#2103 Accuracy of a continuous glucose monitor in people with diabetes on peritoneal dialysis

Abstract Background and Aims Globally 40% of people receiving peritoneal dialysis (PD) have a diagnosis of diabetes. Good glycaemic control can by complicated by the obligate glucose load associated with PD. Monitoring and treatment of people with diabetes (PwD) on PD has been limited by reduced reliability of traditional markers of glycaemia such as HbA1c in end-stage kidney disease. Continuous glucose monitors (CGMs) have been increasingly used in the care of PwD over the last 20 years. Monitors worn on the upper arm or abdomen measure glucose in interstitial fluid at regular intervals over the 24 hour period. CGMs are currently unlicensed in PD populations as they have not been tested in large scale studies. Concerns remain around accuracy in the presence of potentially interfering substances generated by the breakdown of osmotic agents used in PD fluid (e.g. Icodextrin). We aimed to assess the analytical and clinical accuracy of the FreeStyle Libre(FSL) CGM system in PwD on maintenance PD. Method We conducted a single centre observational study of 10 PwD who had been on PD for at least 3 months. For the 10 day study period participants wore a blinded research model FSL monitor which measured interstitial glucose levels every 15 minutes. Results from the FSL were compared with; Analytical accuracy was assessed using three methods; Clinical accuracy was assessed using the Parkes consensus error grid to assign each pair of readings to one of five risk zones and quantify the clinical implications of any difference between the two measurements. The International Organisation for Standardisation guidelines stipulates 99% of pairs should be in zones A and B, denoting little or no effect on clinical actions. Results Of 11 recruited participants, 10 completed the 10 day study period resulting in 76 paired CGM-YSI measurements and 375 paired CGM-CBG measurements. 7 participants were on APD and 3 on CAPD. All participants were prescribed a PD regime that included one daily dwell with Icodextrin. The aggregate MARD was 9.4% (95% CI 7.9, 10.4) for CGM-YSI matched pairs and 17.6% (95% CI 17, 19.3) for CGM-CBG matched pairs. Spearman's r was 0.94 (p < 0.0001) for CGM-YSI and 0.89 (p < 0.0001) for CGM-CBG. From Bland-Altman analysis, the systematic error of the CGM compared with YSI and CBG was −0.81 mmol/l (SD 0.81) and −1.6 mmol/l (SD 1.9) respectively. Parkes consensus error grid analysis showed that compared with YSI and CBG respectively, 100% and 99.9% of sensor values were in the clinically acceptable zones A and B (Fig. 1). Conclusion We have demonstrated satisfactory performance of the FSL monitoring system by both analytical and clinical metrics in this cohort of PD patients using treatment regimens including Icodextrin based fluids. CGMs provide much more information for the PwD and their clinical care team compared with random CBGs (which in themselves have been shown to exhibit significant between user variability) and time averaged markers such as HbA1c. They allow detection of asymptomatic hypoglycaemia and direct monitoring of the systemic effects of any change to the PD prescription resulting in a larger or smaller glucose load. These results should be reassuring for clinicians wishing to utilise the greater depth of information provided by CGMs to improve clinical care for their patients with diabetes on PD.