Info

□ <1.0 □ 1.0-3.0 ■ >3.0 3 -i C-Reactive Protein (mg/L)

<130 130-160 >160 LDL cholesterol (mg/dL)

Fig. 4. hsCRP provides prognostic information at all levels ofLDL-C and at all levels of the Framing-ham risk score. (Reproduced from ref. 12.)

Framingham estimate of 10-year risk (%)

<130 130-160 >160 LDL cholesterol (mg/dL)

Fig. 4. hsCRP provides prognostic information at all levels ofLDL-C and at all levels of the Framing-ham risk score. (Reproduced from ref. 12.)

of the National Cholesterol Education Program (68) but also at all levels of risk specified by the Framingham algorithm (12). After adjusting for all components ofthe Framingham risk score, the RR associated with increasing hsCRP quintiles at baseline were 1.0, 1.3, 1.4, 1.7, and 1.9 (p for trend <0.001) for all participants, and 1.0, 1.6, 1.5, 1.8, and 2.2 (p for trend <0.001) for those not taking postmenopausal hormone therapy. For the hsCRP cut points of <1, 1 to <3, and 3 mg/L or greater, the adjusted RRs were 1.0 (referent), 1.2 (95% CI: 0.9-1.5), and 1.5 (95% CI: 1.2-1.9) (p for trend <0.001) for all participants, and 1.0 (referent), 1.1 (95% CI: 0.9-1.6), and 1.5 (95% CI: 1.2-1.9) (p for trend <0.001) for those not taking postmenopausal hormone therapy (69).

The results from the Women's Health Study confirm earlier findings from a nested case-control analysis within the Physicians' Health Study, which showed that baseline levels of hsCRP were associated with incident MI and thromboembolic stroke even after control for cholesterol level in a large cohort of healthy middle-aged men followed for 14 yr (543 end points were included in the analysis) (40). More recent analyses of the Physicians' Health Study data indicate that hsCRP remains an independent predictor of MI in men after adjustment for all components of the Framingham risk score; compared with men with hsCRP levels <1 mg/L, those with hsCRP levels between 1 and 3 mg/L were 70% more likely to develop MI, and those with hsCRP levels >3 mg/L were more than twice as likely (RR: 2.2; 95% CI: 1.2-3.8) to do so (70).

These additive effects of hsCRP beyond the traditional Framingham risk score have also been observed in other large prospective studies. In a case-cohort analysis conducted among 12,819 apparently healthy middle-aged men and women followed for 6 yr in the multiethnic ARIC study (608 cases), the RR ofincident CHD for those with baseline hsCRP levels <1, 1-3, and >3 mg/L were 1.0 (referent), 1.31 (95% CI: 0.96-1.80), and 1.72 (95% CI: 1.24-2.39), respectively, after adjusting for age, sex, race, smoking status, systolic blood pressure, LDL-C, high-density lipoprotein cholesterol (HDL-C), and diabetes (21). Similar findings were obtained in a 6.6-yr follow-up of 3435 German men ages 45-74 yr participating in the MONICA Augsburg study; after adjustment for the 10-yr Framingham CHD risk score (categorized as <6, 6-10, 11-14, 15-19, and ©20%), the RR of incident

Fig. 5. Framingham-adjusted RRs of future coronary events according to baseline levels of hsCRP of <l, 1-3, and >3 mg/L in major cohort studies. (Data from refs. 10,12,21,39, and56—58.) *The Reykjavik data were reported for hsCRP tertiles rather than for hsCRP cutpoints of <1, 1-3, and >3 mg/L.

Fig. 5. Framingham-adjusted RRs of future coronary events according to baseline levels of hsCRP of <l, 1-3, and >3 mg/L in major cohort studies. (Data from refs. 10,12,21,39, and56—58.) *The Reykjavik data were reported for hsCRP tertiles rather than for hsCRP cutpoints of <1, 1-3, and >3 mg/L.

MI or sudden cardiac death among men with hsCRP levels <1, 1-3, and >3 mg/L were 1.0 (referent), 1.44 (95% CI: 0.95-2.17), and 2.21 (95% CI: 1.49-3.27), respectively (56). In subgroup analyses, the predictive utility of hsCRP was most apparent among men at intermediate coronary risk—i.e., those whose 10-yr Framingham scores were between 10 and 20%. In nested case-control analyses conducted within the large Nurses' Health Study (32,826 initially healthy women with a baseline blood sample were followed for 8 yr, during which time 249 incident CHD cases occurred) and the Health Professionals Follow-up Study (18,225 initially healthy men with a baseline blood sample were followed for 6 yr, during which time 266 incident CHD cases occurred), the pooled RR comparing hsCRP levels of ©3 mg/L to levels <1 mg/L was 1.46 (95% CI: 1.08-2.04) in a model that controlled for the effects of age, hypertension, ratio oftotal cholesterol to HDL-C, smoking, and diabetes (58). After additional adjustment for alcohol intake, body mass index (BMI), physical activity, parental history of CHD before age 60, and, for women, postmenopausal hormone therapy, the risk estimate remained largely unchanged (RR: 1.68; 95% CI: 1.18-2.38). Indeed, ofrecent large-scale population-based studies, only the Rotterdam Study failed to find that the addition ofhsCRP to a model with classic cardiovascular risk factors improved risk prediction. In this nested case-control study conducted among 7983 Dutch men and women age 55 yr and older who were followed for 5 yr for incident MI, the age- and sex-adjusted RR for the extreme quartile comparison (hsCRP >3.02 vs <0.82 mg/L) was 2.0 (95% CI: 1.1-3.4), but the association was attenuated after additional adjustment for smoking, total cholesterol, HDL-C, hypertension, diabetes, BMI, and family history of premature MI (RR: 1.2; 95% CI: 0.6-2.2) (p for trend across quartiles <0.50) (55). Nevertheless, the consistency of the Framingham-adjusted risk estimates in the Women's Health Study, Physicians' Health Study, ARIC, MONICA, Nurses' Health Study, and Health Professionals Follow-up Study provides compelling support for the incorporation of hsCRP into existing prediction algorithms (Fig. 5).

Data from a study in Reykjavik, Iceland that included 2459 incident CHD events during a 20-yr follow-up of18,569 men and women also highlight the clinical utility of hsCRP as an independent predictor of risk even in a population with elevated lipid levels (57). In this study, baseline hsCRP levels were associated with an approximate 50% increase in future vascular risk not only after adjustment for the traditional Framingham risk factors but also after further control for triglycerides, BMI, and indices of pulmonary function (hsCRP ©2.0 vs <0.78 mg/L; RR: 1.45; 95% CI: 1.25-1.68). During the first 10 yr of follow-up, an even higher risk was observed (RR: 1.84; 95% CI: 1.49-2.28). The reported RRs from Iceland were calculated using an hsCRP cut point of 2.0 mg/L rather than the recommended 3.0 mg/L and thus likely underestimate the predictive ability of hsCRP. Even so, the prognostic value of hsCRP in this cohort was virtually identical to that of high blood pressure and statistically similar to that of smoking.

Additional data supporting the addition of hsCRP to the Framingham risk evaluation come from two primary prevention trials. Among 5742 healthy men and women enrolled in the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS), a primary prevention trial of lovastatin, each quartile increase in baseline hsCRP predicted a 21% increase in the risk of a first cardiovascular event (95% CI: 4-41%), an association that persisted after adjustment for all individual components of the Framingham risk score (48). In a cross-sectional survey of1666 men and women without CVD participating in the Pravastatin Inflammation/CRP Evaluation trial, hsCRP level was minimally correlated with each variable in the 10-yr Framingham risk score and only modestly correlated with the score itself (r < 0.30) (71).

Although the relationship between hsCRP and CHD has been more extensively studied than that between hsCRP and stroke, data from the Framingham Study suggest that hsCRP is also an independent predictor of cerebral thrombosis. Among 1462 respondents followed for 12-14 yr for the development of a first ischemic stroke or transient ischemic attack, there were significant increases in risk across hsCRP quartiles after factoring out the effects of smoking, ratio of total cholesterol to HDL-C, systolic blood pressure, and diabetes (60). Among women, the RRs associated with increasing hsCRP quartiles (<1.00, 1.02-3.19, 3.20-7.31, ©7.33 mg/L) were 1.0 (referent), 1.2 (95% CI: 0.63-2.27), 1.6 (95% CI: 0.89-2.93), and 2.1 (95% CI: 1.9-3.83) (p for trend = 0.008). Among men, the corresponding RRs (for hsCRP quartiles of <1.08, 1.10-3.00, 3.03-6.80, ©6.90 mg/L) were 1.0 (referent), 0.9 (95% CI: 0.46-1.86), 1.5 (95% CI: 0.80-2.87), and 1.6 (95% CI: 0.873.13) (p for trend = 0.04).

Clinical interpretation of hsCRP is most conveniently performed using cut points of <1, 1-3, and >3 mg/L. However, recent findings from the Women's Health Study suggest that the cardiovascular risk gradient extends across the full range of hsCRP levels. That is, the absolute risk of CVD is extremely low for those 10-15% of individuals with hsCRP levels <0.5 mg/L, and, by contrast, the risk is extremely high when hsCRP levels exceed 10 mg/L (Fig. 6) (69). Moreover, as shown in Fig. 7, there is a clear risk gradient associated with a five-tier classification scheme for hsCRP (<0.5, 0.5 to <1, 1 to <3, 3 to <10, and ©10 mg/L) across all levels of the Framingham risk score (70). A monotonic increase in cardiovascular risk associated with rising hsCRP levels was observed not only among individuals with estimated 10-yr Framingham risk scores of between 10 and 20% but also among those with lower scores. These data suggest that an hsCRP-modified Fram-ingham risk score could improve the prediction of CHD events in general population settings.

Fig. 6. Clinical predictive value for future cardiovascular (CV) events of very low (<0.5 mg/L) and very high (>10 mg/L) levels of hsCRP. (Reproduced from ref. 69.)

Calculated Framingham

Fig. 7. RRs of future coronary events according to a five-category classification scheme for hsCRP and Framingham risk score. (Reproduced from ref. 70.)

Calculated Framingham

Fig. 7. RRs of future coronary events according to a five-category classification scheme for hsCRP and Framingham risk score. (Reproduced from ref. 70.)

It should be noted that most of the epidemiological studies cited were conducted in white populations ofpredominantly European ancestry. With some exceptions (e.g., the aforementioned Honolulu Heart Program [53,61]), few data on the utility of hsCRP as a predictor of incident CVD in nonwhite cohorts exist. Additional investigations in racially diverse populations are needed.

Was this article helpful?

0 0

Post a comment