To evaluate the interobserver agreement of visual coronary plaque characteristics by 320-slice multidetector computed tomography (MDCT) in three populations with low, intermediate and high CAD prevalence and to identify determinants for the reproducible assessment of these plaque characteristics. 150 patients, 50 asymptomatic subjects from the general population (low CAD prevalence), 50 symptomatic non-acute coronary syndrome (non-ACS) patients (intermediate CAD prevalence), and 50 ACS patients (high CAD prevalence), matched according to age and gender, were retrospectively enrolled. All coronary segments were evaluated for overall image quality, evaluability, presence of CAD, coronary stenosis, plaque composition, plaque focality, and spotty calcification by four readers. Interobserver agreement was assessed using Fleiss' Kappa (κ) and intra-class correlation (ICC). Widely used clinical parameters (overall scan quality, presence of CAD, and determination of coronary stenosis) showed good agreement among the four readers, (ICC = 0.66, κ = 0.73, ICC = 0.74, respectively). When accounting for heart rate, body mass index, plaque location, and coronary stenosis above/below 50 %, interobserver agreement for plaque composition, presence of CAD, and coronary stenosis improved to either good or excellent, (κ = 0.61, κ = 0.81, ICC = 0.78, respectively). Spotty calcification was the least reproducible parameter investigated (κ = 0.33). Across subpopulations, reproducibility of coronary plaque characteristics generally decreased with increasing CAD prevalence except for plaque composition, (limits of agreement: ±2.03, ±1.96, ±1.79 for low, intermediate and high CAD prevalence, respectively). 320-slice MDCT can be used to assess coronary plaque characteristics, except for spotty calcification. Reproducibility estimates are influenced by heart rate, body size, plaque location, and degree of luminal stenosis.