Abstract Background Microflora-dependent trimethylamine-N-oxide (TMAO) formation, which results from intake of choline and L-carnitine-rich food, shows promise as a predictor of cardiovascular disease (CVD) risk, but these associations have not been examined in ethnically diverse populations. In a multiethnic population-based study of adults in Canada, we assessed the stability of TMAO and L-carnitine in stored serum samples and their association with intimal medial thickness, prevalent risk factors, and clinical events. Methods In a randomly sampled cross-sectional study of 1286 Canadians, fasting serum samples were collected and stored. In 292 consecutive individuals (99 CVD cases and 193 unmatched control subjects), L-carnitine and TMAO concentrations were assessed using validated analytical approaches. Results The mean (± SD) TMAO level was 1.998 ± 3.13 μM and L-carnitine was 42.29 ± 11.35 μM. The relative levels of the samples did not appreciably change after 3 freeze-thaw cycles (coefficient of variation, 5.6% and 4.7%, respectively). No significant association between L-carnitine levels and prevalent CVD was found, with adjustment for covariates (odds ratio, 1.57; 95% confidence interval, 0.58-4.26; P trend = 0.65), for highest vs lowest quintile group. TMAO levels showed a significant, graded association with prevalent CVD (odds ratio, 3.17; 95% confidence interval, 1.05-9.51; P trend = 0.02). After further adjustment for diabetes status, meat, fish, and cholesterol intake, the association remained significant. No significant association between carotid intimal medial thickness and L-carnitine ( P  = 0.64) or TMAO ( P  = 0.18) was found. Conclusions Serum TMAO and L-carnitine analysis on stored samples is reliable. Our findings support an association between TMAO with prevalent CVD in a multiethnic population. This finding requires replication in larger studies in which dietary intake and stored serum samples exist.