A randomised trial without an intervention is difficult to envisage. In today's Lancet, however, Nic Timpson and colleagues1 take advantage of natural genetic randomisation during sexual reproduction-that of alleles bearing single nucleotide polymorphisms (SNPs) governing the abundance of C-reactive protein (CRP)-to assess whether associations between CRP and components of the metabolic syndrome are causal.
In prospective observational studies, CRP concentration has consistently been linked with cardiovascular events2 as well as with high-risk vascular phenotypes including high blood pressure, diabetes, and the metabolic syndrome.3,4 This fits with the prevailing view that inflammation is critical to atherogenesis. Whilst the similar and well-known associations of blood pressure and cholesterol levels with cardiovascular events are considered causal-because reducing blood pressure or cholesterol reduces cardiovascular risk in randomised trials-the same might not be true for CRP. Associations between CRP and disease could be explained by confounding, because of CRP's associations with other risk factors such as low birthweight, lower sociodemographic position, lack of physical activity, smoking, and abdominal obesity.5,6 Reverse causation might also be at work, whereby inflammatory cytokines from atheroma or adipose tissue raise CRP. (figure 1). While statistical adjustment can reduce confounding, not all confounding factors are known, or accurately measured. Moreover, adjustment requires a judgment about mechanism. For example, if blood pressure or diabetes mediate rather than confound the association between CRP and cardiovascular events, adjusting for these factors would lead to underestimation of the causal association. How might we get better insight into causation? Mechanistic studies in vitro have yielded conflicting results. Potentially proatherogenic and blood pressure-raising effects of CRP on vascular cells and tissues7 might have been mediated by proinflammatory bacterial peptides or sodium azide present in commercial CRP preparations.8,9 The increased atheroma formation in apolipoprotein-E-deficient mice that overexpress human CRP was not reproducible.10 A randomised trial of a selective CRP-lowering therapy is required, because randomisation would ensure that measured and unmeasured confounders were evenly distributed between placebo and intervention groups (figure 2). Unfortunately, no selective CRP-lowering drug exists. Statins lower CRP and have beneficial effects on cardiovascular disease, but the benefits may be adequately explained by cholesterol lowering. Alleles of the gene encoding CRP exist that influence circulating CRP concentration. These are transmitted from parent to offspring at random. Therefore, factors that could confound associations of CRP with components of the metabolic syndrome should be distributed evenly in those who do, and those who do not, have high-CRP alleles (figure 2). Moreover, because genotype is determined before onset of disease, the possibility of reverse causation is also overcome.11,12 Timpson and coworkers confirm that CRP is associated with metabolic syndrome, and that SNPs and haplotypes of the CRP gene are strongly associated with differences in CRP concentration. As expected, haplotypes inferred from these SNPs were not associated with factors, such as smoking, that could confound the association of CRP with the metabolic syndrome. By quantifying the associations of CRP haplotypes with blood pressure, triglycerides, HDL-cholesterol, adiposity, and insulin resistance, Timpson and colleagues made indirect but unconfounded estimates of the association of CRP with these components of the metabolic syndrome. In all cases, the indirect genetic estimates were smaller than the directly observed associations of CRP with the same factors, and in some cases they were in the opposite direction. This analysis suggests that the directly observed associations of CRP with the metabolic syndrome are affected by residual confounding and/or reverse causation, leading to the conclusion that CRP is unlikely to be a causal factor in its development. These findings could have wider implications. First, the data suggest that components of the metabolic synd-rome probably act as confounders, and do not mediate the association between CRP and cardiovascular events. Thus, it is probably legitimate to adjust for these factors when examining associations of CRP with cardiovascular events in observational studies. Second, the study identifies a set of genetic tools for examining whether CRP associations with cardiovascular events are causal. This is valuable because it would provide a test for claims that CRP-lowering could be as important as cholesterol-lowering in the prevention of cardiovascular events. How robust are Timpson and colleagues' conclusions? First, it is assumed that CRP genotype acts only by altering the circulating concentration of CRP but not its function, an assumption that appears reasonable with current evidence. Second, it assumes that a lifelong, genetically-determined increase in CRP does not lead to compensatory changes in other systems to inhibit potentially adverse effects of a higher CRP level.11 This could lead to the erroneous conclusion that raised CRP in later life as a result of adverse behaviours is not important in the development of the metabolic syndrome or cardiovascular disease. Finally, because differences in CRP level by genotype are small, very large studies will be required to reduce the present imprecision of the unconfounded genetic estimates of the effect of CRP on the metabolic syndrome. Even larger studies will be needed to assess whether or not associations of CRP with cardiovascular events are causal, and it is likely that studies will need to be pooled to achieve adequate sample size. Nevertheless, Mendelian randomisation offers the potential for insight into causation beyond that usually possible from observational epidemiology.11 Many other circulating biomarkers are also associated with cardio-vascular disease risk but, as with CRP, specific drugs that lower their concentration do not exist, and there are uncertainties as to whether such associations are causal. If it were possible to distinguish probable causal from non-causal links using Mendelian randomisation, this might help prioritise the development of new drugs for cardiovascular disease prevention. With the recent publication of databases of human DNA sequence variation,13 the availability of genetic tools for Mendelian randomisation studies is increasing. Carefully phenotyped cohorts, such as Timpson and colleagues', will facilitate such analyses and are testimony to the importance of a strong interface between genetics and epidemiology.


References
1) Timpson NJ, Lawler DA, Harbord RM, et al. C-reactive protein and its role in metabolic syndrome: mendelian randomisation study. Lancet 2005; 366: 1954-59.
2) Danesh J, Wheeler JG, Hirschfield GM, et al. C-reactive protein and other circulating markers of infiammation in the prediction of coronary heart disease. N Engl J Med 2004; 350: 1387-97.
3) Pradhan AD, Cook NR, Buring JE, Manson JE, Ridker PM. C-reactive protein is independently associated with fasting insulin in nondiabetic women. Arterioscler Thromb Vasc Biol 2003; 23: 650-55.
4) Sesso HD, Buring JE, Rifai N, Blake GJ, Gaziano JM, Ridker PM. C-reactive protein and the risk of developing hypertension. JAMA 2003; 290: 2945-51.
5) Pepys MB, Hirschfield GM. C-reactive protein: a critical update. J Clin Invest 2003; 111: 1805-12.
6) Sattar N, McConnachie A, O'Reilly D, et al. Inverse association between birth weight and C-reactive protein concentrations in the MIDSPAN Family Study. Arterioscler Thromb Vasc Biol 2004; 24: 583-87.
7) Verma S, Wang CH, Li SH, et al. A self-fulfilling prophecy: C-reactive protein attenuates nitric oxide production and inhibits angiogenesis. Circulation 2002; 106: 913-19.
8) Taylor KE, Giddings JC, van den Berg CW. C-reactive protein-induced in vitro endothelial cell activation is an artefact caused by azide and lipopolysaccharide. Arterioscler Thromb Vasc Biol 2005; 25: 1225-30.
9) Pepys MB. CRP or not CRP? That is the question. Arterioscler Thromb Vasc Biol 2005; 25: 1091-94.
10) Hirschfield GM, Gallimore JR, Kahan MC, et al. Transgenic human C-reactive protein is not proatherogenic in apolipoprotein E-deficient mice. Proc Natl Acad Sci USA 2005; 102: 8309-14.
11) Davey Smith G, Ebrahim S. 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 2003; 32: 1-22.
12) Keavney B. Commentary: Katan's remarkable foresight: genes and causality 18 years on. Int J Epidemiol 2004; 33: 11-14.
13) Palmer LJ, Cardon LR. Shaking the tree: mapping complex disease genes with linkage disequilibrium. Lancet 2005; 366: 1223-34.