The evidence linking low-carbohydrate diets (LCD) to CVD is controversial, and results from epidemiological studies are inconsistent. We aimed to assess the relationship between LCD patterns and coronary artery Ca (CAC) scores from computed tomography in the Multi-Ethnic Study of Atherosclerosis cohort. Our sample included 5614 men and women free of clinical CVD at baseline (2000–2002), who had a FFQ, a baseline measure and ≥1 measure of CAC during follow-up. We excluded those with implausible energy intake or daily physical activity. The overall, animal-based and plant-based LCD scores were calculated based on intakes of macronutrients. Relative risk regression and robust regression models were used to examine the cross-sectional and longitudinal relationship between LCD score quintile and CAC outcomes, after adjustment for multiple cardiovascular risk factors. The mean age of participants was 63 years. The median intakes of total carbohydrate, fat and protein were 53·7, 30·5 and 15·6 % energy/d, respectively. Among 2892 participants with zero CAC scores at baseline, 264 developed positive scores during 2·4-year follow-up (11–59 months). Among those with positive scores at baseline, the median increase in CAC was 47 units over the course of follow-up. The overall, the animal-based and the plant-based LCD scores were not associated with CAC prevalence, incidence and progression. In conclusion, diets low in carbohydrate and high in fat and/or protein, regardless of the sources of protein and fat, were not associated with higher levels of CAC, a validated predictor of cardiovascular events, in this large multi-ethnic cohort.