary revascularization. A myocardial infarction was confirmed if symptoms met WHO criteria and was accompanied by diagnostic electrocardiographic changes or abnormal elevations in serum cardiac enzymes. Stroke was defined as a new focal neurologic deficit of sudden onset and vascular origin that persisted for more than 24 hours. Clinical information and radiographic brain scans were used to distinguish hemorrhagic from ischemic strokes. The inter-observer agreement in the classification of stroke into hemorrhagic and ischemic subtypes in the WHS has been shown to be excellent (21).
Among WHS participants, 28,345 women provided blood samples that were stored in liquid nitrogen until the time of analysis. These samples underwent lipid analysis and evaluation for high-sensitivity CRP in a core laboratory certified by the National Heart, Lung, and Blood Institute/Centers for Disease Control and Prevention Lipid Standardization program. Levels of total cholesterol and HDL-C were measured enzymatically on a Hitachi 911 autoanalyzer (Roche Diagnostics, Basel, Switzerland) while LDL-C was determined directly (Genzyme, Cambridge, Mass). Levels of apolipoproteins B100 and A-I were measured by an immunoturbidometric technique on the Hitachi 911 analyzer (22). High-sensivity CRP was measured using a validated immunoturbidometric method (Denka Seiken, Tokyo, Japan) (23). Of the samples received by the core laboratory, 27,748 (98%) underwent successful evaluation for each biomarker. Non-HDL-C was calculated by subtracting HDL-C from total cholesterol.
We restricted our analysis to women who were not using hormone therapy at the time of lipid analysis (N=15,632) because of guidelines issued by the U.S. Department of Health and Human Services (24) and because hormone therapy is known to alter lipid and inflammatory biomarker levels (25–27).
Population distributions were computed for each marker and Spearman correlation coefficients were used to test for interrelationships between each of the lipid subfractions and hs-CRP. Baseline levels of each blood parameter were divided into increasing tertiles. We used Cox proportional hazards models to calculate hazard ratios (HRs) and 95% confidence intervals for future CHD and ischemic stroke by comparing the 2nd and 3rd tertiles using the lowest tertile as the reference group. HRs were adjusted on an a priori basis for age, randomized treatment assignment, Framingham blood pressure category, body mass index (BMI), diabetes, and smoking status. For any study participant who suffered both a CHD event and an ischemic stroke, only the first event was used in this analysis. Tests for trends across tertiles of each biomarker were addressed by entering a single ordinal term based on the median value for that biomarker within each tertile. Data analysis was conducted using SAS statistical software version 9.1 (SAS Institute Inc., Cary, NC).
Mean (SD) age at baseline for the 15,632 initially healthy women followed for this portion of the study was 53.5 (7.7) years and the mean (SD) body mass index (BMI) was 26.3 (5.3) kg/m2. As previously reported (28), total cholesterol, LDL-C, non-HDL-C, and apolipoprotein B100 were all highly correlated with one another (r values ranged from 0.76 to 0.93, all significant to P<0.0001). HDL-C and apolipoprotein A-I were highly correlated with one another (r=0.80) but only weakly correlated with the other lipid variables. By contrast, the correlation between hs-CRP and the lipid variables was weaker, ranging from r = −0.33 (P<0.0001) with HDL-C to 0.15 (P<0.0001) with LDL-C.
Over the follow-up period of 10 years, 468 subjects suffered a first cardiovascular event (132 ischemic stroke, 336 coronary heart disease). As expected, there were significant differences in mean age, mean BMI, history of hypertension, diabetes, and smoking status across the groups of women who did not suffer a cardiovascular event and those who reached either cardiovascular endpoint (Table 1). The median levels of the lipid values, their ratios, and hs-CRP in each of the three groups of women are also shown in Table 1. Women with incident CHD were more likely to have a parental history of myocardial infarction before the age of 60 than women with ischemic stroke, and had slightly lower HDL-C and apolipoprotein A-I levels. Perhaps because of these differences in HDL-C and apolipoprotein A-I, lipid ratios were slightly higher among women with incident CHD when compared to women with incident stroke.
The fully adjusted HRs for developing future ischemic stroke and coronary heart disease for each of the single lipid variables and hs-CRP are presented in Table 2. After adjustment for age (years), blood pressure (Framingham categories), diabetes, BMI (kg/m2), current smoking status and randomized treatment assignment, increasing tertiles of total cholesterol, non-HDL-C, HDL-C, and hs-CRP were all associated with future ischemic stroke (Ptrend = 0.02, 0.03, 0.02, and 0.003, respectively). Specifically, the HR of future stroke for those women in the highest as compared with the lowest tertile was 1.91 (95% CI, 1.13–3.21) for total cholesterol, 1.72 (95% CI, 1.03–2.86) for non-HDL-C, 0.57 (95% CI, 0.36–0.92) for HDL-C, and 2.76 (95% CI, 1.51–5.05) for hs-CRP. While the associations between increasing tertiles of LDL-C and apolipoproteins B100 and A-I were not statistically significant, the estimates of the HRs were in the expected direction (HR = 1.29, 95% CI, 0.83–2.02 for LDL-C; HR = 1.47, 95% CI, 0.88–2.44 for apolipoprotein B100, and HR = 0.72, 95% CI, 0.47–1.11 for apolipoprotein A-I), consistent with the direction of effect for other highly correlated lipid variables. Among the significant risk determinants for future ischemic stroke, the magnitude of association as measured by the likelihood ratio (LR) χ2 statistic was 213.7 for hs-CRP, 208.0 for total cholesterol, 207.3 for HDL-C, and 206.1 for non-HDL-C.
The fully adjusted HRs for lipid ratios as determinants of ischemic stroke are summarized in Table 3. The ratios of total to HDL cholesterol, LDL-C to HDL-C and apolipoprotein B100 to HDL-C all associated with future ischemic stroke (Ptrend = 0.004, 0.01, and 0.03, respectively), while the ratio of apolipoprotein B100 to A-I was of borderline significance (Ptrend = 0.08). Specifically, the HRs of future ischemic stroke for those in the highest as compared with the lowest tertile were 1.95 (95% CI, 1.16–3.26) for total to HDL cholesterol ratio, 1.77 (95% CI, 1.08–2.90) for the LDL-C to HDL-C ratio, 1.42 (95% CI, 0.88–2.30) for the ratio of apolipoproteins B100 to A-I, and 1.62 (95% CI, 0.98–2.68) for the apolipoprotein B100 to HDL-C ratio. The ratio of total to HDL cholesterol had a LR χ2 statistic of 210.1, which was similar to that of many of the single lipid measures, but smaller than the statistic for hs-CRP (LR χ2 statistic=213.7).
As anticipated, all the measured lipid levels and hs-CRP were associated with future CHD (Table 2). After adjustment, the magnitude of association appeared to be greatest between HDL-C, with a HR of CHD for the highest as compared with the lowest tertile equal to 0.38 (95% CI, 0.27–0.52; LR χ2 statistic=435.2), although the highly correlated variables non-HDL-C (HR 2.93, 95% CI, 2.04–4.21, LR χ2 statistic=431.1) and apolipoprotein B100 (HR 3.25, 95% CI, 2.18–4.82; LR χ2 statistic=429.7) also showed close association with future CHD.
All of the lipid ratios strongly associated with future coronary heart disease (Table 3). In contrast to the single lipid measurements, where non-HDL-C and apolipoprotein B100 were more closely associated with CHD than total cholesterol, the lipid ratio with the strongest association was the total cholesterol to HDL-C ratio (HR for highest vs. lowest tertile, 4.20; 95% CI, 2.79–6.32; LR χ2 statistic=457.4). Similarly, the hazard ratio for future CHD for the top versus the bottom tertile for the ratio of apolipoprotein B100 to HDL-C was 3.98 (95% CI, 2.65–5.96; LR χ2 statistic=452.6).
Figure 1 and Figure 2 present the adjusted HRs for each lipid variable, hs-CRP and the lipid ratios for those in the highest compared with the lowest tertile. The association between each of the lipid variables and ischemic stroke is similar in direction of effect to its association with coronary heart disease. Visually, the HRs for future CHD for all of the lipid variables appear to be slightly stronger than the HR of future ischemic stroke for the same lipid, although we performed no formal statistical comparison. In contrast, the HR for future ischemic stroke for those in the highest tertile of hs-CRP appears to be somewhat larger than the HR of future CHD, although the 95% confidence intervals overlap (HR 2.76; 95% CI, 1.51–5.05 for ischemic stroke vs. HR 1.66;95% CI, 1.17–2.34 for CHD) and no formal statistical comparison was made.
In order to clarify whether this apparent difference in the strength of association between lipid levels and ischemic stroke as compared with CHD was due to a high proportion of cardioembolic strokes caused by atrial fibrillation, we repeated our analysis after excluding all patients who reported either prevalent atrial fibrillation at the time of enrollment in the WHS or subsequently developed the dysrrhythmia during the 10 years of follow-up. Of the 132 women who suffered an ischemic stroke as their first cardiovascular event, 5 had reported atrial fibrillation upon admission to the study and 18 developed atrial fibrillation during the course of follow-up, for a total of 109 women with ischemic stroke who had no history of atrial fibrillation. The point estimates for the HRs of future stroke for the third versus the first tertile of each of the lipid values, their ratios, and hs-CRP did not change substantially. Specifically, after multi-variable adjustment, the HRs of future ischemic stroke for those in the highest as compared with the lowest tertile were 1.87 (95% CI, 1.08–3.24; Ptrend = 0.03) for total cholesterol, 1.30 (95% CI, 0.80–2.10; Ptrend = 0.16) for LDL-C, 0.53 (95% CI, 0.32–0.89; Ptrend = 0.01) for HDL-C, 1.64 (95% CI, 0.95–2.82, Ptrend = 0.05) for non-HDL-C, 0.70 (95% CI, 0.44–1.12; Ptrend = 0.12) for apolipoprotein A-I, 1.42 (95% CI, 0.82–2.46; Ptrend = 0.17) for apolipoprotein B100 and 2.65 (95% CI, 1.39–5.08; Ptrend 0.007) for hs-CRP. Similarly, those estimates did not change for the lipid ratios, with the HRs of future ischemic stroke for the top versus the bottom tertile equal to 2.24 (95% CI, 1.26–4.00; Ptrend = 0.002) for the total cholesterol to HDL-C ratio, 1.81 (95% CI, 1.06–3.10; Ptrend = 0.02) for the LDL-C to HDL-C ratio, 1.46 (95% CI, 0.87–2.47; Ptrend = 0.07) for the ratio of apolipoproteins B100 to A-I, and 1.68 (95% CI, 0.97–2.91, Ptrend = 0.03) for the ratio of apolipoprotein B100 to HDL-C. Aside from the point estimate of the HR for the top versus the bottom tertile of the total cholesterol to HDL-C ratio, which changed from 1.95 (95% CI, 1.16–3.26) to 2.24 (95% CI, 1.26–4.00), or approximately 15%, no estimate for any of the HRs for extreme tertiles changed by more than 7% after we excluded women with incident or prevalent atrial fibrillation from the analysis.
During follow-up only 31 hemorrhagic strokes occurred, and we found no significant linear relationship between hemorrhagic stroke and any of the lipid variables or hs-CRP, using both minimally-adjusted (for age and randomized treatment assignment) and fully-adjusted proportional hazards models.
Finally, because there is evidence in the literature that exercise (29) and alcohol consumption (30) may be related to stroke risk, and both are known to be related to cholesterol (and in particular HDL-C) levels (31,32), we constructed a Cox model which included exercise (exercise <1 time per week, 1–3 times per week, and at least 4 times per week) and alcohol consumption (less than 3 drinks per month, 1–6 drinks per week, and at least 1 drink per day) in addition to the other variables included in our fully adjusted model. None of the HRs for extreme tertiles of any of the lipid values or hs-CRP changed by more than 5% with the addition of alcohol and exercise to the previously constructed model.
In this prospective cohort of initially healthy women, we directly compared standard lipid measures, apolipoproteins B100 and A-I, and hs-CRP as risk determinants for future ischemic stroke as compared with coronary heart disease. Overall, we found that lipid levels are significantly associated with a risk of future ischemic stroke in a direction similar to that of coronary heart disease. However, for all lipid measures, the magnitude of effect appeared smaller for future ischemic stroke than for CHD. While hs-CRP was a significant predictor of both clinical events, the magnitude of effect, if anything, was somewhat greater for ischemic stroke than for CHD. We could demonstrate no linear relationship between future hemorrhagic stroke and lipid or hs-CRP levels. No substantive differences were observed after exclusion of those with prevalent or incident atrial fibrillation.
We believe these data are clinically relevant for several reasons. First, a clear link between lipids and future stroke has not been established. In our population of initially healthy U.S. women, non-HDL-C and HDL-C were associated with future ischemic stroke risk in a way that was clearly dose-dependent, while total cholesterol was also associated with future ischemic stroke in a somewhat weaker dose-dependent manner. There was a trend towards an increased risk of ischemic stroke with increasing tertiles of apolipoproteins B100, A-I and LDL-C, although the trend was not significant. All the lipid ratios except the ratio of apolipoprotein B100 to A-I showed a consistently increasing risk of stroke with increasing levels of the ratio.
As anticipated, lipid levels were all clearly associated with the risks of incident coronary heart disease. In all cases but one (the ratio of apolipoprotein B100 to A-I), the direction and magnitude of the association between the measured lipid variables and the endpoints of ischemic stroke and coronary heart disease were not significantly different. While ischemic stroke represents a heterogeneous disorder (33), these data provide evidence that in our population, hyperlipidemia confers a risk of both ischemic heart disease and ischemic cerebrovascular disease.
Previous literature has noted the lack of association between lipids and ischemic stroke. Some studies have been unable to differentiate between hemorrhagic and ischemic stroke, and have suggested that cholesterol’s positive association with ischemic stroke may be concealed by a negative relationship with hemorrhagic stroke (11). While a relationship between lipids and ischemic and hemorrhagic stroke was suggested by evidence from the MR FIT trial (34), we could not demonstrate a linear relationship between the risk of hemorrhagic stroke and lipid levels. We are significantly limited, however, by the number of hemorrhagic strokes (n=31) in our population. Other early research reports may have been limited by sample size (9,10), combining fatal and incident ischemic stroke (35), or examining only fatal stroke (12).
More recently, researchers working on a number of well-characterized cohorts such as the Atherosclerosis Risk in Communities Study (ARIC) (14) and the Physician’s Health Study (15) have reported similarly negative results. In the ARIC cohort, Shahar and colleagues reported a total of 305 ischemic strokes, 144 of which were in women. While they do not report the risk associated with elevated levels of total cholesterol, non-HDL-C or the total cholesterol to HDL-C ratio, which were the lipid measures that demonstrated the strongest association with future ischemic stroke in our cohort, the HRs of ischemic stroke among women for the highest versus the lowest quartile of LDL-C (HR 1.33; 95% CI, 0.81–2.20), apolipoprotein B100 (HR 1.61; 95% CI 0.96–2.69) and HDL-C (HR 0.68; 95% CI, 0.36–1.27) are quite similar to the HRs for extreme tertiles of those lipid measures we report here. An alternative explanation is that the population of women included in the WHS has fewer established risk factors for ischemic stroke than that recruited for ARIC, such that elevated lipid levels play a more central role in the pathophysiology of ischemic stroke among these women.
While Bowman and colleagues were unable to show a relationship between cholesterol levels and ischemic stroke in a nested-case-control study of 296 ischemic strokes from the Physician’s Health Study (15), other studies performed exclusively in men have demonstrated a positive association between lipids and ischemic stroke (34–37). Investigators working in a cohort of subjects enrolled in a health maintenance organization found an OR of 1.6 (95% CI, 1.3–2.0) of ischemic stroke for the highest versus the lowest quintile of total cholesterol and a protective effect of the highest levels of HDL-C (OR=0.8, 95% CI, 0.6 to 1.0 for extreme quintiles) (33), results that are consistent with recently published findings from Korea (38). Sacco and colleagues, using an ethnically diverse community-based cohort in New York City, also demonstrated a protective effect of high HDL-C levels on the risk of ischemic stroke (39).
The association we report between hs-CRP and future ischemic stroke has been demonstrated previously in both men (2) and in cohorts composed of men and women (16,17). We were able to confirm that relationship in our population of women, and to demonstrate that the relationship between hs-CRP and future ischemic stroke (HR 2.76; 95% CI 1.51–5.05; LR χ2 statistic=213.7) was at least as strong as that of the strongest lipid risk determinant, the ratio of total to HDL cholesterol (HR 1.95; 95% CI 1.16–3.26; LR χ2 statistic=210.1). While the lipid values were highly correlated with one another, the correlations between hs-CRP and the lipid parameters were weaker (range, 0.15 to −0.33) (28). This observation is consistent with the hypothesis that both inflammation and hyperlipidemia contribute to the development of ischemic stroke and is concordant with prior data that markers of inflammation, such as fibrinogen (18) and the leukocyte count (19), are related to incident stroke.
Another possibility is that while most ischemic stroke is the result of atherosclerotic processes, cardioembolic stroke (e.g., thromboembolism from atrial fibrillation) may not be closely associated with hyperlipidemia. While hs-CRP is a well established marker of the systemic inflammation intrinsic to atherosclerosis in this cohort (5) and others (2,6–8), elevated levels of hs-CRP have also been associated with the presence of chronic and paroxysmal atrial fibrillation (40) and may predict incident atrial fibrillation (41). By capturing subjects who are predisposed to both etiologies of ischemic stroke, hs-CRP may serve as a more potent marker of future ischemic stroke than lipids alone. However, our findings that excluding subjects with atrial fibrillation does not substantially alter the point estimates of the hazard ratios for any of the lipids, their ratios, or hs-CRP does not support this hypothesis.
The reliability of our laboratory assays and the large scale, prospective nature of the cohort serve to reduce the possibility of laboratory error, bias or chance as explanations for our findings. We elected to exclude women who were using hormone therapy at the time of lipid analysis. While doing so may limit the generalizability of our results somewhat, we did so on an a priori basis because of guidelines issued by the U.S. Preventive Task Force (24), and because of evidence in the literature that hormone use alters both lipid levels (25,26) and hs-CRP levels (27) and is associated with socioeconomic status, education, exercise and other unmeasured confounders of CHD and ischemic stroke events (42,43). Because hormone therapy use is declining (44,45), we believe that our results are applicable to a large segment of the U.S. population. Our study included only middle-aged women participating in a randomized controlled trial, who are more likely to be healthy than their peers. We measured lipid and inflammatory biomarkers once and simple intra-individual variability may have altered circulating lipid levels, causing us to misclassify the exposure. However, the overall relationship between lipids and cardiovascular disease in our study (28) is similar to that in other studies conducted predominantly in men (4). In addition, statin use was very low at the time of lipid analysis in the WHS (20), and the initiation of statins among women with high lipid levels would tend to bias our results towards the null.
Our analysis supports the use of standard lipid measures such as total cholesterol, HDL-C, non-HDL-C, and the ratio total cholesterol to HDL-C and hs-CRP in the assessment of risk for ischemic stroke in addition to coronary heart disease. While the risk of coronary heart disease for a given level of lipid may be somewhat higher than the risk of ischemic stroke, the risks are quite similar in direction. Inflammation, as measured by hs-CRP, and hyperlipidemia appear to play a key role in the development of cerebrovascular atherosclerosis and ischemic stroke, consistent with their well-known role in the pathophysiology of coronary atherosclerosis and coronary heart disease.
We are indebted to the participants in the Women’s Health Study for their conscientious collaboration and to the entire staff of the Women’s Health Study for their expert assistance.
Financial Support: The research for this article was supported by grants from the Donald W. Reynolds Foundation (Las Vegas, NV), the Leduq Foundation (Paris, France), and the Doris Duke Charitable Foundation (New York, NY). The Women’s Health Study is supported by grants from the National Heart, Lung, and Blood Institute (HL-43851) and the National Cancer Institute (CA-47988).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of Interest: Dr. Ridker is listed as a co-inventor on patents held by the Brigham and Women’s Hospital that pertain to the use of inflammatory biomarkers in cardiovascular disease.