Lifestyle interventions and low glycemic diets have potential in diabetes prevention. However, dietary monitoring relies on self-report, which is prone to under-reporting. This observational study investigated the correlation between continuous glucose monitoring (CGM) metrics and glycemic load (GL) or daily macronutrients consumption. Based on one week of CGM data, actigraphy measurements, and food diaries, we investigated correlations between GL per meal, and 19 CGM metrics, selected based on 20 studies identified via a systematic literature review. Furthermore, we generated linear mixed models to predict GL and macronutrients intake using moderately correlated CGM metrics. Forty-eight healthy participants (27 women, average age of 28.2 years, average body mass index (BMI) of 23.4 kg/m2) were included. We found significant positive moderate correlations (P < .0004) between GL and area under the curve (ρ = 0.40, two-hour window), relative amplitude (ρ = 0.40, three hours and ρ = 0.42, four hours), standard deviation (SD) (ρ = 0.41, four hours), and variance (ρ = 0.43, four hours). Significant positive moderate correlations (P < .0004) were observed between carbohydrate and SD (ρ = 0.45), variance (ρ = 0.44), and mean amplitude of glycemic excursions (MAGE) (ρ = 0.40) over 24 hours. We obtained one valid mixed linear model for predicting GL from CGM metrics obtained two hours after food intake. A second model predicted energy intake using moderately correlated CGM metrics, body composition, sleep duration, and physical activity. We demonstrated moderate correlations between GL and CGM metrics in healthy populations. These CGM metrics were successfully used to predict GL or energy intake.
Ong et al. (Tue,) studied this question.