Glucotypes reveal new patterns of glucose dysregulation

Heather Hall, Dalia Perelman, Alessandra Breschi, Patricia Limcaoco, Ryan Kellogg, Tracey McLaughlin, Michael Snyder, Heather Hall, Dalia Perelman, Alessandra Breschi, Patricia Limcaoco, Ryan Kellogg, Tracey McLaughlin, Michael Snyder

Abstract

Diabetes is an increasing problem worldwide; almost 30 million people, nearly 10% of the population, in the United States are diagnosed with diabetes. Another 84 million are prediabetic, and without intervention, up to 70% of these individuals may progress to type 2 diabetes. Current methods for quantifying blood glucose dysregulation in diabetes and prediabetes are limited by reliance on single-time-point measurements or on average measures of overall glycemia and neglect glucose dynamics. We have used continuous glucose monitoring (CGM) to evaluate the frequency with which individuals demonstrate elevations in postprandial glucose, the types of patterns, and how patterns vary between individuals given an identical nutrient challenge. Measurement of insulin resistance and secretion highlights the fact that the physiology underlying dysglycemia is highly variable between individuals. We developed an analytical framework that can group individuals according to specific patterns of glycemic responses called "glucotypes" that reveal heterogeneity, or subphenotypes, within traditional diagnostic categories of glucose regulation. Importantly, we found that even individuals considered normoglycemic by standard measures exhibit high glucose variability using CGM, with glucose levels reaching prediabetic and diabetic ranges 15% and 2% of the time, respectively. We thus show that glucose dysregulation, as characterized by CGM, is more prevalent and heterogeneous than previously thought and can affect individuals considered normoglycemic by standard measures, and specific patterns of glycemic responses reflect variable underlying physiology. The interindividual variability in glycemic responses to standardized meals also highlights the personal nature of glucose regulation. Through extensive phenotyping, we developed a model for identifying potential mechanisms of personal glucose dysregulation and built a webtool for visualizing a user-uploaded CGM profile and classifying individualized glucose patterns into glucotypes.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1. Summary of methods.
Fig 1. Summary of methods.
We enrolled 57 participants with different diagnoses for diabetes and continuously monitored their interstitial glucose levels for 2–4 weeks. Spectral clustering was used to classify different patterns of glycemic responses based on their variability. We then compared different classes of glucose variability with common clinical parameters and in relation to the effect of meal with standardized nutrient content. Finally, we analyzed insulin metabolism to elucidate potential physiological mechanisms underlying glycemic dysregulation detected through our classification. CGM, continuous glucose monitoring; IR, insulin resistance; IS, insulin sensitivity.
Fig 2. Classification of CGM with classes…
Fig 2. Classification of CGM with classes of glycemic signatures.
(A-C) Segregation of the 2.5-hour windows into the three classes of glycemic signatures derived from spectral clustering. The lines in each panel show an example of the glycemic signatures in each class. This separation of windows explains approximately 73% of the variance. (D) One day of CGM data for 3 separate individuals. Color indicates classification of glycemic signatures. Note that since overlapping windows were used for clustering and classification, some periods of the day have multiple classifications. (E) Heat map showing the fraction of time individuals spent in each of the glycemic classes. Rows represent unique individuals in the cohort, while columns represent each of the glycemic signature classes shown in A-C. Color of the tiles corresponds to the fraction of time spent in each class, with 1 being 100% of the time. There were 238 windows per participant (S3 Data). Rows of individuals are arranged according to hierarchical clustering. CGM, continuous glucose monitoring.
Fig 3. Correlation between glycemic signature classes…
Fig 3. Correlation between glycemic signature classes and measures of glucose homeostasis.
(A) Forest plots for each of the glucotypes. A Pearson’s correlation test was used to determine the correlation between the clinical metabolic tests—listed 1 per line—and the fraction of time spent in each glucotype class (S5 Data). The center dot line is the resulting correlation coefficient, with the line representing the corresponding 95% confidence interval. (B) Forest plot with the lines representing age and BMI. (C) OGTT 2hr is plotted against the fraction of time in the severe glucotype for each individual. The line of best fit is shown in blue with the 95% confidence interval shaded in gray. The correlation coefficient, r, was derived from a Pearson’s correlation test. OGTT 2hr, blood glucose concentration 2 hours after the start of oral glucose tolerance test; SSPG, steady-state plasma glucose.
Fig 4. Correlation between carbohydrate content and…
Fig 4. Correlation between carbohydrate content and frequency of severe glycemic signature responses.
(A) Heat map of glycemic signature class responses to standardized meals. Rows represent individuals, and columns represent meals. Color indicates classification of response, and intensity indicates fraction of total responses (S7 Data). The number of total responses per individual corresponds to the number of standard meals each participant ate. Individuals are sorted by hierarchical clustering (left) based on the fraction of responses in a given class, regardless of the type of meal that triggered the response. (B) Contingency table reporting the number of responses to standardized meals assigned to a given glycemic signature class. Meals are sorted based on their net carbohydrate content (total carbohydrates–fiber). The size of the dots is proportional to the number of windows, and the intensity is the chi-squared ratio between observed and expected counts. There is a significant association between net carbohydrate content and severity of the response (p-value: 0.06, chi-squared test). “PB”: bread and peanut butter; “Bar”: PROBAR protein bar; “CF”: cornflakes and milk. For both panels A and B, the classification of glycemic responses to meals is based on the entire CGM profiles rather than on the initial set of clustered windows (see Methods). CGM, continuous glucose monitoring.
Fig 5. Diabetes classification and prevalence of…
Fig 5. Diabetes classification and prevalence of severe variability.
(A) Principal component (labeled “PC”) analysis of common measures used to describe glucose control and evaluate CGM data (S5 Data). Individual participants are colored based on their glucotype, the glycemic signature class in which they spent the majority of their time. The dot size is proportional to the fraction of time spent with severe variability. (B) Same principle component analysis, but participants are colored based on diabetes diagnosis. This diagnosis was based on the ADA Guidelines of HbA1c, fasting blood glucose, and blood glucose concentration at 2 hours after the start of an OGTT. (C) Box and whisker plot of fraction of time spent with severe variability for nondiabetic, prediabetic, and diabetic individuals. The classification is based on all time windows, not restricting to the ones after standardized meals. (D) Proportion of CGM data in prediabetic and diabetic glycemic ranges defined in the ADA Guidelines. Participants are grouped by their diabetes diagnosis and colored by their glucotype. Some normoglycemic participants who demonstrated the severe glucotype reached prediabetic glycemic levels up to 15% of the time and diabetic glycemic levels 2% of the time. ADA, American Diabetes Association; CGM, continuous glucose monitoring; OGTT, oral glucose tolerance test.
Fig 6. Glucotypes and insulin metabolism.
Fig 6. Glucotypes and insulin metabolism.
(A) Comparison of glycemic response, insulin secretion, and insulin sensitivity across different glucotypes (low, moderate, severe) (S5 Data). Each column is a participant. The glycemic response is shown in the horizontal panels as fasting blood sugar concentration and as blood glucose concentration 2 hours after oral glucose load in an OGTT. Dashed lines for fasting blood sugar and 2-hour glucose in OGTT correspond to the ADA thresholds for prediabetes and diabetes. Insulin secretion rate was calculated using deconvolution of C-peptide concentrations at 0, 30, and 120 minutes during OGTT (ISEC software [25], see Methods). The average derivative of the insulin secretion curve is shown as aggregate measure of insulin secretion rate over the whole OGTT time course. Insulin sensitivity is shown as glucose concentration from SSPG test, with higher values reflecting greater insulin resistance. (B, C, D, E, F) Insulin secretion rate (black solid line) and blood glucose concentration (dots and purple solid line) during OGTT for 5 individuals (S8 Data). Shaded areas in panels B-F aggregate data of distribution of insulin secretion rate for participants diagnosed with diabetes (pink), prediabetes (green), and nondiabetic (blue). Each panel represents an individual with a different physiology. (B) Insulin-sensitive individual with normal insulin secretion and a low blood glucose 2 hours postprandially; (C) Individual diagnosed with diabetes demonstrating high glycemic concentrations despite high insulin secretion; (D) Nondiabetic individual with normal fasting blood glucose but high 2-hour OGTT value in setting of insulin resistance and absolute but relative deficiency of insulin secretion; (E) Nondiabetic individual with insulin resistance with low insulin secretion, characterized by early glucose rise after load; (F) Nondiabetic individual with insulin resistance and high compensatory insulin secretion with relatively normal postprandial glucose following oral glucose load. ADA, American Diabetes Association; ISEC, Insulin SECretion; OGTT, oral glucose tolerance test; SSPG, steady-state plasma glucose.

References

    1. American Diabetes Association. American Diabetes Association Standards of Medical Care in Diabetes—2017. Diabetes Care—The Journal of Clinical and Applied Research and Education. 2016;40 (Supplement 1) p. 1–142.
    1. Tabá G, Herder C, Rathmann W, Brunner E, Kivimäk M. Prediabetes: A High-Risk State for Developing Diabetes. The Lancet. 2012;379(9833):2279–2290.
    1. McCulloch DK. Overview of medical care in adults with diabetes mellitus. UpToDate Online Boston. 2009. Available from:
    1. Ahlqvist E, Storm P, Käräjämäki A, Martinell M, Dorkhan M, Carlsson A, et al. Novel subgroups of adult-onset diabetes and their association with outcomes: a data-driven cluster analysis of six variables. The Lancet Diabetes and Endocrinology. 2018; 6(5):361–369 10.1016/S2213-8587(18)30051-2
    1. Holst JJ, Knop FK, Vilsbøll T, Krarup T, Madsbad S. Loss of incretin effect is a specific, important, and early characteristic of type 2 diabetes. Diabetes Care. 2011;34(Supplement 2):S251–S257.
    1. Nathan DM, Davidson MB, DeFronzo RA, Heine RJ, Henry RR, Pratley R, et al. Impaired Fasting Glucose and Impaired Glucose Tolerance. Diabetes Care. 2007;30(3):753–759. 10.2337/dc07-9920
    1. Group DPPR. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. New England journal of medicine. 2002;346(6):393–403. 10.1056/NEJMoa012512
    1. Tuomilehto J, Lindström J, Eriksson JG, Valle TT, Hämäläinen H, Ilanne-Parikka P, et al. Prevention of Type 2 Diabetes Mellitus by Changes in Lifestyle among Subjects with Impaired Glucose Tolerance. New England Journal of Medicine 2001;344(18):1343–1350. 10.1056/NEJM200105033441801
    1. Li G, Zhang P, Wang J, Gregg EW, Yang W, Gong Q, et al. The long-term effect of lifestyle interventions to prevent diabetes in the China Da Qing Diabetes Prevention Study: a 20-year follow-up study. The Lancet. 2008;371(9626):1783–1789.
    1. Herman WH, Hoerger TJ, Brandle M, Hicks K, Sorensen S, Zhang P, et al. The cost-effectiveness of lifestyle modification or metformin in preventing type 2 diabetes in adults with impaired glucose tolerance. Annals of internal medicine. 2005;142(5):323–332.
    1. Zeevi D, Korem T, Zmora N, Israeli D, Rothschild D, Weinberger A, et al. Personalized Nutrition by Prediction of Glycemic Responses. Cell. 2015;163(5):1079–1094. 10.1016/j.cell.2015.11.001
    1. Borg R, Kuenen J, Carstensen B, Zheng H, Nathan D, Heine R, et al. Real-life glycaemic profiles in non-diabetic individuals with low fasting glucose and normal HbA1c: the A1C-Derived Average Glucose (ADAG) study. Diabetologia. 2010;53(8):1608–1611. 10.1007/s00125-010-1741-9
    1. Garber AJ. Postprandial dysmetabolism and the heart. Heart failure clinics. 2012;8(4):563–573. 10.1016/j.hfc.2012.06.004
    1. Lind M, Tuomilehto J, Uusitupa M, Nerman O, Eriksson J, Ilanne-Parikka P, et al. The association between HbA1c, fasting glucose, 1-hour glucose and 2-hour glucose during an oral glucose tolerance test and cardiovascular disease in individuals with elevated risk for diabetes. PLoS ONE. 2014;9(10):e109506 10.1371/journal.pone.0109506
    1. Bonora E. Postprandial peaks as a risk factor for cardiovascular disease: epidemiological perspectives. International journal of clinical practice Supplement. 2002;(129):5–11.
    1. Jung HS. Clinical Implications of Glucose Variability: Chronic Complications of Diabetes. Endocrinology and Metabolism. 2015;30(2):167 10.3803/EnM.2015.30.2.167
    1. Monnier L, Mas E, Ginet C, Michel F, Villon L, Cristol JP, et al. Activation of oxidative stress by acute glucose fluctuations compared with sustained chronic hyperglycemia in patients with type 2 diabetes. JAMA. 2006;295(14):1681–1687. 10.1001/jama.295.14.1681
    1. Crenier L, Abou-Elias C, Corvilain B. Glucose Variability Assessed by Low Blood Glucose Index Is Predictive of Hypoglycemic Events in Patients With Type 1 Diabetes Switched to Pump Therapy. Diabetes Care. 2013;36(8):2148–2153. 10.2337/dc12-2058
    1. Monnier L, Colette C, Leiter L, Ceriello A, Hanefeld M, Owens D, et al. The effect of glucose variability on the risk of microvascular complications in type 1 diabetes. Diabetes Care. 2007;30(1):185–6– author reply 187–8.
    1. Suh S, Kim JH. Glycemic Variability: How Do We Measure It and Why Is It Important? Diabetes & metabolism journal. 2015;39(4):273.
    1. Albert A, Rajagopal R. Smart Meter Driven Segmentation: What Your Consumption Says About You. IEEE Transactions on Power Systems. 2013;28(4):4019–4030.
    1. Tuzcu V, Nas S. Dynamic time warping as a novel tool in pattern recognition of ECG changes in heart rhythm disturbances. IEEE International Conference on Systems, Man and Cybernetics, 2005, pp. 182–186 Vol. 1 10.1109/ICSMC.2005.1571142
    1. Mazze RS, Strock E, Wesley D, Borgman S, Morgan B, Bergenstal R, et al. Characterizing glucose exposure for individuals with normal glucose tolerance using continuous glucose monitoring and ambulatory glucose profile analysis. Diabetes technology & therapeutics. 2008;10(3):149–159.
    1. Nomura K, Saitoh T, Kim GU, Yamanouchi T. Glycemic profiles of healthy individuals with low fasting plasma glucose and HbA1c. ISRN endocrinology. 2011;2011(435047):6 10.5402/2011/435047.
    1. Hovorka R, Soons PA, Young MA. ISEC: a program to calculate insulin secretion. Computer methods and programs in biomedicine. 1996;50(3):253–264.
    1. Chang W, Cheng J, Allaire J, Xie Y, McPherson J. shiny: Web Application Framework for R; 2018 [cited 2018 June 28]. Available from: .
    1. Pei D, Jones CN, Bhargava R, Chen YD, Reaven G. Evaluation of octreotide to assess insulin-mediated glucose disposal by the insulin suppression test. Diabetologia. 1994;37(8):843–845.
    1. Greenfield M, Doberne L, Kraemer F, Tobey T, Reaven G. Assessment of insulin resistance with the insulin suppression test and the euglycemic clamp. Diabetes. 1981;30(5):387–392.
    1. Van Cauter E, Mestrez F, Sturis J, Polonsky KS. Estimation of insulin secretion rates from C-peptide levels: comparison of individual and standard kinetic parameters for C-peptide clearance. Diabetes. 1992;41(3):368–377.
    1. Zhao et al. Systematic Clustering of Transcription Start Site Landscapes. PLoS ONE. 2011;6(8): e23409 10.1371/journal.pone.0023409
    1. Calinski T, Harabasz J. A dendrite method for cluster analysis. Communications in Statistics. 1974;3(1):1–27.
    1. Halkidi M, Batistakis Y, Vazirgiannis M. On clustering validation techniques. Journal of intelligent information systems. 2001;17(2–3):107–145.
    1. Hennig C. fpc: Flexible Procedures for Clustering; 2018 [cited 2018 June 28]. Available from: .
    1. Batista GE, Wang X, Keogh EJ. A complexity-invariant distance measure for time series In: Proceedings of the 2011 SIAM international conference on data mining. April 28–30, 2011. Arizona, USA: SIAM; p. 699–710. 10.1137/1.9781611972818.60
    1. Giorgino T, et al. Computing and visualizing dynamic time warping alignments in R: the dtw package. Journal of statistical Software. 2009;31(7):1–24.
    1. Von Luxburg U. A tutorial on spectral clustering. Statistics and Computing. 2007;17(4).
    1. David Meyer and Christian Buchta (2018). proxy: Distance and Similarity Measures. R package version 0.4–22. Available from:
    1. Bersch SD, Azzi D, Khusainov R, Achumba IE, Ries J. Sensor data acquisition and processing parameters for human activity classification. Sensors (Basel, Switzerland). 2014;14(3):4239–4270.
    1. Wang B, Mezlini AM, Demir F, Fiume M, Tu Z, Brudno M, et al. Similarity network fusion for aggregating data types on a genomic scale. Nature methods. 2014;11(3):333 10.1038/nmeth.2810
    1. Wojcicki JM. “J”-index. A new proposition of the assessment of current glucose control in diabetic patients. Hormone and metabolic research. 1995;27(1):41–42.
    1. Service FJ, Molnar GD, Rosevear JW, Ackerman E, Gatewood LC, Taylor WF. Mean amplitude of glycemic excursions, a measure of diabetic instability. Diabetes. 1970;19(9):644–655.
    1. Baghurst PA. Calculating the mean amplitude of glycemic excursion from continuous glucose monitoring data: an automated algorithm. Diabetes technology & therapeutics. 2011;13(3):296–302.

Source: PubMed

3