Prediabetes represents a substantial burden worldwide, particularly in the Asia Pacific region.¹ Recent 2024 estimates indicate that impaired glucose tolerance (IGT) affects 12% of adults, while impaired fasting glucose (IFG) affects 9.2%, translating into hundreds of millions of individuals at increased risk of developing type 2 diabetes (T2D).¹ At ATTD-Asia 2025, Professor Chow, Yee-Kuan Elaine, Clinical Associate Professor in the Department of Medicine and Therapeutics at The Chinese University of Hong Kong, examined how artificial intelligence (AI) and wearable technologies may support earlier detection, improved risk stratification, and more personalized approaches to T2D prevention.¹

The long-term implications of prediabetes are considerable.¹ Modelling data presented during the session showed that a 20-year-old with prediabetes may spend more than half of their remaining lifetime living with T2D, highlighting the importance of early identification and timely intervention.¹ While intensive lifestyle intervention (ILI) has been shown to reduce progression to T2D by 40%-50% in landmark trials such as the Da Qing Diabetes Prevention Study and the US Diabetes Prevention Program, Prof. Chow noted that translating these programs into routine practice remains challenging.¹ Such interventions are resource-intensive, difficult to scale, and often lack personalization, contributing to suboptimal engagement and retention in real-world settings.¹

Challenges also exist in the diagnosis of prediabetes.¹ Current definitions vary across organizations, and reliance on fasting plasma glucose or HbA1c alone may fail to identify individuals at highest risk, particularly in East Asian populations where post-load dysglycemia is more prevalent.¹ Evidence from longitudinal studies indicates that abnormalities in post-prandial glucose often precede rises in fasting glucose, suggesting that conventional static measures may miss early metabolic dysfunction.¹ In this context, continuous glucose monitoring (CGM) offers a complementary approach to screening and diagnosis.¹ CGM provides dynamic information on glucose patterns, capturing both fasting and post-prandial excursions, and can be self-administered in the home setting, potentially improving access and uptake.¹ Importantly, CGM-derived metrics such as glucose variability have been shown to reflect underlying abnormalities in insulin secretion and insulin sensitivity, even in individuals without overt T2D.¹

Beyond detection, CGM metrics are increasingly being explored as meaningful endpoints in prediabetes intervention studies.¹ Prof. Chow highlighted recent work demonstrating how dietary interventions, when evaluated using CGM-based post-prandial glucose measures, can reveal differential metabolic responses not apparent from fasting glucose alone.¹ These findings support the role of CGM in refining both risk assessment and intervention evaluation.¹

The integration of AI with CGM and other wearable devices further extends these capabilities.¹ AI algorithms can process large volumes of multidimensional data, including glucose patterns, physical activity, dietary inputs, and contextual information—to identify correlations and predict individual post-prandial glucose responses.¹ This enables more personalized lifestyle guidance, including precision nutrition strategies tailored to individual metabolic phenotypes.¹ Emerging evidence also suggests that machine-learning models incorporating clinical and microbiome features may help distinguish sub-phenotypes of prediabetes and guide targeted interventions.¹

Real-world and clinical studies presented during the session illustrated the feasibility of these approaches.¹ Digital health programs integrating CGM, activity trackers, smart scales, and AI-enabled coaching have demonstrated improvements in glycemic measures over time.¹ In addition, pragmatic trial had shown that AI-led T2D prevention programs can achieve outcomes comparable to human-led interventions, supporting their potential role in extending preventive care to larger populations.¹ Looking ahead, Prof. Chow outlined future developments in wearable technology and AI, including multimodal sensors, improved meal recognition, pre-emptive behavioral nudging, and culturally contextualized interventions.¹ She also emphasized the need for training algorithms on diverse datasets and ensuring equitable access, robust validation, and appropriate human oversight.¹

In conclusion, while traditional T2D prevention programs are effective, they face limitations in scalability and personalization.¹ The use of CGM, multimodal wearables, and AI analytics offers new opportunities for earlier detection, personalized lifestyle intervention, and sustained engagement in prediabetes care.¹ When combined with pharmacological strategies and broader public health efforts to address obesogenic environments, these innovations may play an important role in shaping future approaches to T2D prevention.¹

In an interview with Omnihealth Practice, Professor Chow discussed how AI-enabled digital tools and wearables are reshaping early detection and prevention strategies in prediabetes, particularly in Asian populations.

Q1. Why is early detection of prediabetes especially challenging in APAC populations, and why is early intervention important?

Professor Chow: Prediabetes is usually silent and asymptomatic. In many Asian populations, the earliest abnormality is isolated postprandial glucose elevation, while fasting glucose and HbA1c often remain normal. As a result, standard screening can miss many at-risk individuals, and dynamic tests such as OGTT, which are less practical in routine care, are often required. Early intervention is crucial because these changes reflect the earliest loss of beta-cell function, when glucose abnormalities are still reversible. Timely intervention at this point can restore normal glucose tolerance and prevent progression to diabetes and related complications.

Q2: How can CGM patterns and real-time feedback guide early lifestyle interventions in prediabetes, and what are the potential behavioral benefits and risks of CGM use?

Professor Chow: CGM allows individuals and clinicians to visualize real-time glucose patterns, particularly postprandial responses to foods and the glucose-lowering effects of physical activity. This immediate feedback increases awareness and supports informed, personalized choices, such as substituting foods that trigger large glucose spikes or using exercise strategically to improve glucose control. Such feedback can reinforce motivation and promote positive behavioral change. However, responses to CGM vary. Some individuals may become anxious or distressed by frequent alerts and over-monitor their glucose, while others may disengage and stop using the device. Therefore, CGM use should be flexible and tailored to the individual, with appropriate guidance to maximize benefit while minimizing anxiety.

Q3: How do AI- and digital-led programs enhance prediabetes prevention and shared decision-making compared with traditional interventions, and what practical barriers must be addressed for routine adoption?

Professor Chow: AI- and digital-led programs improve diabetes and prediabetes prevention by increasing reach and flexibility. They allow lifestyle education and personalized feedback to be delivered at scale, while supporting shared decision-making through joint interpretation of health data. However, many tools operate in a regulatory grey zone as consumer wellness products with limited transparency on how AI models are trained and validated, particularly in Asian populations. Engagement can also be lower in purely digital programs without human support. Clear regulatory frameworks, population-specific evidence, and hybrid models that combine digital tools with healthcare professional input are needed for wider adoption.