Identifying subgroups of people at risk for type 2 diabetes

Analysis of glycemic and non-glycemic characteristics in individuals at risk for type 2 diabetes identifies six putative subgroups that differ in terms of risk of diabetes progression and complications.

Type 2 diabetes (T2D) impacts over four hundred million people globally, causing enormous health and economic burdens, in large part due to complications of the disease including nephropathy, retinopathy, and cardiovascular disease1. An even larger group of people have risk factors for developing diabetes in the future2 and represent a growing population in need of clearer diagnostic criteria and approaches for early intervention. Identifying people with an elevated risk of developing T2D has practical implications, as there are interventions that can prevent or delay T2D onset3, and earlier disease diagnosis is thought to improve outcomes4. Current screening strategies cast a wide net, such that under some criteria up to 35% of adults in a given population may be considered to have prediabetes2, with only a small fraction (~5–10%) expected to progress to T2D each year3. Additionally, the pathophysiology underlying development of diabetes remains poorly understood, and delineating these mechanisms will likely improve strategies to classify and manage at-risk individuals. In this issue of Nature Medicine, Wagner et al.5 are able to stratify individuals at risk for T2D using multiple phenotypic measures, both glycemic and non-glycemic, to identify subgroups of people differing in likelihood of developing T2D and associated complications (Fig. 1).Fig. 1: Identifying subgroups at risk from type 2 diabetes.Wagner et al. performed cluster analysis on individuals at risk for T2D from two prospective studies (discovery cohort TUEF/TULP, N = 899; replication Whitehall II, N = 6,810). In the discovery cohort, they included in their model estimates of insulin secretion and insulin resistance from an oral glucose tolerance test, MRI-based measurement of visceral and subcutaneous adipose tissue, 1H-MR-spectroscopy-based measurement of hepatic fat content, HDL cholesterol, and polygenic T2D risk score. In the replication cohort a related set of available simpler phenotypes were used: estimates of insulin secretion and insulin resistance from an oral glucose tolerance test, fasting triglycerides, HDL cholesterol, waist circumference, hip circumference, and BMI. The authors performed ‘partitioning around medoids’ (PAM) clustering and identified six subgroups of individuals in both analyses. These six clusters displayed distinct risks of progression to T2D, as well as risks of renal and cardiovascular outcomes.In current clinical practice, international guidelines recommend identifying individuals who have a high probability of developing T2D based on risk factors (such as obesity, family history of diabetes) and then implementing screening with glycemic blood tests, which are sometimes preceded by a screening questionnaire6,7. Those who have glycemic measures that are elevated, but in an intermediate range that fall below thresholds for T2D diagnosis, have been categorized as having ‘prediabetes’ or ‘intermediate hyperglycemia’. Identifying individuals at risk for T2D is useful for preventing or delaying the onset of disease3 and for earlier management leading to improved outcomes4; however, different diagnostic criteria for prediabetes have been proposed7,8 that all show considerable heterogeneity in terms of the ability to predict T2D progression in an individual3. In fact, up to 59% of people diagnosed with prediabetes may spontaneously return to normoglycemia within five years of follow-up9. Several randomized controlled trials have demonstrated that lifestyle and/or pharmacological interventions aimed at those at risk for T2D improve outcomes (reviewed in ref. 10); however, interventions may not be necessary or cost-effective for all people with prediabetes11. Identifying subsets of individuals at highest risk of adverse outcomes will help to better target these interventions to those who need them most.Wagner et al. analyzed data from participants in two prospective studies, including only those who did not have T2D initially. The discovery cohort, the Tuebingen Family study and Tuebingen Lifestyle Program (TUEF/TULIP), included German adults with a history of prediabetes, a family history of diabetes, a body mass index (BMI) greater than 27 kg/m2, or a history of gestational diabetes. The authors studied the deep phenotyping data of 899 participants from TUEF/TULIP, including estimates of insulin secretion and insulin resistance from an oral glucose tolerance test, magnetic resonance imaging (MRI)-based measurement of visceral and subcutaneous adipose tissue, 1H-MR-spectroscopy-based measurement of hepatic fat content, high-density lipoprotein (HDL) cholesterol, and polygenic T2D risk score. The authors carried out partitioning around medoids (PAM) clustering on these data and identified six subgroups of individuals. The replication cohort, Whitehall II, was a study of London-based civil servants aged 33–55 years. Six corresponding clusters were also identified in the 6,810 Whitehall II participants using a related set of available simpler phenotypes: estimates of insulin secretion and insulin resistance from an oral glucose tolerance test, fasting triglycerides, HDL cholesterol, waist circumference, hip circumference, and BMI.By carrying out longitudinal analysis of individuals in the clusters, the authors assessed development of T2D as well as renal and cardiovascular disease outcomes over a mean of 4.1 years in the discovery and 16.3 years in the replication cohort; all-cause mortality was also assessed in the replication cohort. Two clusters (3 and 5) were notable for having the highest risk of T2D progression in both studies. These two clusters had the highest proportion of individuals with abnormal glycemia at baseline; however, the authors found that cumulative diabetes risk in these two subgroups was still higher than glucose-only stratification approaches. A different cluster (6) also had increased glycemia at baseline, but lower risk of T2D progression than clusters 3 and 5, yet increased risk of renal disease as well as all-cause mortality; the authors hypothesize that the constellation of higher risk of kidney disease at relatively lower glycemic risk is related to an observed increase in renal sinus fat volume and reduced burden of genetic variants connected physiologically to pancreatic β-cell function in the individuals in this cluster12 compared to those in the other clusters.In addition to providing a tool for risk stratification, these clusters point to differing underlying pathophysiology in each subgroup of individuals and hence offer exciting opportunities for gaining mechanistic insight into the pathogenesis of T2D and its complications. What physiologic process, for example, is promoting renal decline in cluster 6 that appears to be unlinked to diabetes risk? The different cluster profiles also raise the interesting possibility that there may be interventions best suited for those in particular high-risk clusters. Can we identify differences in disease biology among the clusters that may guide novel therapeutic strategies? Equally important, the lower risk clusters may represent sets of individuals for whom recommended measures could be safely relaxed, for example, by reducing monitoring intervals and therefore avoid unnecessary intervention and promote more cost-effective use of resources.Machine-learning approaches, as used here by Wagner et al., for phenotypically driven subtyping of complex conditions have gained great attention in their recent application to diabetes, whereby subgroups were identified that differed in time to insulin use and risk of complications13. It is worth noting that while such clustering approaches have conceptual appeal in discretely subclassifying patients, it is also possible that outcomes (such as risk of T2D or nephropathy) could be more precisely predicted at the individual level by carrying out regression modeling of each outcome separately using the multiple phenotypic variables, rather than categorizing individuals into discrete subgroups. This approach of using simple clinical measures in a regression model for individuals with T2D was recently shown to outperform clustering for prediction of nephropathy risk and response to treatment14. Future research will benefit from further investigation and comparison of both discrete clustering and predictive regression modeling approaches.As the prevalence of both prediabetes and T2D continues to rise across the globe1, the ability to identify those at highest risk and intervene to curb this epidemic is direly needed. The analysis by Wagner et al. serves as a proof of concept that substructure exists within individuals at risk for T2D and that subgroups with distinct risk profiles can be reproducibly identified. With additional replication in future studies, these subgroups of people at risk for T2D offer the exciting potential for improved screening practices and targeted prevention strategies.

References1.Federation, I.D. IDF Diabetes Atlas, 9th edn, (Brussels, Belgium, 2019).2.National Diabetes Statistics Report, 2020. (Center for Disease Control and Prevention, U.S. Department of Health and Human Services, Atlanta, GA, 2020).3.Tabák, A. G., Herder, C., Rathmann, W., Brunner, E. J. & Kivimäki, M. Lancet 379, 2279–2290 (2012).Article 

Google Scholar 
4.Colagiuri, S., Cull, C. A. & Holman, R. R., UKPDS Group. Diabetes Care 25, 1410–1417 (2002).Article 

Google Scholar 
5.Wagner, R. et al. Nat. Med. https://doi.org/10.1038/s41591-020-1116-9 (2020)6.International Diabetes Federation. www.idf.org/managing-type2-diabetes (2017).7.American Diabetes Association. Diabetes Care 43, S14–S31 (2020).Article 

Google Scholar 
8.Global report on diabetes. (World Health Organization, Geneva, 2016).9.Richter, B., Hemmingsen, B., Metzendorf, M. I. & Takwoingi, Y. Cochrane Database Syst. Rev. 10, CD012661 (2018).PubMed 

Google Scholar 
10.Crandall, J.P. et al. Nat. Clin. Pract. Endocrinol. Metab. 4, 382–393 (2008).Article 

Google Scholar 
11.Diabetes Prevention Program Research Group. Diabetes Care 26, 2518–2523 (2003).Article 

Google Scholar 
12.Udler, M. S. et al. PLoS Med. 15, e1002654 (2018).Article 

Google Scholar 
13.Ahlqvist, E. et al. Lancet Diabetes Endocrinol. 6, 361–369 (2018).Article 

Google Scholar 
14.Dennis, J. M., Shields, B. M., Henley, W. E., Jones, A. G. & Hattersley, A. T. Lancet Diabetes Endocrinol. 7, 442–451 (2019).Article 

Google Scholar 
Download referencesAuthor informationAffiliationsDiabetes Unit and Center for Genomic Medicine, Department of Medicine, Massachusetts General Hospital, Boston, MA, USAMiriam S. UdlerCorresponding authorCorrespondence to
Miriam S. Udler.Ethics declarations

Competing interests
The author declares no competing interests.

About this articleCite this articleUdler, M.S. Identifying subgroups of people at risk for type 2 diabetes.
Nat Med 27, 23–25 (2021). https://doi.org/10.1038/s41591-020-01208-2Download citation

Via Source link

Most Popular


All About information only. Please consult your Doctor for any illness.

Copyright © 2020 Safety Health News. Powered by Wordpress.

To Top