The paraoxonase (PON1) PON1-Q192R and PON1-L55M polymorphisms have been inconsistently associated with vascular disease. Plasma PON1 activity phenotypes vary markedly within genotypes and were, therefore, expected to add to the informativeness of genotype for predicting vascular disease. The case-control sample included 212 age- and race-matched men (mean age 66.4 years). The 106 carotid artery disease (CAAD) cases had >80% carotid stenosis, and the 106 controls had <15%. Two PON1 substrate hydrolysis rates (paraoxon [POase] and diazoxon [DZOase]) were significantly lower in cases than in controls and were significant predictors of CAAD by use of logistic regression (POase, P=0.005; DZOase, P=0.019). DZOase predicted vascular disease independently of lipoprotein profile, high density lipoprotein subfractions, apolipoprotein A-I, and smoking. PON1-192 and PON1-55 genotypes or haplotypes did not predict case-control status unless the activity phenotype was also included as a predictor by use of logistic regression. When phenotype was included as a predictor, PON1-192 and PON1-55 genotypes or combined haplotypes were significant predictors (P<0.05). In conclusion, examining PON1-192 and/or PON1-55 genotypes alone may mistakenly lead to the conclusion that there is no role of PON1 in CAAD. These results support the benefit of a "level crossing" approach that includes intervening phenotypes in the study of complexly inherited disease.