Eleven population-based studies (54,336 patients) and 27 diagnostic cohort studies (4,787 patients) met the inclusion criteria. Follow-up ranged from 0.1 to 15.1 years in the population-based studies. The quality score ranged from 16 to 27 out of 27 for population based studies and eight to 14 out of 15 for diagnostic cohort studies.

__Population-based studies__: Ultrasound assessment of carotid plaque had a higher diagnostic accuracy for the prediction of future myocardial infarction than that of CIMT; the AUC for carotid plaque was 0.64 (95% CI 0.61 to 0.67) and for CIMT 0.61 (95% CI 0.59 to 0.64). The higher diagnostic accuracy of carotid plaque remained significant after adjusting for covariates in multivariate meta-regression. Negative predictive values of carotid plaque for future myocardial infarction over 10 years was also higher (96.0%, 95% CI 95.3 to 96.4) than for CIMT (95.2%, 95% CI 94.5 to 96.0). The annual event rates of myocardial infarction after negative results were 0.40% (95% CI 0.36 to 0.47) for carotid plaque and 0.47% (95% CI 0.42 to 0.55) for CIMT.

__Diagnostic cohort studies__: Ultrasound assessment of carotid plaque had a higher diagnostic accuracy (AUC 0.76, 95% CI 0.73–0.80) for the prediction of coronary artery disease than that of CIMT (AUC 0.74, 95% CI 0.72–0.76); the difference did not remain significant after adjusting for covariates in multivariate meta-regression. Pre-specified subgroup analyses showed CIMT measured in both the carotid bulb and internal carotid artery (AUC 0.79, 95% CI 0.77–0.81) had a higher diagnostic accuracy than when CIMT was measured only in the carotid artery (AUC 0.6, 95% CI 0.65–0.71).

Results of *post hoc* sensitivity analyses (details given) did not substantially differ from the main analysis.

Publication bias was not observed in the population-based studies, but this may have been present for the diagnostic cohort studies; it was estimated that six studies needed to be imputed to make the plot symmetrical.