C-reactive protein in healthy and in sick populations.

suspected vascular risk factors such as smoking and obesity; a marker of chronic infective processes possibly correlated with coronary heart disease risk such as persistent infection with Chlamydia pneumoniae; or a marker of the extent of subclinical disease (since atherosclerosis is partly an inflammatory lesion). In these cases, C-reactive protein itself would not be of causal relevance to coronary heart disease. Despite the uncertainties about the relevance of C-reactive protein to coronary heart disease in the general population, a number of studies have investigated C-reactive protein in other settings, including long-term studies of populations defined on the basis of pre-existing conditions such as vascular disease, diabetes and chronic renal disease. Compared with studies in the general population, interpretation of studies in these ‘sick’ populations is complicated by an increased likelihood that factors related to pre-existing disease itself may alter both C-reactive protein values and coronary heart disease risk. Such distortions might be particularly likely in patients with pre-existing coronary heart disease. This is because C-reactive protein values can be influenced not only by the severity of disease itself but also by the effects of medications prescribed for coronary heart disease (for example, certain ‘statin’ drugs may lower plasma C-reactive protein values) and of changes in habits and risk factors following a diagnosis of coronary heart disease (for example, smoking cessation and weight reduction). Even if studies in sick populations attempt adjustments for such potentially distorting factors, the results may still be of uncertain validity. Residual biases are possible both because baseline values of some confounders may be inaccurate measurements of their long-term ‘usual’ values, thereby resulting in inadequate statistical adjustments; and because some possible confounders may not be measured at all. The study of Garcia-Moll et al. in this issue illustrates the challenges of investigating C-reactive protein in sick populations. They made C-reactive protein measurements in 911 British patients with chronic stable angina subsequently monitored for a median duration of about 18 months. The authors correctly point out that the 29 myocardial infarctions and deaths from coronary heart disease recorded in their study were too few for reliable analyses of the predictive ability of C-reactive protein. By comparison, three previously published long-term prospective studies of C-reactive protein in cohorts defined on the See page 1598 for the article to which this Editorial refers