Blood levels of C-reactive protein (CRP), a marker of inflammation, are related to cardiovascular disease risk. To determine cross-sectional correlates in the elderly, we measured CRP in 400 men and women older than 65 years and free of clinical cardiovascular disease at baseline as part of the Cardiovascular Health Study. Only 2% of the values were greater than 10 mg/L, the cut-point usually used to identify inflammation. CRP levels appeared tightly regulated, since there were strong bivariate correlations between CRP and the following: inflammation-sensitive proteins such as fibrinogen (r = .52); measures of fibrinolysis such as plasmin-antiplasmin complex (r = .23); pack-years of smoking (r = .30); and body mass index (r = .24; all P values < or = .001). The association with pack-years was independent of the length of time since cessation of smoking. CRP levels were also associated with coagulation factors VIIc, IXc, and Xc; HDL cholesterol (negative) and triglyceride; diabetes status; diuretic use; ECG abnormalities; and level of exercise. Because of effect modification, two multiple linear regression prediction models were developed for CRP, one each for never smokers and ever smokers. An a priori physiologic model was used to guide these analyses, which disallowed the use of other inflammation-sensitive variables such as fibrinogen. In never smokers, the independent predictors were body mass index (+), diabetes status (+), plasmin-antiplasmin complex (+), and the presence of ECG abnormalities (+); this model predicted 15% of the CRP population variance. In ever smokers, the predictors were body mass index (+), plasmin-antiplasmin complex (+), pack-years of smoking (+), HDL cholesterol (-), and ankle-arm blood pressure index (-); this model predicted 42% of the population variance. We conclude that levels of CRP in the healthy elderly are tightly regulated and reflect lifetime exposure to smoking as well as level of obesity, ongoing level of fibrinolysis, diabetes status, and level of subclinical atherothrombotic disease. Moreover, exposure to smoking affects the relation of CRP to these other factors.