Polymorphisms in the CRP Gene and Cardiovascular EventsPolymorphisms in the

To the Editor: In their study of polymorphisms in the C-reactive protein (CRP) gene, Dr Lange and colleagues¹ found that genetic variation in the CRP gene was associated with CRP levels and that genetic variants were differentially associated with cardiovascular disease (CVD) events in white and black participants. The authors base their hypothesis on an analogous earlier study showing that the association of a PSSK9 variant with both low-density lipoprotein cholesterol (LDL-C) and CVD suggested that lifelong reduction in LDL-C would reduce CVD.

The likely effect of treatments targeting modifiable risk factors for CVD depends on robust causal inference of the risk factor−disease relationship, which is aided by mendelian randomization methods. Because these methods involve formally combining the CRP gene and gene-disease associations,² it is disappointing that Lange et al did not use their results to provide an estimate of the causal effect of CRP on CVD risk for direct comparison with similar studies.

This study found no association between genetic variation and carotid artery wall intima-media thickness, and the authors note that this “suggests a greater involvement of CRP in the transition from subclinical to clinical disease than in atherosclerosis progression.” This would not explain the largely null findings with respect to nonfatal events in this study. An effect of CRP on case fatality but not disease development is possible and in line with findings in rodent models of beneficial effects of reduction in CRP levels after myocardial infarction (MI).³ However, such an effect in humans cannot be established from a study of the nature or size of Lange et al; this would require a very large clinical cohort followed prospectively, with assessment of the effect of CRP (using the CRP gene as an instrument) on disease progression and case fatality. We agree that the associations found in Lange et al could be explained by chance and that the findings need replication in other larger studies.

It is important to be aware of the effects of publication bias within genetic epidemiology. Previously published null studies in this area⁴ (including references 14, 27, and 41-43 in the article by Lange et al) may not be well known. In contrast with the authors' assertion, most of these studies included a greater number of cases than theirs did.