Why Digital Biostatistics is the Future of Diabetes Mellitus and Glucose Management?

Diabetes Mellitus represents a major public health challenge, currently affecting approximately 12% of the U.S. population. A particularly concerning issue, especially in Type 2 Diabetes, is the high rate of undiagnosed cases. In 2020, the Centers for Disease Control and Prevention (CDC) reported that around 21% of diabetes cases in the U.S. remained undetected [2]. Our recent research suggests that this percentage may be even higher [5].

The prevalence of sedentary lifestyles, particularly in developed countries, contributes to a projected increase in diabetes cases. Currently affecting around 9.3% of the global population, the prevalence is expected to rise to 10.2% by 2030 and 10.9% by 2045 . This trend highlights the urgent need for effective public health strategies.

The economic impact is also considerable. In the U.S. alone, the total cost of diabetes is estimated at $412.9 billion, including $306.6 billion in direct medical expenses and $106.3 billion in indirect costs [ 9]. Beyond the financial burden, diabetes significantly affects both life expectancy and quality of life, largely due to late diagnoses and poor glycemic control, often resulting in serious complications such as cardiovascular disease.

There is an increasing need to develop precision medicine and public health strategies based on data-driven approaches. These strategies include novel screening methods and personalized treatment plans using dynamic information from wearable devices and electronic health records.

Ten years ago, personalized medicine in diabetes was not widely implemented [6]. Diabetes Mellitus (DM) encompasses various hyperglycemic conditions, which the American Diabetes Association (ADA) classifies into four major categories. Among these, Type 1

diabetes mellitus (T1DM) is an autoimmune disease characterized by progressive β-cell destruction, ultimately requiring lifelong exogenous insulin therapy [3]. Accurate diagnosis of T1DM—often supported by islet autoantibody testing-and subsequent insulin prescription already represent established forms of personalized medicine. Moreover, emerging evidence suggests that some oral medications originally indicated for Type 2 Diabetes may serve as valuable adjunctive therapies in insulin-requiring T1DM [7]. Despite the identification of more than 40 genes implicated in T1DM, their direct clinical utility remains limited [10].

In recent years, novel classification systems based on clustering analyses have further refined the stratification of both T1DM and Type 2 Diabetes, leading to more precise clinical decision-making [12]. This move toward precision medicine focuses on identifying predictors of differential drug efficacy, enabling clinicians to select the most effective treatment for each patient. Notably, recent studies have revealed significant variability in individual responses to all major non-insulin therapies that follow metformin, including sulfonylureas, thiazolidinediones, DPP-4 inhibitors, GLP-1 receptor agonists, and SGLT2 inhibitors [4]. Reinforcement learning algorithms have also emerged as promising tools for generating personalized therapy recommendations, with some reports indicating they may even surpass traditional dietary interventions such as the Mediterranean diet in improving glycemic control [1] .

Conclusions

The widespread adoption of continuous glucose monitors (CGMs) has been a pivotal development in digital diabetes care, particularly in type 1 diabetes. CGMs have significantly reduced the proportion of time spent in the hypoglycemic range in patients with type 1 diabetes from approximately 5% to 1% especially in children, where insulin loop control systems have shown remarkable success. Clinical trials on AI-based systems for diabetes management are now appearing in leading medical journals [11].

Clearly, the future of glucose management lies in mathematical modeling. The race is on to develop the most accurate, interpretable, and clinically meaningful models capable of capturing glucose fluctuations across different time scales. Functional representations of glucose dynamics, such as the concept of glucodensities [8], offer a novel way to interpret

 

Figure 1: Glucodensity profile of two individuals.

continuous glucose data. These models go beyond simple metrics like time-in-range (e.g., the percentage of time spent between 70-140mg/dL ), which may not sufficiently capture metabolic differences among diverse patient populations.

This ongoing revolution in diabetes research is being driven by AI and mathematical modeling. However, it is crucial to recognize the uniqueness of each patient. The heterogeneity of human physiology and glucose metabolism often limits the practical utility of some deep learning approaches. In medicine, interpretability and clinical relevance must remain top priorities, especially for clinical trials and regulatory agencies.

References

[1] Orly Ben-Yacov, Anastasia Godneva, Michal Rein, Smadar Shilo, Dmitry Kolobkov, Netta Koren, Noa Cohen Dolev, Tamara Travinsky Shmul, Bat Chen Wolf, Noa Kosower, et al. Personalized postprandial glucose response-targeting diet versus mediterranean diet for glycemic control in prediabetes. Diabetes care, 44(9):1980-1991, 2021.

[2] Prevention CDC. National diabetes statistics report: estimates of diabetes and its burden in the united states, 2020. Atlanta, GA: US Deportment of Health and Human Services, 2020.

[3] Jane L Chiang, M Sue Kirkman, Lori MB Laffel, and Anne L Peters. Type 1 diabetes through the life span: a position statement of the american diabetes association. Diabetes care, 37(7):2034, 2014.

[4] John M Dennis. Precision medicine in type 2 diabetes: using individualized prediction models to optimize selection of treatment. Diabetes, 69(10):2075-2085, 2020.

[5] Geronimo Heilmann, Sandra Trenkamp, Clara Möser, Maria Bombrich, Martin Schön, Iryna Yurchenko, Klaus Strassburger, Marcos Matabuena Rodríguez, Oana-Patricia Zaharia, Volker Burkart, et al. Precise glucose measurement in sodium fluoride-citrate plasma affects estimates of prevalence in diabetes and prediabetes. Clinical Chemistry and Laboratory Medicine (CCLM), (0), 2023.

[6] Jeffrey W Kleinberger and Toni I Pollin. Personalized medicine in diabetes mellitus: current opportunities and future prospects. Annals of the New York Academy of Sciences, 1346(1):45-56, 2015.

[7] Harold E Lebovitz. Adjunct therapy for type 1 diabetes mellitus. Nature Reviews Endocrinology, 6(6):326-334, 2010.

[8] Marcos Matabuena, Alexander Petersen, Juan C Vidal, and Francisco Gude. Glucodensities: A new representation of glucose profiles using distributional data analysis. Statistical methods in medical research, 30(6):1445-1464, 2021.

[9] Emily D Parker, Janice Lin, Troy Mahoney, Nwanneamaka Ume, Grace Yang, Robert A Gabbay, Nuha A ElSayed, and Raveendhara R Bannuru. Economic costs of diabetes in the us in 2022. Diabetes Care, 47(1):26-43, 2024.

[10] Flemming Pociot, Beena Akolkar, Patrick Concannon, Henry A Erlich, Cécile Julier, Grant Morahan, Concepcion R Nierras, John A Todd, Stephen S Rich, and Jørn Nerup. Genetics of type 1 diabetes: what's next? Diabetes, 59(7):1561, 2010.

[11] R Paul Wadwa, Zachariah W Reed, Bruce A Buckingham, Mark D DeBoer, Laya Ekhlaspour, Gregory P Forlenza, Melissa Schoelwer, John Lum, Craig Kollman, Roy W Beck, et al. Trial of hybrid closed-loop control in young children with type 1 diabetes. New England Journal of Medicine, 388(11):991-1001, 2023.

[12] Robert Wagner, Martin Heni, Adam G Tabak, Jürgen Machann, Fritz Schick, Elko Randrianarisoa, Martin Hrabě de Angelis, Andreas L Birkenfeld, Norbert Stefan, Andreas Peter, et al. Pathophysiology-based subphenotyping of individuals at elevated risk for type 2 diabetes. Nature medicine, 27(1):49-57, 2021.