Carriers of the APOEepsilon4 allele have consistently shown higher, and epsilon2 carriers have lower, plasma cholesterol and coronary heart disease (CHD) risk compared with epsilon3 homozygotes. An epsilon4:smoking interaction was observed in NPHSII, consistent with context dependency of the epsilon4 effect on CHD risk. In this study, APOE genotype was determined in 3787 male British civil servants followed for fatal and non-fatal myocardial infarction for median of 5.8 years, with 159 validated CHD events. APOE genotype was associated with expected effects on lipid traits (all P <0.0001). We tested the hypothesis that APOEepsilon4 was not a risk factor for CHD among non-smokers. In non-smokers, the odds ratio (OR) for epsilon2 and epsilon4 carriers were 0.51 (0.27, 0.97) and 0.70 (0.46, 1.08), respectively, compared with epsilon3 homozygotes. Thus epsilon2 carriers showed expected risk-protection, but despite 80% power to detect an OR in epsilon4 subjects of 1.71 (i.e. of magnitude increase reported in prospective studies), the epsilon4 non-smokers showed no increased risk compared with epsilon3 homozygotes. Smoking prevalence in this study was low (12.8%), but smokers had higher CHD risk which was of similar magnitude in risk in all genotypes [(OR 1.57 (1.03, 2.40)]. These data, therefore, provide strong corroborative evidence that there is no elevated risk of CHD in epsilon4 non-smokers, but failed to confirm the epsilon4:smoking interaction on risk. This supports the context dependency of the epsilon4 risk effect, but the low smoking incidence in the Whitehall men reduced our ability to examine a smoking:genotype interaction.