Cardiac Digital Twins Could Revolutionize Diabetes Heart Treatment
A landmark review maps how computational heart models can decode diabetic cardiomyopathy and predict responses to SGLT2i and GLP-1 drugs.
Summary
Diabetes now affects 830 million people worldwide and dramatically raises cardiovascular risk through complex, multi-scale changes from individual proteins to the whole circulatory system. This review from Imperial College London is the first to comprehensively map how cardiac computational models — digital twins of the heart — can simulate diabetic cardiomyopathy and predict responses to treatments like SGLT2 inhibitors and GLP-1 receptor agonists. These models span metabolic, electrophysiological, mechanical, and hemodynamic scales. The authors argue that such tools could accelerate drug development, reduce reliance on expensive cardiovascular outcome trials, and enable personalized treatment strategies for diabetic patients with varying cardiovascular risk profiles.
Detailed Summary
Diabetes has grown from affecting 200 million people in 1990 to 830 million in 2022, making it one of the most urgent global health crises. Both type 1 and type 2 diabetes substantially increase cardiovascular disease risk — including hypertension, coronary artery disease, heart failure, and arrhythmias. A distinct subset of patients develops diabetic cardiomyopathy: myocardial dysfunction occurring in the absence of other cardiovascular risk factors. Critically, diabetic women face higher CVD risk than diabetic men, and early-onset type 2 diabetes carries a more aggressive cardiovascular burden than late-onset disease, complicating risk stratification and treatment selection.
The pathophysiology of the diabetic heart is strikingly multi-scale. At the cardiomyocyte level, chronic hyperglycemia impairs glucose uptake, upregulates free fatty acid (FFA) metabolism, and triggers lipotoxicity and mitochondrial dysfunction. This reduces ATP availability, disrupts calcium handling proteins — including SERCA, NCX, L-type calcium channels, and ryanodine receptors — and causes ionic remodeling with reduced potassium currents and altered sodium currents. The net result is prolonged action potential duration, intracellular calcium and sodium overload, increased arrhythmia risk, and contractile dysfunction. At the organ level, these cellular changes combine with collagen accumulation and fibrosis to produce concentric remodeling, diastolic dysfunction, atrial dilation, and hypertrophy.
The review provides the first comprehensive taxonomy of cardiac computational models applied to diabetes. These range from metabolic models simulating ATP production from glucose and FFAs within cytosol and mitochondria, to action potential models capturing ionic remodeling, to calcium handling models, to whole-heart electromechanics models incorporating patient-specific geometries derived from imaging data. Crucially, these can be coupled with circulatory system models to capture hemodynamic effects of diabetic vascular disease — including increased arterial stiffness, impaired microvascular vasodilation, and elevated preload from diabetic nephropathy.
A major focus is the cardiac effects of SGLT2 inhibitors and GLP-1 receptor agonists, both of which have demonstrated substantial cardiovascular benefits in outcome trials but whose precise cardioprotective mechanisms remain incompletely understood. SGLT2i are hypothesized to act through osmotic diuresis reducing preload, direct myocardial metabolic effects shifting substrate utilization, and inhibition of the sodium-hydrogen exchanger. GLP-1 RAs may act through direct cardiomyocyte receptors, anti-inflammatory effects, and hemodynamic improvements. Computational models offer a platform to test these competing hypotheses in silico, potentially replacing or augmenting expensive cardiovascular outcome trials that currently require years and thousands of patients.
The authors outline a roadmap for next-generation multi-scale, multi-physics digital twins that integrate metabolism, electrophysiology, mechanics, and perfusion. These models could incorporate sex differences, obesity, and heart failure comorbidities to generate personalized cardiovascular risk predictions. Key challenges include model validation against clinical data, computational cost, and regulatory acceptance for clinical deployment. Despite these hurdles, the authors argue that cardiac digital twins represent a transformative opportunity to accelerate therapeutic development and enable precision medicine for the 830 million people living with diabetes globally.
Key Findings
- Diabetes prevalence has surged from 200 million in 1990 to 830 million in 2022, with CVD remaining the leading cause of death in this population
- Early-onset type 2 diabetes carries significantly higher myocardial infarction and stroke risk than both late-onset type 2 and type 1 diabetes
- Diabetic women face disproportionately higher CVD risk than diabetic men, though the underlying mechanisms remain poorly characterized
- Diabetic cardiomyocytes show reduced transient outward K+ current, fast and slow rectifier K+ currents, and altered late Na+ current — all contributing to prolonged APD and arrhythmia risk
- SERCA, NCX, L-type calcium channel, and ryanodine receptor activity are all reduced in diabetic myocytes, causing intracellular Ca2+ and Na+ overload and contractile dysfunction
- SGLT2 inhibitors and GLP-1 receptor agonists show substantial cardiac benefits in CVOTs, but their direct vs. indirect cardioprotective mechanisms remain unresolved — a gap computational models could address
- This is the first review to systematically map cardiac computational models across all scales (metabolic, electrophysiological, mechanical, hemodynamic) as applied to diabetic heart disease
Methodology
This is a narrative review article from Imperial College London's National Heart and Lung Institute, synthesizing existing literature on cardiac computational models applied to diabetes and anti-diabetic treatments. The review covers models spanning subcellular metabolism, action potential dynamics, calcium handling, whole-heart electromechanics, and circulatory system hemodynamics. No original experimental data or patient cohorts were analyzed; the authors draw on animal studies (rodent, rabbit), human imaging datasets, and published computational modeling frameworks. No formal statistical methods or meta-analytic pooling were employed.
Study Limitations
As a narrative review, this paper does not perform systematic literature search or meta-analysis, introducing potential selection bias in the studies cited. The computational models described are largely validated in animal models or small human cohorts, and their clinical translation faces significant hurdles including regulatory acceptance, computational cost, and the need for large-scale validation datasets. The authors acknowledge that consensus on the definition and existence of diabetic cardiomyopathy itself remains lacking, complicating model development and validation targets.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.