France (mean ± SEM: 1.68 ± 0.03 vs. 1.58 ± 0.03 mg L )1 ; P ¼ 0.01). This difference appeared stronger in controls than in cases (Table 1), even though the interaction did not reach significance (P ¼ 0.38).As two studies have suggested a threshold effect in which low levels of protein Z increased the risk of thrombosis [4,6], we have stratified the population into two groups according to the 10th percentile of the distribution in controls. In the whole cohort, 10 percent (n ¼ 58) of healthy controls had protein Z levels below 0.93 mg L )1 (10th percentile ¼ 0.93 mg L )1). Eleven percent (n ¼ 32) of the patients had protein Z levels below this cut-off (P ¼ 0.58), which means that protein Z below 0.93 mg L )1 was not a risk factor for coronary disease.This was homogeneous across countries as 12% (n ¼ 21) and 10% (n ¼ 12) of the patients, in France and Northern Ireland, respectively, had protein Z levels below the 10th percentile which was set in controls for each country separately (P ¼ 0.47 and 0.80 for case-control difference in France and Northern Ireland, respectively).In conclusion, antigenic protein Z levels do not seem to have a major effect on the risk of coronary heart disease in healthy men followed for 5 years.