Discordance between postprandial plasma glucose measurement and continuous glucose monitoring

Background: There has been growing interest in studying postprandial glucose responses using continuous glucose monitoring (CGM) in nondiabetic individuals. Accurate measurement of glucose responses to meals can facilitate applications such as precision nutrition and early detection of diabetes. Objectives: We aimed to quantify the discordance between simultaneous postprandial glucose measurements made using plasma and CGM. Methods: We studied 10 nondiabetic older adults who randomly consumed 9 predefined meals of varying macronutrient compositions. Glucose was measured for 8 h after the meal by the CGM, blood samples for plasma collection were taken regularly, and plasma glucose was quantified using gold-standard laboratory measurement and a fingerstick blood glucose meter. The primary outcome measured was the mean absolute relative difference (MARD) of CGM and fingerstick measurements relative to the gold standard. Secondary outcomes were Bland-Altman statistics, Clarke Error Grid, and time in range metrics. Additional subgroup analyses were performed by stratifying the postprandial glucose measurements based on the macronutrient composition of the meals. Results: When compared against the gold-standard postprandial glucose measurements, the fingerstick meter was more accurate (MARD: 8.0%; 95% CI: 7.6%, 8.6%) than the CGM (MARD: 13.7%; 95% CI: 13.1%, 14.3%; P < 0.0001). After the meals, Bland-Altman analysis demonstrated that the CGM underestimated the 8-h gold-standard glucose rise by 12.8 mg/dL on average (P < 0.0001), whereas the fingerstick meter did so by 5.5 mg/dL on average (P < 0.0001). The CGM underestimated the time spent in the 70-180 mg/dL range (P = 0.002) and overestimated the time spent <70 mg/dL (P < 0.0001) compared with the other 2 methods. Conclusions: We discovered discordance between gold standard, fingerstick, and CGM in measuring plasma glucose concentrations after a meal. Consequently, emerging applications of CGM in healthy individuals, such as precision nutrition and diabetes onset prediction, may need to account for these discordances.