(Circulation. 2007;115:e476.)
© 2007 American Heart Association, Inc.
Correspondence |
Department of Medicine II, Johannes Gutenberg University, Mainz, Germany
McMaster University and Hamilton General Hospital, Hamilton, Canada
We thank Drs Ridker and Everett for their interest in our work,1 and we commend Ridkers pioneering work describing an association between C-reactive protein (CRP) and the risk of myocardial infarction or stroke. Subsequent studies have both confirmed and refuted these original observations. The former studies "controlled" or "adjusted" for fewer other risk factors; when they did so, they dichotomized variables (a weaker approach) rather than using them as continuous variables. By contrast, the latter studies have incorporated adjustments for other markers (especially of abdominal obesity, because visceral fat is a source of various cytokines or N-terminal-pro-brain natriuretic peptide, which is emerging as a strong risk factor2) and have used more sensitive statistical approaches to adjustment. In most studies, after full adjustment the odds ratios between extreme quantiles have been too modest (generally under 2, but sometimes under 1.5) to be clinically useful or justify the additional expense of measuring CRP. In this respect, the study by Bos et al3 is unique because it addresses different subtypes of stroke, including subtypes of ischemic strokes such as large-artery hemispheric strokes, small-vessel lacunar stroke, and hemorrhagic strokes. There was no independent association between CRP concentration and incident stroke or its subtypes. The limitations of CRP as a predictor with incremental value (either using the methods of area under the curve or statistics) have been discussed by Lloyd-Jones4 et al and have been demonstrated elegantly by Danesh et al.5 Despite their reservations, inflammation (of which CRP may be a generalized marker) may still be causally related to the development of atherosclerosis or its clinical consequences. No amount of epidemiological analyses of observational studies can clarify causality unless the odds ratios are very large (eg, smoking) and the various rules of causation postulated by Bradford-Hill are met (eg, as with low-density lipoprotein or elevated blood pressure). More than 100 associations of risk markers with coronary heart disease or stroke have been reported, but the majority (including some persuasive hypotheses such as hormone therapy, homocysteine lowering, or antioxidant vitamins) have not been confirmed.
What is needed are novel research methods (perhaps using Mendelian randomization, although genetic studies do not provide any support of causality so far)6 or well-designed trials with interventions that specifically lower CRP (or other inflammatory markers) or block their biological effects, to assess whether clinical events can be lowered. We urge investigators who are proponents of "the inflammation hypothesis" to use such approaches to further the field, and to provide much more persua-sive evidence. For the moment, the importance of CRP as a useful clinical predictor of cardiovascular disease events or as a key component in the causation of cardiovascular disease remains unproven.
| Acknowledgments |
|---|
None.
| References |
|---|
|
|
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Circulation Home | Subscriptions | Archives | Feedback | Authors | Help | AHA Journals Home | Search Copyright © 2007 American Heart Association, Inc. All rights reserved. Unauthorized use prohibited. |