Cardiac output (CO) measurement is crucial for the guidance of therapeutic decisions in critically ill and high-risk surgical patients. Newly developed completely non-invasive CO technologies are commercially available; however, their accuracy and precision have not recently been evaluated in a meta-analysis. We conducted a systematic search using PubMed, Cochrane Library of Clinical Trials, Scopus, and Web of Science to review published data comparing CO measured by bolus thermodilution with commercially available non-invasive technologies including pulse wave transit time, non-invasive pulse contour analysis, thoracic electrical bioimpedance/bioreactance, and CO2 rebreathing. The non-invasive CO technology was considered acceptable if the pooled estimate of percentage error was <30%, as previously recommended. Using a random-effects model, sd, pooled mean bias, and mean percentage error were calculated. An I2 statistic was also used to evaluate the inter-study heterogeneity. A total of 37 studies (1543 patients) were included. Mean CO of both methods was 4.78 litres min−1. Bias was presented as the reference method minus the tested methods in 15 studies. Only six studies assessed the random error (repeatability) of the tested device. The overall random-effects pooled bias (limits of agreement) and the percentage error were −0,13 [−2.38 , 2.12] litres min−1 and 47%, respectively. Inter-study sensitivity heterogeneity was high (I2=83%, P<0.001). With a wide percentage error, completely non-invasive CO devices are not interchangeable with bolus thermodilution. Additional studies are warranted to demonstrate their role in improving the quality of care.