Accuracy of Continuous Glucose Monitoring in Hemodialysis Patients with Diabetes

Objective:

In the general population continuous glucose monitoring (CGM) provides convenient and less-invasive glucose measurements than conventional self-monitored blood glucose, and results in reduced hypo-/hyperglycemia and increased time-in-target glucose range. However, accuracy of CGM vs. blood glucose is not well-established in hemodialysis patients.

Research Design and Methods:

Among 31 maintenance hemodialysis patients with diabetes hospitalized from 10/2020-5/2021, we conducted protocolized glucose measurements using Dexcom G6 CGM vs. blood glucose, with the latter measured before each meal and at night, plus every 30-minutes during hemodialysis. We examined CGM-blood glucose correlations and agreement between CGM vs. blood glucose using Bland-Altman plots, percentage of agreement, mean and median absolute relative differences (ARDs), and consensus error grids.

Results:

Pearson and Spearman correlations for averaged CGM vs. blood glucose levels were 0.84 and 0.79, respectively; Bland-Altman showed the mean difference between CGM and blood glucose was ~+15 mg/dl. Agreement rates using /20 criteria were 48.7%, 47.2%, and 50.2% during the overall, hemodialysis, and non-hemodialysis periods, respectively. MARD was ~20% across all time periods; median ARD was 19.4% during the overall period and was slightly lower during non-hemodialysis (18.2%) vs. hemodialysis periods (22.0%). Consensus error grids showed nearly all CGM values in clinically acceptable zones A (no harm) and B (unlikely to cause significant harm).

Conclusion:

In hemodialysis patients with diabetes, while MARD values were higher than traditional optimal analytic performance thresholds, error grids showed nearly all CGM values in clinically acceptable zones. Further studies are needed to determine whether CGM improves outcomes in hemodialysis patients.