Continuous glucose monitor overestimates glycemia, with the magnitude of bias varying by postprandial test and individual – a randomized crossover trial

Abstract

Background

Continuous glucose monitors (CGM) are used to characterize postprandial glycemia, yet no study has directly tested how different test foods/beverages alter CGM accuracy.

Objectives

Assess glycemic responses to test foods/drinks using CGM compared with capillary sampling (criterion).

Methods

Fifteen healthy females (n = 9) and males (n = 6) completed 7 laboratory visits in a randomized crossover design with ≥48 h washout between visits. During each visit, participants consumed an oral carbohydrate challenge comprising either 50 g glucose or equivalent 50 g carbohydrate as whole fruits, 50 g carbohydrate as blended fruit, 50 g carbohydrate as commercially available fruit smoothie, 50 g carbohydrate as commercially available fruit smoothie ingested over 30 ± 4 min, 50 g carbohydrate as commercially available fruit smoothie with 5 g inulin, 30 g carbohydrate as commercially available fruit smoothie. The glycemia was recorded from both CGM and capillary samples every 15 min for 120 min and expressed as incremental areas under the curve. The glycemic index (GI) was calculated relative to 50 g glucose where appropriate. Exploratory analyses examined 1) interindividual heterogeneity of CGM bias compared with criterion and 2) whether CGM bias could be improved with adjustment for baseline differences.

Results

CGM-estimated fasting and postprandial glucose concentrations were (mean ± standard deviation) 0.9 ± 0.6 and 0.9 ± 0.5 mmol/L higher than capillary estimates, respectively(both, P < 0.001). CGM bias varied by postprandial test such that GI for 50 g carbohydrate as commercially available fruit smoothie was higher with CGM (69; 95% confidence interval: 48, 99) compared with capillary (53; 95% confidence interval: 40, 69; P = 0.05). Furthermore, differences in CGM compared with capillary fasting glucose concentrations varied by participant (P = 0.001). Unadjusted, CGM overestimated time >7.8 mmol/L by ∼4-fold, and adjustment for baseline differences reduced this overestimate to ∼2-fold (both P < 0.01).

Conclusions

CGM overestimated glycemic responses in numerous contexts. At times, this can mischaracterize the GI. In addition, there is interindividual heterogeneity in the accuracy of CGM in estimating fasting glucose concentrations. Correction for this difference reduces, but does not eliminate, postprandial overestimate of glycemia by CGM. Caution should be applied when inferring absolute or relative glycemic responses to foods using CGM, and capillary sampling should be prioritized for accurate quantification of glycemic response.

This trial was registered at clinicaltrials.gov as NCT06333184.