AI spots are the risk of hidden diabetes even when the test results seem normal

Millions may miss a risk of early diabetes. Nevertheless, AI models show why your blood sugar spikes could be more important than your test results.

Study: Multimodal Corrélates of glucose points in people with normal regulation of glucose, pre-diabetes and type 2 diabetes. Meanwhile, Image credit: Andrey_popov / Shutterstock

In a recent article published in the journal Nature MedicineThe researchers analyzed data of more than 2. Similarly, 400 people in two cohorts to identify the models in glucose peaks and develop personalized glycemic risk profiles.

They have discovered significant differences in glucose point schemes between people with type 2 diabetes (T2D). Consequently, those with prediabetes or ai spots risk hidden diabetes normoglycemia. Moreover, Their multimodal risk model could help practitioners identify prediabetic individuals at higher risk to develop a T2D.

Diabetes. Meanwhile, prediabetes affect a large segment of the adult American population, but standard diagnostic tools such as glycated hemoglobin (HBA1C) and fasting glucose do not reflect the complete complexity of glucose regulation.

Many factors. In addition, including stress, composition of the microbiome, sleep, physical activity, genetics, diet and age, can influence fluctuations in blood sugar, in particular post-meal tips (defined as an increase of at least 30 mg / dl in 90 minutes), which have been observed even in healthy individuals.

Previous studies have explored these variations using continuous glucose monitoring (CGM). Similarly, but their scope was limited to prediabetic and normoglycemic individuals, often lacking in representation of under-represented groups in biomedical research.

To fill this gap, the study of progress conducted a nationwide clinical trial involving 1,137 various participants ai spots risk hidden diabetes (48.1% of groups historically underrepresented in biomedical research) ranging from normoglycemia to T2D. In addition, The researchers have collected a wide range of self -detached. Therefore, sensors based, including physiological, life, biological, demographic and clinical information.

This multimodal approach allowed the development of a more nuanced understanding of glycemic control. Therefore, the individual variability of glucose points.

The study was aimed at creating complete glycemic risk profiles that could improve early detection. intervention for pre -divaratic individuals at risk of progressing to diabetes, offering a personalized alternative to conventional diagnostic measures such as HBA1C.

The researchers used data from two cohorts: Progress (a digital clinical trial based in the United States). HPP (an Israeli observation study). Progress has entered adults with. without T2D for 10 days of CGM, while collecting data on the intestinal microbiome, genomics, heart rate, sleep, food and activity.

Participants also provided samples of stools, blood and ai spots risk hidden diabetes saliva at home and shared their electronic health files. Exclusion criteria included conditions such as recent use of antibiotics. pregnancy, type 1 diabetes and other health factors that could interfere with CGM or metabolism. Recruitment was completely distant using social networks and invitations to an electronic health file.

CGM data were processed at one -minute intervals and glucose points were defined using specific thresholds. Six key glycemic metrics have been calculated, in particular average glucose, time in hyperglycemia and the duration of points.

Lifestyle data was collected using a food journalization application and portable trackers. Genomics. microbiome data were analyzed using standard tools, and composite measures such as polygenic risk scores and the diversity of microbiomas have been calculated.

An automatic learning model has been developed to assess the T2D risk based on multimodal data (demography. anthropometry, CGM, food intake and intestinal microbiome), and its performance has been tested both ai spots risk hidden diabetes in progress and HPP cohorts. Statistical analyzes included covariance analysis, Spearman correlations and priming for meaning tests and model assessment.

From the 1. 137 participants registered, 347 were included in the final analysis, 174 of which were normoglycemic, 79 were prediabetic and 94 had a T2D.

Researchers have observed significant differences in the metrics of glucose tips in the states of diabetes. such as nocturnal hypoglycemia, the resolution time of tips, the average level of glucose and the time spent in hyperglycemia. These differences have been the most pronounced between T2D. other groups, with pre -divided individuals showing metrics statistically closer to normoglycemia than T2D for key measures such as the frequency of points and intensity.

The diversity of intestinal microbiomas was negatively correlated with most measures of glucose tips. which suggests that a healthier microbiome profile is linked to better glucose control.

The higher rest heart rate. the body ai spots risk hidden diabetes mass index (BMI) and HBA1C were associated with lower glycemic results, while physical activity was linked to more favorable glucose models. Interestingly, the higher carbohydrate intake was associated with a faster point resolution but with more frequent and intense points.

The team has developed a binary classification model based on multimodal data that distinguished the normoglycemic of high -precision. T2D people. When applied to the external data set (HPP). the model has retained high performance, and it has succeeded in identifying substantial variability of risk levels in prediabetic individuals with similar HBA1C values.

These results suggest that multimodal glycemic profiling can improve risk prediction. individual surveillance beyond standard diagnostic tools, in particular for prediabetes.

The study underlines that the traditional diagnoses of diabetes. such as HBA1C, fail to grasp individual variations in glucose metabolism.

Using CGM alongside multimodal data based on genomics. lifestyle and microbiome, researchers have identified significant differences ai spots risk hidden diabetes in glucose peaks through normoglycemic, prediabetic and T2D individuals, with prediabetics showing a stronger similarity with normoglycemia than with T2D in several key metrics.

The multimodal risk model focused on automatic learning. validated in an external cohort, has revealed a great variability of the risk in prediabetic individuals with identical HBA1C levels, supporting its added value compared to standard metrics.

The forces include the cohort of decentralized and diversified progress (with a representation of 48.1% of under-represented groups) and the collection of real world data. However. the limits imply potential biases of the variability of devices, inaccuracies in typing biases, self -declared prejudices, food challenges and the use of antihyperglycemic drugs.

A larger validation and longitudinal research is necessary to confirm predictive usefulness and clinical relevance.

In the end. this study demonstrates the potential of remote multimodal data to improve early detection, prediabetes risks personalized T2D stratification and prevention, paving ai spots risk hidden diabetes the way for more precise and inclusive diabetes care.