Abstract
BACKGROUND: Hyperglycaemia after noncardiac surgery is rarely detected outside of the critical care environment, yet occurs commonly and is associated with excess complications including infections and myocardial injury. Systematic, prospectively collected data regarding the accuracy of continuous glucose monitoring commenced immediately before surgery and throughout the early perioperative period are lacking. METHODS: We prospectively enrolled patients aged >50 yr undergoing noncardiac surgery who required at least 24 h of hospital stay. We used real-time continuous glucose monitoring (Dexcom G7 sensor, placed in the upper outer arm) (Dexcom, San Diego, CA, USA) with reference values from arterial blood glucose measurements by amperometry. The primary outcome was the overall mean difference (bias) before surgery, at end of surgery, and 24 h after surgery (Bland-Altman analysis). Secondary outcomes included the mean absolute relative difference and surveillance error grid analyses. RESULTS: We compared paired blood (73% arterial) and continuous glucose monitoring glucose values at each prespecified timepoint in 118 participants (64/118 [54%] female; mean age: 66 [range: 51-89] yr; 25% with diabetes mellitus). The overall bias between continuous glucose monitoring and blood glucose from measurements at each of the three timepoints in the first 24 h after induction of anaesthesia was 0.38 mM (95% confidence interval [95% CI]: 0.23-0.53; n=340 paired readings). Bias decreased from before the start of surgery (1.08 mM [95% CI: 0.87-1.29]; n=116) to 0.15 mM at the end of surgery (95% CI: -0.15 to 0.46; n=113). Mean absolute relative difference ranged from 12.0% to 18.3%. Error grid analyses found that >98% continuous glucose monitoring values were within acceptable risk ranges. CONCLUSIONS: The accuracy of state-of-the-art continuous glucose monitoring is sufficient for perioperative use and could enhance perioperative surveillance of dysglycaemia. CLINICAL TRIAL REGISTRATION: ISRCTN46862025.