CONTEXT AND OBJECTIVE Although glucometers have not been validated for intensive care units, they are regularly used. The aim of this study was to compare and assess the accuracy and clinical agreement of arterial glucose concentration obtained using colorimetry (Agluc-lab), capillary (Cgluc-strip) and arterial (Agluc-strip) glucose concentration obtained using glucometry and central venous glucose concentration obtained using colorimetry (Vgluc-lab). DESIGN AND SETTING Cross-sectional study in a university hospital. METHOD Forty patients with septic shock and stable individuals without infection were included. The correlations between measurements were assessed both in the full sample and in subgroups using noradrenalin and presenting signs of tissue hypoperfusion. RESULTS Cgluc-strip showed the poorest correlation (r = 0.8289) and agreement (-9.87 +/- 31.76). It exceeded the limits of acceptable variation of the Clinical and Laboratory Standards Institute in 23.7% of the cases, and was higher than Agluc-lab in 90% of the measurements. Agluc-strip showed the best correlation (r = 0.9406), with agreement of -6.75 +/- 19.07 and significant variation in 7.9%. For Vgluc-lab, r = 0.8549, with agreement of -4.20 +/- 28.37 and significant variation in 15.7%. Significant variation was more frequent in patients on noradrenalin (36.4% versus 6.3%; P = 0.03) but not in the subgroup with hypoperfusion. There was discordance regarding clinical management in 25%, 22% and 15% of the cases for Cgluc-strip, Vgluc-lab and Agluc-strip, respectively. CONCLUSION Cgluc-strip should be avoided, particularly if noradrenalin is being used. This method usually overestimates the true glucose levels and gives rise to management errors. CLINICAL TRIAL REGISTRATION ACTRN12608000513314 (registered as an observational, cross-sectional study).