https://orcid.org/0000-0003-4719-2268
Figures
Abstract
The artificial pancreas is a closed-loop insulin delivery system that automatically regulates glucose levels in individuals with type 1 diabetes. In-silico testing using simulation environments accelerates the development of better artificial pancreas systems. Simulation environments need an accurate model that captures glucose dynamics during exercise to simulate real-life scenarios. We proposed six variations of the Bergman Minimal Model to capture the physiological effects of moderate exercise on glucose dynamics in individuals with type 1 diabetes. We estimated the parameters of each model with clinical data using a Bayesian approach and Markov chain Monte Carlo methods. The data consisted of measurements of plasma glucose, plasma insulin, and oxygen consumption collected from a study of 17 adults with type 1 diabetes undergoing aerobic exercise sessions. We compared the models based on the physiological plausibility of their parameters estimates and the deviance information criterion. The best model features (i) an increase in glucose effectiveness proportional to exercise intensity, and (ii) an increase in insulin action proportional to exercise intensity and duration. We validated the selected model by reproducing results from two previous clinical studies. The selected model accurately simulates the physiological effects of moderate exercise on glucose dynamics in individuals with type 1 diabetes. This work offers an important tool to develop strategies for exercise management with the artificial pancreas.
Citation: Alkhateeb H, El Fathi A, Ghanbari M, Haidar A (2021) Modelling glucose dynamics during moderate exercise in individuals with type 1 diabetes. PLoS ONE 16(3): e0248280. https://doi.org/10.1371/journal.pone.0248280
Editor: Pasquale Palumbo, University of Milano Bicocca, ITALY
Received: September 25, 2020; Accepted: February 24, 2021; Published: March 26, 2021
Copyright: © 2021 Alkhateeb et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data cannot be shared publicly because of legal restrictions. These restrictions are imposed by the IRCM local ethics committee. Data are available from the Institut de recherches cliniques de Montréal (contact via info@ircm.qc.ca) for researchers who meet the criteria for access to confidential data.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Type 1 diabetes (T1D) is a chronic disease in which the immune system attacks and destroys the pancreatic beta cells resulting in the loss of insulin secretion [1]. Individuals with T1D need life-long insulin therapy, which requires either multiple daily injections or the use of an insulin pump that infuses insulin subcutaneously, guided by glucose measurements. A glucose sensor is a wearable device that continuously monitors blood glucose levels [2]. However, despite current technologies, individuals with T1D still spend significant amount of time outside the glucose target range, placing them at risk for devastating long-term complications such as heart attack, stroke, blindness, kidney disease, and amputation [3].
The artificial pancreas system is a promising solution for individuals with T1D. It is a closed-loop delivery system composed of a glucose sensor, an infusion pump, and a control algorithm. The artificial pancreas can be a single-hormone system that infuses insulin or a dual-hormone system that infuses insulin and glucagon [2]. The control algorithm automatically computes the hormonal doses based on glucose levels. Randomized clinical trials have shown that the artificial pancreas improves glucose control and reduces hypoglycemia [4].
Regular exercise is recommended by the American Diabetes Association [5] for its numerous beneficial effects [6], yet, around 63% of individuals with T1D are inactive [7], with the fear of hypoglycemia being the greatest barrier to exercise [8]. Even with the use of the artificial pancreas, exercise management remains a challenge [9], and new exercise control algorithms are needed to reduce the occurrence of exercise-induced hypoglycemia [10].
Enhancing the control algorithms can be accelerated by performing in-silico testing in simulation environments [2] before moving into clinical trials. Simulation environments contain virtual patients, each represented by a set of parameters used to simulate glucoregulatory models [11–14]. In order to test the safety, efficacy, and limitations of the control algorithm before moving into clinical trials, the glucoregulatory models need to incorporate the exercise effect on glucose dynamics.
A few models are reported that describe the exercise effect on glucose dynamics. In particular, Roy and Parker [15] extended the Bergman Minimal Model [16] to incorporate the effects of exercise on glucose levels. Parameter estimates in their exercise model were based on data from healthy individuals performing different exercise protocols, but the model was validated by reproducing results from a previous exercise study performed by individuals with T1D. Ewings et al. [17] modified Roy and Parker’s model, and estimated model parameters using data from individuals with T1D. Breton [18] proposed another extension of the Minimal Model using heart rate as a marker of energy expenditure. Model parameters were estimated using data from individuals with T1D undergoing exercise during hyperinsulinemic clamp protocol. Dalla Man et. al [19] incorporated an extension of Breton’s exercise model into a simulation environment of the glucose–insulin system [20]. They proposed three candidate models and performed in-silico testing to compare and validate the models. Their selected model accounts for exercise duration and intensity. A recent model by Schneider [21] has been developed by expanding the Bergman Minimal Model [16] to capture the effects of aerobic and anaerobic exercise on glucose dynamics in individuals with T1D. The model uses heart rate to distinguish between aerobic and anaerobic exercise. In-silico experiments showed that the extension model captures the effects stated within the literature. Kim et al. [22] developed a model to predict the hormonal changes during exercise. Their model was validated with experimental data where non-diabetic participants performed moderate intensity exercise for 60 min. Palumbo et al. [23] introduced some modifications to Kim’s model to simulate the effects on personal metabolic homeostasis during exercise of different durations and intensities. Hernandez et al. [24] proposed another exercise model using the percentage of maximum oxygen consumption and the percentage active muscle mass as inputs to calculate the change in hepatic glucose production, peripheral glucose uptake, and peripheral insulin uptake. Resalat et al. [25] incorporated Hernandez’s model into a model of single- and dual-hormone virtual patient population. Other exercise models have been proposed [26–29]. In all previous studies, the models were either proposed as stand-alone models and not compared to other candidate models or were not developed from real T1D clinical data.
In this paper, we proposed six physiologically-motivated variations of the Bergman Minimal Model [16] to capture the effects of moderate exercise on glucose dynamics in individuals with T1D. We estimated the parameters of the models using clinical data from individuals with T1D utilizing a Bayesian approach with Markov chain Monte Carlo methods. We compared the models based on the physiological plausibility of their parameters estimates and the deviance information criterion (DIC). We validated the performance of the selected model by replicating two published clinical studies.
Methods
A. Experimental data
Model parameters were estimated using data collected by Taleb et al [30]. The study included 17 adults with T1D who underwent four interventions. The study compared the efficacy of the dual-hormone (insulin and glucagon) to the single-hormone (insulin only) artificial pancreas during two types of exercise: continuous and interval. Continuous exercise was set at 60% VO2max and lasted for 60 minutes. Interval exercise consisted of 2-minute alternating periods of 50% and 85% VO2max for 40 minutes, with two 10-minute periods at 45% VO2max at the start and end of the sessions. VO2max values were obtained at the admission visit using a graded exercise test [31]. The study was approved by the IRCM Ethics Committee and conducted according to the declaration of Helsinki. All participants provided written informed consent.
At each intervention visit, the artificial pancreas control began at 15:30, a snack was given at 15:45, and exercise started at 18:00 followed by a 30-minute recovery period. Venous blood samples were withdrawn every 30 minutes before exercise, every 10 minutes during exercise, and every 15 minutes after exercise, totaling 14 data points for plasma glucose, plasma insulin, and plasma glucagon each per intervention. The rate of O2 consumption and CO2 production were determined from expired air samples collected using a mask during exercise. Fig 1 shows a summary of the clinical study.
In this work, only data from single-hormone with moderate exercise interventions were used for our modelling purposes. Participant data that contained episodes of hypoglycemia requiring treatment during or before exercise (defined as plasma glucose <3.3 mmol/l with symptoms or <3.0 mmol/l irrespective of symptoms) were excluded and not used in the modelling. The final data set that was used for parameter estimation included data from 11 participants. Fig 2 shows mean glucose and insulin levels for single-hormone with moderate exercise data for all 11 participants used in this work. This work was approved by the McGill Ethics Committee.
Shaded area is exercise period. Values are mean and standard deviation (n = 11).
B. Model structure
The proposed models are different variations of the Bergman Minimal Model [16]. The suggested models were motivated by physiological findings and developed as sets of differential equations able to describe important physiological interactions during exercise. More specifically, the effects of glucose and insulin levels, exercise intensity, and exercise duration on glucose levels were captured. We propose six variations from the following equations structure:
(1)
(2)
where
and
Inc1(t) and inc2(t) define the six variations, as described in Table 1, and model parameters are defined in Table 2.
Models 1–3.
Glucose enters the muscle cells via facilitated diffusion through the GLUT4 transporters [32, 33]. Exercise increases the number of GLUT4 transporters on the cell membrane by increasing the transcription in muscle cells and increasing the translocation to the cell membrane [34, 35]. This leads to an increase in the rate of glucose uptake by two separate mechanisms. First, a rapid short-term increase in insulin-independent glucose uptake, which has been associated with muscle contractions [36, 37]. Second, a broader, longer-lasting increase in insulin-dependent glucose uptake due to the increase sensitivity to the actions of insulin in muscle cells [38, 39]. The combination of these two mechanisms on the rate of glucose uptake is additive [39–43].
Model 1 simulates the rise in the rate of glucose uptake by adding a fixed increase in insulin sensitivity and glucose effectiveness at exercise onset. Model 2 assumes that the increase in glucose effectiveness is negligible. Model 3 assumes that the increase in insulin sensitivity is negligible.
Model 4.
As the intensity of the exercise increases, the rate of hepatic glycogenolysis decreases while the rate of glucose uptake increases [44–48]. Model 4 mimics this phenomenon by making the increase in glucose effectiveness and insulin sensitivity rise with increasing exercise intensity. The percentage of the maximal oxygen consumption (%VO2max) is used to quantify exercise intensity.
Model 5.
As the duration of the exercise increases, the rate of glucose uptake increases, which results in lower glucose levels [48]. Model 5 relates the increase in glucose effectiveness and insulin sensitivity to how long the individual has been exercising.
C. Parameter estimation and model selection
A Bayesian approach was adopted for the estimation process with the use of Markov chain Monte Carlo (MCMC) methodology [49]. A combination of prior knowledge and subject data was used to generate samples from the parameter’s posterior distributions. The median of the posterior distribution was used as a point estimate of parameter estimates.
The differential equations describing the models were solved numerically using initial conditions, model parameters, and input data from plasma insulin and plasma glucagon levels. The MCMC was implemented using WinBUGS version 1.4 [50, 51], extended by WBDiff package version 1.9.4 (MRC Biostatistics Unit, Cambridge, U.K.). The measurement errors associated with plasma glucose were assumed to be normally distributed with zero mean and coefficient of variation (CV) of 2% [52].
DIC was used to compare models based on their goodness of fit and complexity. The model with lowest DIC makes the best predictions and is considered the best performing model [53]. DIC is defined as DIC = ‘goodness of fit’ + ‘complexity’. The goodness of fit is captured via the deviance:
(3)
Complexity is measured via the estimation of the ‘effective number of parameters’:
(4)
where Eθ|y[D] is the expected value of D(θ) given y (i.e., the posterior mean deviance) and D(Eθ|y[θ]) is the deviance evaluated at the posterior mean of the parameters. The DIC is then formally defined as:
(5)
The best model was chosen based on: (i) the DIC score, (ii) parameter’s identifiability and physiological plausibility, and (iii) a CV below 100%. Parameter values were used to generate virtual patients and conduct in-silico experiments.
D. Details of model validation
For the purpose of model validation, we performed in-silico simulations to reproduce results of two previously completed clinical studies [44, 54]. We generated a virtual population of 12 individuals with T1D from the parameter sets that were estimated. Meal and insulin models used for the validation are reported in the S1 File.
Study 1.
The protocol of the first simulated study reflected a clinical study conducted by Rabasa-Lhoret et al. [44] in 8 individuals with T1D. In this study, postprandial exercise of different intensities and durations took place 90 minutes after a breakfast of 75 g carbohydrates with various premeal insulin bolus reductions. Eight protocols were simulated as shown in Table 3.
Study 2.
The second simulated study replicated a clinical study conducted by Zaharieva et al. [54], where 17 individuals with T1D had three basal rate reductions followed by 60-min exercise sessions of 50% VO2max (the exercise was divided into four 15-min bouts with 5-min rest periods in between). The insulin reductions included: 1) pump stop at exercise onset, 2) an 80% basal reduction set 90 min pre-exercise, and 3) a 50% basal reduction set 90 min pre-exercise.
Validation using the Hovorka model.
We also performed in-silico simulation using the Hovorka model [55] to reproduce results of a published clinical study by Haidar et al. [56]. A virtual population of 30 individuals were generated. Model parameters were sampled from a prior log-normal distribution with a mean taken from Wilinska et al. [12] and parameter correlations from healthy individual data [55]. Exercise effects were implemented using the parameter sets that were estimated in this paper. The equations are listed in the appendix (section D).
Results
A. Parameter estimation and model comparison
For each subject, 100,000 iterations of WinBUGS were run. The last 20,000 samples were used to generate the posterior distributions. Table 4 summarizes the results for parameter estimates. All model parameters were physiologically plausible and posteriorly identifiable, with CV not exceeding 50% in all cases.
Model 1, having a fixed increase in both glucose effectiveness and insulin sensitivity, showed the DIC of 471. By neglecting the increase in glucose effectiveness, Model 2 produced a higher DIC of 731. Model 3 produced a better DIC compared with Model 2 by neglecting the increase in insulin sensitivity (731 vs 705). Model 4, incorporating the effect of exercise intensity on rate of glucose uptake, reduced DIC to 450. Model 5, incorporating the effect of exercise duration on rate of glucose uptake, did not reduce DIC compared with Model 4; its DIC was the highest with value 750 (i.e., worst performance). Combining the effect of both intensity and duration in Model 6 produced the lowest DIC of 425 (i.e., best performance). Fig 3 shows model fits on two sample occasions. Model fits for all participants can be found in the S1 Fig in S1 File.
Shaded area is exercise period.
As a result, Model 6’s performance suggests that during exercise there is an increase in glucose effectiveness based on exercise intensity, and an increase in insulin sensitivity based on exercise intensity and duration. Additionally, all of the parameters of Model 6 were identified with good sensitivity (CV < %30). Visual inspection of Model 6’s fits confirms good performance of the model to fit the data. The weighted residuals of Model 6 are shown in Fig 4; weighted residuals indicate the ability of the model to represent the input-output relationship of the subjects without bias.
Values are median (interquartile range) (n = 11).
B. Model validation
The comparison between our simulation results and the results of the published clinical study of Rabasa-Lhoret et al. [44] demonstrates comparable trends in glucose levels before, during, and 60-minutes post exercise. Fig 5 shows a 50% and a 75% reduction in breakfast bolus followed by a 60-minute exercise session at 50% VO2max. Similar to the clinical study’s conclusion, a 75% reduction in premeal bolus resulted in a safer glycemic profile with a decreased risk of hypoglycemia compared with a 50% reduction in premeal bolus. The remaining graphs can be found in the S2–S5 Figs in S1 File.
Shaded area is exercise. Values are mean and standard deviation (n = 11). Error bars represent the standard error.
Similarly, simulation results demonstrate that Model 6 was able to reproduce the results obtained during the clinical study of Zaharieva et al. [54]. Fig 6 shows that by the end of exercise, a basal reduction of 80% 90 min pre-exercise showed the smallest drop in glucose levels compared with a basal reduction of 50% 90 min pre-exercise and stopping the pump at exercise onset.
Shaded area is exercise. Values are mean and standard deviation (n = 11). Error bars represent the standard error.
The quantitative comparison between our simulation results using the Hovorka model [55] and the results of the published clinical study of Haidar et al. [56] are shown in Table 5. The outcomes are comparable, and the p values are not significant. Furthermore, it can be seen from Table 5 that the generated in-silico participants are representative of the real clinical participants.
Discussion
Mathematical models that incorporate the effects of exercise on glucose dynamics play an important role in accelerating the development of safe and effective artificial pancreas systems usable in normal life conditions. In this paper, we proposed six variations of the Bergman Minimal Model [16], compared them on the basis of their physiological plausibility and DIC, and validated them against experimental data.
In people with T1D, exercise increases the rate of their glucose uptake [57]. However, specific details of how and what causes this increase is not completely understood. At least two components are responsible, an insulin-independent component and an insulin-dependent component [58]; they depend on exercise intensity, exercise duration, or a combination of both, as well as other factors. We proposed variations of the Bergman Minimal Model [16] to explore the effects of these factors. The first model simulated a fixed, exercise-induced rise in both glucose effectiveness and insulin sensitivity. Models 2 assumed a zero increase in insulin sensitivity (e1 = 0), and Models 3 assumed a zero increase in glucose effectiveness (e2 = 0). Model 4 related the increase in glucose effectiveness and insulin sensitivity to exercise intensity; meanwhile, Model 5 related the increase in glucose effectiveness and insulin sensitivity to exercise duration. Finally, Model 6 related the increased glucose effectiveness to exercise intensity and related the increased insulin sensitivity to exercise intensity and duration. In our system identification problem, the input to our system is exercise, which is not modified by glucose levels, and thus, this system identification problem is considered in open-loop settings. The insulin infusion rate is indeed altered by glucose levels, but this is considered an inner loop to the system. In addition, insulin infusion rate has been mostly suspended during exercise.
The outcomes of our work are consistent with findings from literature. Models 1 produced a higher DIC compared with Models 4 and 6 which suggests that adding a fixed increase on both glucose effectiveness and insulin sensitivity for the duration of the exercise is inaccurate. This agrees with two previous studies [46, 47] that demonstrated a varying rise in the rate of glucose uptake during moderate exercise. In Model 2, neglecting the increase in glucose effectiveness produced a higher DIC value. The high DIC supports the conclusion indicated by Romeres et al. [59] that the increase in insulin-independent rate of glucose uptake during moderate exercise is approximately 2–3 folds, which is an increase that cannot be neglected. By neglecting the increase in insulin sensitivity, Model 3 produced a lower DIC and improved fits compared with Model 2. This result supports the observation made by previous studies [48, 60, 61] showing that the increase in insulin-independent rate of glucose uptake is rapid at exercise onset, and quickly returns to its initial value after the end of exercise; meanwhile, the increase in insulin-dependent rate of glucose uptake is slow at exercise onset, and takes longer to returns to its initial value after the end of exercise. This is evident by the improvement of Model 3’s DIC, suggesting that during exercise, the increase in glucose effectiveness has a more-noticeable effect on glucose uptake compared with the increase in insulin sensitivity.
A particularly interesting finding in our work is related to the comparison between the models that included the effect of exercise intensity vs exercise duration. Adding the effect of exercise intensity on the increases in glucose effectiveness and insulin sensitivity produced a low DIC value and good fits as shown in Model 4, confirming findings in literature [47, 48] that higher exercise intensity results in higher glucose uptake. Duration alone, on the other hand, did not result in better fits as shown in Model 5, indicating that exercise duration has less effect on glucose uptake rates than exercise intensity. Furthermore, a combination of exercise intensity and exercise duration produced the lowest DIC and best fits as seen in Model 6, which agrees with findings from literature [48, 59]. Romeres et al. [59] showed that the rise in insulin-independent glucose uptake rate does not change for the duration of the exercise; meanwhile, the rise in insulin-dependent glucose uptake rate continues to increase as the exercise duration increases. Additionally, Nguyen et al. [48] showed that exercise intensity affects both insulin-independent and insulin-dependent rates of glucose uptake. Because of Model 6’s DIC value and good fits, it was chosen as the best model to capture the exercise-induced changes in glucose dynamics.
In-silico validation showed that Model 6 successfully predicted changes in glucose levels observed in clinical studies of moderate exercise. The simulations used virtual patients with the Model 6’s parameters estimated from the single-hormone closed-loop moderate-exercise data. The comparison between our simulated results with Model 6 and the clinical study of Rabasa-Lhoret et al. [44] demonstrated similar qualitative changes in glucose levels during and post-exercise, leading to similar clinical conclusions regarding hypoglycemia risks with different exercise intensities (25–75% VO2max), durations (30–60 min), and pre-meal insulin bolus reductions (25–75% reductions). However, there was some quantitative differences with both studies that are likely due to the mismatch between the meals used in the clinical study and those modelled in the simulations, as well as the mismatch in insulin absorption kinetics and insulin sensitivities between the virtual patients and those specific to the real patients recruited in the clinical study. In-silico validation using the Model 6 integrated with Hovorka model showed good quantitative results. The comparison with the clinical study of Haidar et al. [56] demonstrated no significant difference in the percentage of time spent between 4–8 mmol/L, time spent between 4–10 mmol/L, time spent below 4 mmol/L, and time spent below 3.5 mmol/L.
We used rich data to estimate model parameters including plasma glucose, plasma insulin, and oxygen uptake [30]. However, the data were still limited by certain factors. First, the duration of pre-exercise period is not long enough to accurately estimate p1 and p2. Thus, a priori values, taken from literature, were used to produce plausible values. Another limitation is lacking a direct measure of the rate of glucose uptake in our data which require complex experimental procedures with the use of tracers and varying insulin doses in order to give more information about the effects of insulin-dependent and insulin-independent rates of glucose uptake, and thus help estimate e1 and e2 more accurately. Additionally, the data did not include multiple exercise durations which could help develop models with more complex relationship instead of the linear relationship of glucose effectiveness and insulin sensitivity on exercise duration (as is the case in Model-5 and Model-6). Finally, due to lack of post-exercise data, we were not able to model early and late post-exercise period.
In this work, only the effects of moderate exercise were modeled. Interval exercise is associated with higher glucose levels than moderate exercise [9, 62]. This is evident by the data of interval exercise [30] used in this work. In most participants, glucose levels increased during exercise. The structure of our models did not allow the glucose levels to increase during the exercise period, therefore, data from interval exercise produced poor fits and high DIC values (2–3 times higher than current DICs). Thus, the validity of Model 6 was restricted to moderate exercise alone unlike other model [21]. Models that can describe other types of exercise such as interval, vigorous, and anaerobic exercise as well as post-exercise effects could be the subject of future research. Additionally, the effect of long periods of exercise (> 2–3 hours) on glucose dynamics is not completely understood. Therefore, Model 6’s performance for longer exercise durations cannot be validated yet. Furthermore, fat metabolism was not considered when developing the models because exercise took place in earlier postprandial, where carbohydrates and glycogen fulfill the demands of energy and fat metabolism is relatively very low.
The current model does not take into consideration the effects of other factors, such as age and fitness levels as seen in other models [23]. Literature show differences between adults and adolescents in regard to the immediate effect of exercise [63] as a result of the higher growth hormone levels in adolescents [64], which are known to be an opposing effect to insulin action [65, 66], as well as the difference in muscle mass and insulin sensitivity between the two groups [67]. Furthermore, findings show that individuals with good fitness level seem to be more prone to exercise-induced hypoglycemia [68] because of their higher insulin sensitivity. Finally, our model’s performance was not compared to other models from literature due to data and other limitations. This could be the subject of future research.
The new exercise model proposed in this work was developed from real T1D clinical data and was compared with other models that simulate different physiological phenomena experienced by individuals with T1D during moderate exercise.
Conclusion
Developing a model that is able to capture the changes in glucose dynamics during exercise is one of the challenges in improving closed-loop systems [69]. Such models continue to be developed to help build a safe and effective artificial pancreas usable in normal life conditions. The new exercise model proposed in this work can be incorporated in a simulation environment, enabling us to perform in-silico testing that may result in better glycemic control during moderate exercise, and reducing the events of exercise-induced hypoglycemia.
References
- 1. Todd JA. Etiology of type 1 diabetes. Immunity. 2010;32(4):457–67. pmid:20412756
- 2. Haidar A. The Artificial Pancreas: How Closed-Loop Control Is Revolutionizing Diabetes. IEEE Control Systems. 2016;36(5):28–47.
- 3. Foster NC, Beck RW, Miller KM, Clements MA, Rickels MR, DiMeglio LA, et al. State of type 1 diabetes management and outcomes from the T1D Exchange in 2016–2018. Diabetes Technology & Therapeutics. 2019;21(2):66–72. pmid:30657336
- 4. Thabit H, Tauschmann M, Allen JM, Leelarathna L, Hartnell S, Wilinska ME, et al. Home use of an artificial beta cell in type 1 diabetes. The New England Journal of Medicine. 2015;373(22):2129–40. pmid:26379095
- 5. Colberg SR, Sigal RJ, Yardley JE, Riddell MC, Dunstan DW, Dempsey PC, et al. Physical activity/exercise and diabetes: a position statement of the American Diabetes Association. Diabetes Care. 2016;39(11):2065–79. pmid:27926890
- 6. Chimen M, Kennedy A, Nirantharakumar K, Pang T, Andrews R, Narendran P. What are the health benefits of physical activity in type 1 diabetes mellitus? A literature review. Diabetologia. 2012;55(3):542–51. pmid:22189486
- 7. Bohn B, Herbst A, Pfeifer M, Krakow D, Zimny S, Kopp F, et al. Impact of physical activity on glycemic control and prevalence of cardiovascular risk factors in adults with type 1 diabetes: a cross-sectional multicenter study of 18,028 patients. Diabetes care. 2015;38(8):1536–43. pmid:26015557
- 8. Brazeau A-S, Rabasa-Lhoret R, Strychar I, Mircescu HJDc. Barriers to physical activity among patients with type 1 diabetes. Diabetes Care. 2008;31(11):2108–9. pmid:18689694
- 9. Riddell MC, Gallen IW, Smart CE, Taplin CE, Adolfsson P, Lumb AN, et al. Exercise management in type 1 diabetes: a consensus statement. The Lancet Diabetes & Endocrinology. 2017;5(5):377–90. pmid:28126459
- 10. Breton MD, Brown SA, Karvetski CH, Kollar L, Topchyan KA, Anderson SM, et al. Adding heart rate signal to a control-to-range artificial pancreas system improves the protection against hypoglycemia during exercise in type 1 diabetes. Diabetes technology & therapeutics. 2014;16(8):506–11. pmid:24702135
- 11. Dalla Man C, Raimondo DM, Rizza RA, Cobelli C. GIM, simulation software of meal glucose—insulin model. Journal of Diabetes Science and Technology. 2007. pmid:19885087
- 12. Wilinska ME, Chassin LJ, Acerini CL, Allen JM, Dunger DB, Hovorka R. Simulation environment to evaluate closed-loop insulin delivery systems in type 1 diabetes. Journal of diabetes science and technology. 2010;4(1):132–44. pmid:20167177
- 13.
Kanderian SS, Weinzimer S, Voskanyan G, Steil GM. Identification of intraday metabolic profiles during closed-loop glucose control in individuals with type 1 diabetes. Journal of Diabetes Science and Technology. 2009.
- 14. Visentin R, Campos-Náñez E, Schiavon M, Lv D, Vettoretti M, Breton M, et al. The UVA/Padova type 1 diabetes simulator goes from single meal to single day. 2018;12(2):273–81.
- 15. Roy A, Parker RS. Dynamic Modeling of Exercise Effects on Plasma Glucose and Insulin Levels. Journal of Diabetes Science and Technology. 2007;1(3):338–47. pmid:19885088
- 16. Bergman RN, Ider YZ, Bowden CR, Cobelli C. Quantitative estimation of insulin sensitivity. American Journal of Physiology-Endocrinology And Metabolism. 1979;236(6):E667. pmid:443421
- 17. Ewings SM, Sahu SK, Valletta JJ, Byrne CD, Chipperfield AJ. A Bayesian network for modelling blood glucose concentration and exercise in type 1 diabetes. Statistical methods in medical research. 2015;24(3):342–72. pmid:24492795
- 18. Breton MD. Physical Activity—The Major Unaccounted Impediment to Closed Loop Control. Journal of Diabetes Science and Technology. 2008;2(1):169–74. pmid:19885195
- 19. Dalla Man C, Breton MD, Cobelli C. Physical Activity into the Meal Glucose—Insulin Model of Type 1 Diabetes: In Silico Studies. Journal of Diabetes Science and Technology. 2009.
- 20. Dalla Man C, Rizza RA, Cobelli C. Meal simulation model of the glucose-insulin system. IEEE Transactions on biomedical engineering. 2007;54(10):1740–9. pmid:17926672
- 21.
Schneider K. Modeling of Exercise Induced Effects on Blood Glucose Dynamics in T1DM Patients. Master thesis, Swiss Federal Institute of Technology (ETH), Zürich, june 2020.
- 22. Kim J, Saidel GM, Cabrera ME. Multi-Scale Computational Model of Fuel Homeostasis During Exercise: Effect of Hormonal Control. Annals of Biomedical Engineering. 2007;35(1):69–90. pmid:17111212
- 23. Palumbo MC, Morettini M, Tieri P, Diele F, Sacchetti M, Castiglione FJPcb. Personalizing physical exercise in a computational model of fuel homeostasis. 2018;14(4):e1006073.
- 24. Hernandez-Ordonez M, Campos-Delgado D. An extension to the compartmental model of type 1 diabetic patients to reproduce exercise periods with glycogen depletion and replenishment. Journal of biomechanics. 2008;41(4):744–52. pmid:18206156
- 25. Resalat N, El Youssef J, Tyler N, Castle J, Jacobs PGJPo. A statistical virtual patient population for the glucoregulatory system in type 1 diabetes with integrated exercise model. 2019;14(7):e0217301.
- 26. Derouich M, Boutayeb A. The effect of physical exercise on the dynamics of glucose and insulin. Journal of biomechanics. 2002;35(7):911–7. pmid:12052393
- 27. Lenart PJ, Parker RS. Modeling exercise effects in type I diabetic patients. IFAC Proceedings Volumes. 2002;35(1):247–52.
- 28. Xie J, Wang Q. A Data-Driven Personalized Model of Glucose Dynamics Taking Account of the Effects of Physical Activity for Type 1 Diabetes: An In Silico Study. Journal of biomechanical engineering. 2019;141(1):011006. pmid:30458503
- 29. Hobbs N, Hajizadeh I, Rashid M, Turksoy K, Breton M, Cinar AJJods, et al. Improving glucose prediction accuracy in physically active adolescents with type 1 diabetes. 2019;13(4):718–27.
- 30. Taleb N, Emami A, Suppere C, Messier V, Legault L, Ladouceur M, et al. Efficacy of single-hormone and dual-hormone artificial pancreas during continuous and interval exercise in adult patients with type 1 diabetes: randomised controlled crossover trial. Diabetologia. 2016;59(12):2561–71. pmid:27704167
- 31. Storer TW, Davis JA, Caiozzo VJJM. Accurate prediction of VO2max in cycle ergometry. sports si, exercise 1990;22(5):704–12. pmid:2233211
- 32. James DE, Brown R, Navarro J, Pilch PFJN. Insulin-regulatable tissues express a unique insulin-sensitive glucose transport protein. 1988;333(6169):183.
- 33. Goodyear LJ, Hirshman MF, Smith RJ, Horton ES. Glucose transporter number, activity, and isoform content in plasma membranes of red and white skeletal muscle. American Journal of Physiology-Endocrinology And Metabolism. 1991;261(5):E556–E61.
- 34. Fushiki T, Wells JA, Tapscott EB, Dohm GL. Changes in glucose transporters in muscle in response to exercise. American Journal of Physiology-Endocrinology And Metabolism. 1989;256(5):E580–E7. pmid:2655468
- 35. Goodyear LJ, Hirshman MF, Horton ES. Exercise-induced translocation of skeletal muscle glucose transporters. American Journal of Physiology-Endocrinology And Metabolism. 1991;261(6):E795–E9. pmid:1662910
- 36. Ploug T, Galbo H, Richter EAJAJoP-E, Metabolism. Increased muscle glucose uptake during contractions: no need for insulin. 1984;247(6):E726–E31.
- 37. Goodyear LJ, King PA, Hirshman MF, Thompson CM, Horton ED, Horton ES. Contractile activity increases plasma membrane glucose transporters in absence of insulin. American Journal of Physiology-Endocrinology And Metabolism. 1990;258(4):E667–E72. pmid:2159218
- 38. Richter EA, Mikines K, Galbo H, Kiens BJJoap. Effect of exercise on insulin action in human skeletal muscle. 1989;66(2):876–85.
- 39. Douen AG, Ramlal T, Cartee GD, Klip A. Exercise modulates the insulin‐induced translocation of glucose transporters in rat skeletal muscle. FEBS letters. 1990;261(2):256–60. pmid:2178971
- 40. Nesher R, Karl IE, Kipnis DM. Dissociation of effects of insulin and contraction on glucose transport in rat epitrochlearis muscle. American Journal of Physiology-Cell Physiology. 1985;249(3):C226–C32. pmid:3898861
- 41. Goodyear LJ, Kahn BB. Exercise, glucose transport, and insulin sensitivity. Annual review of medicine. 1998;49(1):235–61.
- 42. Wallberg-Henriksson H, Constable S, Young D, Holloszy J. Glucose transport into rat skeletal muscle: interaction between exercise and insulin. Journal of applied physiology. 1988;65(2):909–13. pmid:3049515
- 43. Zorzano A, Balon TW, Goodman MN, Ruderman N. Additive effects of prior exercise and insulin on glucose and AIB uptake by rat muscle. American Journal of Physiology-Endocrinology And Metabolism. 1986;251(1):E21–E6. pmid:3524258
- 44. Rabasa-Lhoret R, Bourque J, Ducros F, Chiasson J-L. Guidelines for premeal insulin dose reduction for postprandial exercise of different intensities and durations in type 1 diabetic subjects treated intensively with a basal-bolus insulin regimen (ultralente-lispro). Diabetes Care. 2001;24(4):625–30. pmid:11315820
- 45. Youngs LM, McMahon SK, Davis EA, Ratnam N, Davey RJ, Ferreira LD, et al. Glucose Requirements to Maintain Euglycemia after Moderate-Intensity Afternoon Exercise in Adolescents with Type 1 Diabetes Are Increased in a Biphasic Manner. The Journal of Clinical Endocrinology & Metabolism. 2007;92(3):963–8.
- 46. Guelfi K, Ratnam N, Smythe G, Jones T, Fournier P. Effect of intermittent high-intensity compared with continuous moderate exercise on glucose production and utilization in individuals with type 1 diabetes. American Journal of Physiology-Endocrinology And Metabolism. 2007;292(3):E865–E70. pmid:17339500
- 47. Shetty VB, Fournier PA, Davey RJ, Retterath AJ, Paramalingam N, Roby HC, et al. Effect of Exercise Intensity on Glucose Requirements to Maintain Euglycemia During Exercise in Type 1 Diabetes. The Journal of Clinical Endocrinology & Metabolism. 2016;101(3):972–80. pmid:26765581
- 48. Nguyen T-Tp, Jacobs Pg, Castle Jr, Lm Wilson, Branigan D, Gabo V, et al. 62-LB: Quantifying Insulin-Mediated and Noninsulin Mediated Glucose Disposal during Exercise in Patients with Type 1 Diabetes. Am Diabetes Assoc; 2019.
- 49.
Gilks WR, Richardson S, Spiegelhalter D. Markov chain Monte Carlo in practice: Chapman and Hall/CRC; 1995.
- 50. Lunn DJ, Thomas A, Best N, Spiegelhalter D. WinBUGS-a Bayesian modelling framework: concepts, structure, and extensibility. Statistics and computing. 2000;10(4):325–37.
- 51.
Spiegelhalter D, Thomas A, Best N, Lunn D. WinBUGS user manual. version; 2003.
- 52.
User’s Manual YSI 2300 STAT PLUS: YSI Incorporated; [https://www.ysi.com/file%20library/documents/manuals%20for%20discontinued%20products/ysi-2300-stat-plus-manual-j.pdf.
- 53. Spiegelhalter DJ, Best NG, Carlin BP, Van Der Linde A. Bayesian measures of model complexity and fit. Journal of the royal statistical society: Series b (statistical methodology). 2002;64(4):583–639.
- 54. Zaharieva DP, McGaugh S, Pooni R, Vienneau T, Ly T, Riddell MCJDc. Improved open-loop glucose control with basal insulin reduction 90 minutes before aerobic exercise in patients with type 1 diabetes on continuous subcutaneous insulin infusion. Diabetes Care. 2019;42(5):824–31. pmid:30796112
- 55. Hovorka R, Shojaee-Moradie F, Carroll PV, Chassin LJ, Gowrie IJ, Jackson NC, et al. Partitioning glucose distribution/transport, disposal, and endogenous production during IVGTT. American Journal of Physiology—Endocrinology And Metabolism. 2002;282(5):E992–E1007. pmid:11934663
- 56. Haidar A, Legault L, Messier V, Mitre TM, Leroux C, Rabasa-Lhoret RJTlD, et al. Comparison of dual-hormone artificial pancreas, single-hormone artificial pancreas, and conventional insulin pump therapy for glycaemic control in patients with type 1 diabetes: an open-label randomised controlled crossover trial. The Lancet Diabetes & Endocrinology. 2015;3(1):17–26.
- 57.
J. Roberts AF, Gregory & Maahs, David & E. Taplin, Craig. Type 1 Diabetes Mellitus and Exercise2018.
- 58. Mallad A, Hinshaw L, Schiavon M, Dalla Man C, Dadlani V, Basu R, et al. Exercise effects on postprandial glucose metabolism in type 1 diabetes: a triple tracer approach. American Journal of Physiology-Heart and Circulatory Physiology. 2015. pmid:25898950
- 59. ROMERES D, SCHIAVON M, BASU A, COBELLI C, BASU R, DALLA MAN C. 1862-P: Insulin-Independent Glucose Utilization during Exercise Is Impaired in Type 1 Diabetes: A New Model Based Analysis. Am Diabetes Assoc; 2019.
- 60. Funai K, Schweitzer GG, Sharma N, Kanzaki M, Cartee GDJAJoP-E, Metabolism. Increased AS160 phosphorylation, but not TBC1D1 phosphorylation, with increased postexercise insulin sensitivity in rat skeletal muscle. american journal of physiology endocrinology and metabolism. 2009;297(1):E242–E51. pmid:19435856
- 61.
Breton M, Dalla Man C, King C, Anderson S, Farhy L, Cobelli C, et al., editors. Effect of exercise on insulin action assessed by the minimal model. Proceedings of the Diabetes Technology Meeting, Atlanta, GA; 2006.
- 62. Harmer AR, Chisholm DJ, McKenna MJ, Morris NR, Thom JM, Bennett G, et al. High-intensity training improves plasma glucose and acid-base regulation during intermittent maximal exercise in type 1 diabetes. 2007;30(5):1269–71.
- 63. Ben Brahim N, Place J, Renard E, Breton MD. Identification of main factors explaining glucose dynamics during and immediately after moderate exercise in patients with type 1 diabetes. Journal of diabetes science and technology. 2015;9(6):1185–91. pmid:26481644
- 64. Galassetti PR, Iwanaga K, Crisostomo M, Zaldivar FP, Larson J, Pescatello AJPd. Inflammatory cytokine, growth factor and counterregulatory responses to exercise in children with type 1 diabetes and healthy controls. Pediatric Diabetes. 2006;7(1):16–24. pmid:16489970
- 65. Rizza RA, Mandarino LJ, Gerich JEJD. Effects of growth hormone on insulin action in man: mechanisms of insulin resistance, impaired suppression of glucose production, and impaired stimulation of glucose utilization. Diabetes. 1982;31(8):663–9.
- 66. Hansen I, Tsalikian E, Beaufrere B, Gerich J, Haymond M, Rizza RJAJoP-E, et al. Insulin resistance in acromegaly: defects in both hepatic and extrahepatic insulin action. American Journal of Physiology-Endocrinology and Metabolism. 1986;250(3):E269–E73. pmid:3513613
- 67. Szadkowska A, Pietrzak I, Mianowska B, Bodalska‐Lipińska J, Keenan H, Toporowska‐Kowalska E, et al. Insulin sensitivity in type 1 diabetic children and adolescents. Diabetic Medicine. 2008;25(3):282–8. pmid:18279410
- 68. Al Khalifah R, Suppère C, Haidar A, Rabasa‐Lhoret R, Ladouceur M, Legault LJDM. Association of aerobic fitness level with exercise‐induced hypoglycaemia in Type 1 diabetes. Diabetic Medicine. 2016;33(12):1686–90. pmid:26773719
- 69. Riddell MC, Zaharieva DP, Yavelberg L, Cinar A, Jamnik VK. Exercise and the Development of the Artificial Pancreas. Journal of Diabetes Science and Technology. 2015;9(6):1217–26. pmid:26428933