This article was first published on the Asia-Pacific Biotech News. Click here to view the original article.
Leveraging AI to Tackle the Looming Metabolic Syndrome Crisis
Metabolic syndrome- related conditions, including atherosclerotic cardiovascular disease (ASCVD), type 2 diabetes ( T2D), chronic kidney disease (CKD), and non-alcoholic fatty liver ( NAFLD/NASH), affect one in three adults and cause over 40 per cent of the world’s fatalities .1 They consume US$1.9 trillion2 of global healthcare spend and patient volumes will grow 40 per cent over the next decade ,3 driven by an ageing population, sedentary lifestyles, and over -nutritious diets. Based on pre-COVID statistics, their global healthcare cost burden is forecast to surpass US$5.5 trillion in 2040,2 now likely to be an under-estimate since COVID survivors have a greatly increased risk of cardiac events ,4 T2D5 and CKD6 — a looming global healthcare crisis!
In this article, we outline metabolic syndrome disease dynamics, current treatment strategies, and consequent health economic impact, using kidney disease to illustrate. Then, we explain the key concepts of patient stratification and precision medicine. We conclude by describing how our companies are using AI to address this disease challenge and eventually transform patients’ lives.
Metabolic syndrome conditions stem from a common set of risk factors: dysglycemia, dyslipidaemia, excess weight, and hypertension. A significant variation is observed in patients’ disease journeys regarding the sequence of symptoms ( e.g. hypertension, excess lipids, blood glucose, kidney function), conditions diagnosed (e.g. ASCVD, T2D, CKD), and serious downstream complications ( e.g. heart attack, stroke, vision impairment, foot amputation, kidney failure) plus varying rates of disease progression. For example, Aisha and Mei are two patients whose type 2 diabetes is known to affect their kidney function and have been diagnosed with diabetic kidney disease7 (DKD). Both are aged 50, non -smokers, have sensible diets, and faithfully follow their medication regimen . Over time, Mei struggled to control her blood pressure and cholesterol levels while Aisha experienced worsening kidney function and anaemia . Mei suffered a stroke within seven years whilst Aisha progressed over the next decade to end-stage renal disease (ESRD ), requiring lifelong dialysis. Mei and Aisha ostensibly have the same disease but exhibited markedly different disease progression patterns — these “clinical phenotypes” are examples of distinct DKD patient sub-populations that have been documented in the scientific literature.8
The modern scientific view is that each individual patient experiences one multi-faceted disease journey9 instead of multiple co-morbidities. Since someone can, for example, simultaneously exhibit a T2D phenotype, a NAFLD phenotype and an ASCVD phenotype, the number of potential composite phenotypes is in the hundreds. Within each such phenotype, there could be multiple “patient endotypes”, sub-groups whose progression is driven by different underlying molecular-level factors including genetics. Phenotypes describe variation in the patient’s observed disease journey. Endotypes explain variation in treatment response when patients of otherwise the same phenotype are treated with the same medications.
Current treatment strategies address each symptom separately, e.g. hypertension, dyslipidaemia, dysglycemia or nephropathy. Current drugs only control specific symptoms and reduce the risk of serious complications. For example, many patients are prescribed statins to lower their LDL cholesterol. But a recent study10 showed that 44 per cent of such patients will nevertheless experience a major cardiac event within seven years . Physicians tend to see which disease journey their patient is taking well after that journey is started, by which time much damage might have already occurred. Treatments are one-size -fits-all, not aimed at specific phenotypes or endotypes, with trial-and-error used to find the right set of medications for each patient.
Patients deserve better, and the cost burden on healthcare payers is immense. Consider CKD, a disease which leads to kidney failure and either lifelong dialysis or kidney transplantation. Medicare in the US spends US$80,000 for a kidney transplant and US$65,160 per annum to maintain a patient on dialysis. The latter figure is US$95,785 if the patient is also diabetic. Incidence of heart failure , a common complication, adds a further US$26,064. Medicare is a government programme; private sector figures will be higher.
CKD prevalence is growing, ranking 19th globally amongst diseases for years of life lost, up from 36th in 1990. In Asia , the disease ranks even higher at 14th .11 Singapore ranks 4th12 in the world for the prevalence of kidney failure, with an average 5.7 new patients diagnosed daily. Singapore is also ranked highest globally13 for DKD, with close to two in three new cases of kidney failure attributed to diabetes. The annual global spend on DKD is forecast to grow from US$300 billion today to US$900 billion by 2040.
Patient Stratification and Precision Medicine (“PS+PM”)
Patient stratification assigns patients into distinct groups, e.g. clinical phenotypes or patient endotypes. Precision medicine comprises early diagnosis and targeted treatment strategies based on each stratified group’s unique characteristics. Patient stratification and precision medicine (“PS+PM”) operate hand-in-hand to improve health outcomes through earlier detection and optimal treatments for each patient group. Total healthcare costs are also reduced by treating patients earlier and avoiding waste on sub-optimal treatments.
Each metabolic syndrome endotype’s distinct disease journey results from the complex interactions of the corresponding patients’ genetics with their cumulative environmental and lifestyle experiences. The latter factors affect whether and when certain genes actually express the proteins that run our bodies’ cells . Such interactions are best understood by investigating the patients’ DNA together with their mix of proteins and their metabolites (by -products of their metabolic processes). This is a “deep data ” endeavour, collecting and analysing rich de-identified clinical and genetic data from a large number of patients, necessitating the use of a set of software algorithms colloquially referred to as “artificial intelligence” (AI).
Broadly speaking, there are three PS+PM strategies. Existing laboratory diagnostics have limited accuracy and cannot predict the future course of the patient’s disease. Hence, an effective first strategy involves stratifying patients up-front into phenotypes with different underlying risks of subsequent serious complications, to prioritise the higher-risk ones for treatment well before their condition deteriorates. Doing this focuses scarce healthcare resources on patients that need it most.
For example, urine microalbuminuria and estimated glomerular filtration rate (eGFR) are traditionally used to assess kidney function . However, high variability and insufficient sensitivity have been observed using albuminuria .14 And eGFR readings can be swayed15 by biological variation of creatinine — studies have shown it can significantly over-estimate GFR16 in some situations, implying a healthier patient than the true reality. Hence, current diagnostic markers are potentially too late or limited in identifying CKD onset. And these markers’ historical gradients are unreliable at predicting the future course of the disease. In response, risk-scoring algorithms have been developed, applying AI to a wider set of patient markers to stratify patients according to their risk of future kidney function deterioration.
A second strategy is to further stratify the higher-risk phenotypes into sub-phenotypes or endotypes that respond best to the different treatment options. This ensures the best possible outcome for the patients with what is currently in the physician’s arsenal. For example, DKD patients are treated with T2D and hypertension medications, sometimes multiple drugs for each symptom. There are over 30 different approved T2D drugs in 11 drug classes, and nearly 70 different approved hypertension drugs in 16 drug classes.17 Significant variability in treatment response and patient tolerance for side effects are observed in every drug class. Physicians currently rely on experience18 and trial- and-error to find the right combinations for individual patients. AI techniques applied to large datasets derived from real-world patients are the basis for pursuing this second strategy.
Since many existing drugs merely control symptoms and do not sufficiently address underlying disease drivers, many clinical situations abound, where even the most appropriate available treatments are unsatisfactory. In these scenarios, a third strategy is deployed — investigating certain endotypes and their causal drivers to devise better treatments for those endotypes. Once again, AI deployed on large datasets is the key, with the focus on those clinical situations exhibiting the highest unmet needs.
Mesh Bio’s AI-enabled Approach
Mesh adopts the first and second PS+PM strategies by identifying the most high-risk patients and suggesting more optimal interventions. Its clinical decision support analytics and workflow automation solutions such as DARA® enable data-driven care delivery.
DARA® summarises and visualises findings from a patient’s health screening results by automating complex, tedious risk analyses. Its built-in clinical and lifestyle recommendations engine (covering thousands of medical scenarios) directs patients to targeted care and wellness services. Patients’ risk trends for chronic conditions such as ASCVD and DKD are projected for the next 10 years. The software platform is being further developed to optimise treatment strategies for high-risk patients using existing drugs and interventions, utilising predictive analytics on multidimensional health data from real-world patients.
Mesh’s software platform provides deeper patient insights to primary care physicians, enabling more informed treatment decisions. The visualisations, risk projections and lifestyle recommendations are greatly appreciated by both physicians and patients. And existing care provider customers have reported significant operational efficiency improvements and revenue increases as a result of deploying the platform.
MultiOmic Health’s AI-enabled Approach
MultiOmic adopts the second and third PS+PM strategies. Its AI-enabled MOHSAIC® platform characterises specific patient endotypes to innovate improved or altogether new treatment concepts targeting the underlying causal drivers. These endotype-specific treatment concepts are implemented in partnership with established pharma companies, using either existing drugs (including unapproved ones from prematurely-cancelled R&D programmes previously trialled as one-size-fits-all treatments) or brand new drugs.
Via partnerships with biobanks, healthcare providers and healthtech companies, MultiOmic is assembling rich datasets of clinical and omics data in targeted clinical situations. Its partners provide de-identified clinical information and bio-samples. To ensure comparable data across multiple sources, MultiOmic generates genetic, epigenetic, proteomic, and metabolomic data from these samples using its network of carefully-selected technology providers. MOHSAIC® combines AI-enabled predictive analytics, systems biology modelling (to simulate causal pathways) and targeted validation in wet laboratory experiments.
MultiOmic generates new treatment concepts that transform patient well-being in clinical situations lacking effective treatments. Its biopharma partners benefit from gaining access to innovative R&D programs with compelling health economics rationales.
Metabolic syndrome is a complex long-term disease with substantial heterogeneity of patient experience driven by many factors unique to the patient’s underlying endotype. Patient stratification and precision medicine strategies are critical for transforming patient outcomes. And these strategies can only be implemented with large real-world datasets and AI techniques.
Thank you to Becky Cripps and Weihan Wong for their contributions to this article.
- World Health Organization. (2020, December 9). The top 10 causes of death. World Health Organization. Retrieved from https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death
- Estimated by MultiOmic Health by aggregating multiple disease-specific sources
- Bommer, C., Sagalova, V., Heesemann, E., Manne-Goehler, J., Atun, R., Bärnighausen, T., … & Vollmer, S. (2018). Global economic burden of diabetes in adults: projections from 2015 to 2030. Diabetes care, 41(5), 963-970.
- Xie, Y., Xu, E., Bowe, B., & Al-Aly, Z. (2022). Long-term cardiovascular outcomes of COVID-19. Nature medicine, 28(3), 583-590.
- Watson, C. (2022). Diabetes risk rises after COVID, massive study finds. Nature.
- Torjesen, I. (2021). Covid-19: Infection increases the risk of kidney disease even in mild cases, finds study.
- U.S. Department of Health and Human Services. (n.d.). Diabetic kidney disease. National Institute of Diabetes and Digestive and Kidney Diseases. Retrieved August 1, 2022, from https://www.niddk.nih.gov/health-information/diabetes/overview/preventing-problems/diabetic-kidney-disease#:~:text=Diabetic%20kidney%20disease%20is%20a, with%20diabetes %20has%20kidney%20disease.&text=The%20main%20job%20of% 20the,your%20blood%20to%20make%20urine.
- Montero, R. M., Herath, A., Qureshi, A., Esfandiari, E., Pusey, C. D., Frankel, A. H., & Tam, F. W. (2018). Defining Phenotypes in Diabetic Nephropathy: a novel approach using a cross-sectional analysis of a single centre cohort. Scientific reports, 8(1), 1-8.
- Mechanick, J. I., Garber, A. J., Grunberger, G., Handelsman, Y., & Garvey, W. T. (2018). Dysglycemia-based chronic disease: an American Association of Clinical Endocrinologists position statement. Endocrine Practice, 24(11), 995-1011.
- Lindh, M., Banefelt, J., Fox, K. M., Hallberg, S., Tai, M. H., Eriksson, M., … & Qian, Y. (2019). Cardiovascular event rates in a high atherosclerotic cardiovascular disease risk population: estimates from Swedish population-based register data. European Heart Journal-Quality of Care and Clinical Outcomes, 5(3), 225-232.
- Gross, A. (2018, May 4). Column – diseases of the kidney on rise across asia. MedTech Intelligence. Retrieved August 1, 2022, from https://www.medtechintelligence.com/column/diseases-of-the-kidney-on-rise-across-asia/
- Key Statistics – National Kidney Foundation Singapore. (n.d.). Retrieved August 1, 2022, from https://nkfs.org/about-us/key-statistics/
- Annual data report. USRDS. (n.d.). Retrieved August 1, 2022, from https://adr.usrds.org/2021/end-stage-renal-disease/11-international-comparisons
- Norris, K. C., Smoyer, K. E., Rolland, C., Van der Vaart, J., & Grubb, E. B. (2018). Albuminuria, serum creatinine, and estimated glomerular filtration rate as predictors of cardio-renal outcomes in patients with type 2 diabetes mellitus and kidney disease: a systematic literature review. BMC nephrology, 19(1), 1-13.
- Badrick, T., & Turner, P. (2013). The Uncertainty of the eGFR. Indian Journal of Clinical Biochemistry, 28(3), 242-247.
- Branten, A. J., Vervoort, G., & Wetzels, J. F. (2005). Serum creatinine is a poor marker of GFR in nephrotic syndrome. Nephrology Dialysis Transplantation, 20(4), 707-711.
- MultiOmic Health analysis of drugs approved for use in the United Kingdom
- The Singapore family physician. College of Family Physicians Singapore. (n.d.). Retrieved August 1, 2022, from https://www.cfps.org.sg/publications/the-singapore-family-physician/article/313_pdf
About the Authors
Robert Thong is currently CEO of MultiOmic Health, a next-generation precision medicine venture, integrating computational systems biology with laboratory science to develop and commercialize new therapeutics for metabolic syndrome — the world’s largest healthcare burden.
Andrew Wu is Co-Founder and CEO of Mesh Bio, a Singapore-based digital health startup founded in 2018 – aimed at delivering precision clinical intervention to address challenges in patient management and rising chronic diseases in the region.