A mobile app tool that generates personalized predictions based on individual features can better predict a person’s postprandial glycemic response to diverse foods vs. models based on calorie and carbohydrate content, according to findings published in JAMA Network Open.
“Our tool generating personalized predictions provides individuals with access to actionable, person-specific information for managing their blood glucose levels and maintaining a normoglycemic status,” Helena Mendes-Soares, PhD, a research associate and assistant professor in surgery at Mayo Clinic in Rochester, Minnesota, and colleagues wrote in the study background. “Personalized [postprandial glycemic response] predictions are accessible in the form of a mobile or web-based application, enabling real-time assessment of influences of foods and combinations of foods on blood glucose levels of the individual at the time of consumption.”
Mendes-Soares and colleagues analyzed data from 327 healthy adults without diabetes with access to a mobile device and web browser, recruited between October 2016 and December 2017 for a 6-day study (mean age, 45 years; 78% women). Researchers asked participants to use the DayTwo Food and Activity Logger, a mobile app that logged food and activity information through the week and contained the MyNetDiary food catalogue, as well as wear a continuous glucose monitor (iPro2, Medtronic) and complete manual glucose monitoring (Contour Next Link, Bayer) at least four times per day for its calibration. Participants were asked to maintain normal eating habits except for four standardized meals provided by the researchers, to be consumed as the first meal of the day. Researchers used logged meal times and CGM measurements to calculate area under the curve in the 2 hours after a meal, truncated to the range of 0 mg/dL to 80 mg/dL. Pearson product-moment correlation was used to quantify the accuracy of the predicted postprandial glycemic responses from the model relative to those obtained by CGM, as well as the correlation between the calorie and carbohydrate content in a meal and postprandial glycemic responses estimated from the CGM measurements.
Two days before the beginning of the study week, participants provided a stool sample to analyze microbiome composition.
Within the cohort, mean number of calories logged per meal was 400.1 kcal, and mean carbohydrate content was 37.9 g. For meals used to train the predictive model, carbohydrates composed 43% of nutrients, proteins composed 18% and lipids composed 39%, according to researchers.
The researchers found that glycemic excursions varied widely among participants after consuming a standardized meal of bagel and cream cheese, ranging from 6 mg/dL to 94 mg/dL (mean, 30.7 mg/dL). Carbohydrate sensitivity, defined as the correlation between the carbohydrate content of a meal and the postprandial glucose response, also varied widely, according to researchers; however, there was intraindividual reproducibility of the glycemic response to the bagel and cream cheese meal (Pearson product-moment correlation R = 0.66).
Using information from the participants, researchers retrained a model for personalized, predictive postprandial glycemic responses, using a framework applied to an Israeli population, and tested the predictions against the data set using 10-fold cross-validation. Researchers found that the correlation between postprandial glycemic responses in the model and in those observed in participants was higher vs. the correlation observed between postprandial glycemic responses and the calorie content of the meals (R = 0.62 vs. 0.34) or the carbohydrate content of the meals (R = 0.4).
“Given the R = 0.66 intraindividual correlation observed for standardized meals, our model appears to explain a large percentage of the explainable variance,” the researchers wrote.
Additionally, receiver operating characteristic curves across three threshold values, representing the 50th, 75th and 90th percentiles of all measured postprandial glycemic responses, demonstrated that the researchers’ predictive model “consistently and substantially outperformed the models based on the calorie and carbohydrate content of the meals consumed,” they noted.
“Follow-up clinical studies assessing the changes induced by this approach compared with current practices on cardiometabolic markers of diabetes risk will be required to assess the long-term health benefits of using this tool,” the researchers wrote. “Nonetheless, results presented herein point toward the potentially significant contribution of measurement-based personalized approaches across different populations in harnessing nutrition as a means of improving [postprandial glycemic response], with subsequent reduction of the consequences of prolonged and repetitive exposure to hyperglycemia.” – by Regina Schaffer
Disclosures: The Center for Individualized Medicine at the Mayo Clinic and DayTwo funded this study. Mendes-Soares reports the Mayo Clinic has a financial interest in DayTwo and has received grants and nonfinancial support from the company. Please see the study for all other authors’ relevant financial disclosures.