A GROUNDBREAKING TOOL FOR TYPE 1 DIABETES TREATMENT R&D.
The first (and currently only) in silico diabetes model accepted by the FDA as a substitute for pre-clinical animal testing of new treatment strategies for Type 1 Diabetes Mellitus.
[wptabs]
[wptabtitle] Overview[/wptabtitle]
[wptabcontent]
T1DMS is a computer simulator of the dynamics of the human metabolic glucose-insulin system. It is based on a kinetic model developed from quantitative knowledge of glucose-insulin metabolism in conjunction with data from a large-population human subject study.
Click here to learn more …
In January 2008, T1DMS became the first computer tool accepted by the FDA as a substitute for animal trials in the pre-clinical testing of certain control strategies in Type 1 Diabetes Mellitus. Implemented in Simulink/MATLAB, T1DMS uses a well-defined set of interfaces for testing of closed loop, user-defined treatment scenarios.
Click here to learn more …
TEG offers Simulation Services for the development and optimization of T1D-related devices, protocols, and treatments using the T1DMS. Used in conjunction with the full FDA-accepted in silico subject population, TEG’s T1DMS Simulation Services can be a fast-track to FDA IDE submission.
Click here to learn more …
Click here to read the T1DMS Simulation Services Whitepaper …
Additionally, TEG distributes and supports a Distributed Version of the T1D metabolic simulator (with a smaller “sample” in silico patient population). This Distributed Version provides members of the wider diabetes research community with a T1DMS system for independent research and teaching.
Click here to learn more …
[/wptabcontent]
[wptabtitle] The T1DMS Difference[/wptabtitle]
[wptabcontent]
Most diabetes models provide only average group-level responses to different treatment strategies based on study populations of randomized controlled clinical trials. Conversely, using T1DMS, pre-clinical testing and experiments are conducted at the level of an individual subject providing insight into the intra- and inter-personal differences seen in humans, in response to the treatment protocol.
The key advantages to using T1DMS are:
- Individualized, intra-personal results
- Inter-personal differences are revealed across the spectrum of human variation
- T1DMS is, to our knowledge, the only model with pediatric and adolescent populations – others only have adults
- Model design can be identical to a proposed clinical study*
- Provides individual real-time glucose traces* for visual understanding of insulin-glucose fluctuation
- Population results are calculated from the individual results as in a clinical trial
* Per FDA’s Draft Guidance for Industry and Food and Drug Administration Staff – The Content of Investigational Device Exemption (IDE) and Premarket Approval (PMA) Applications for Low Glucose Suspend (LGS) Device Systems[/wptabcontent]
[wptabtitle] Applications & Benefits[/wptabtitle]
[wptabcontent]
The capabilities of T1DMS have opened up new possibilities in the mission to bring novel treatments, protocols, and devices to diabetes patients. T1DMS gives researchers a well-defined and validated tool to assess diabetes treatment strategies and controls while saving time and money.
Specific R&D benefits include the following:
- T1DMS has been shown to reduce the research, development and IDE approval process from 2.5 years to less than 1 year
- T1DMS can save millions of dollars in bringing new devices and treatments to the marketplace, as an accepted replacement for pre-clinical animal studies
- T1DMS enables rigorous device/protocol testing scenarios on virtual subjects that would not be possible/ethical in animal and human studies
Typical T1DMS simulation studies investigate “real-life” scenarios including:
- Modeling and testing of insulin dosing algorithms
- Subject-specific basal / bolus insulin dosing
- Insulin pumps and continuous glucose monitors
- Meal patterns and amounts based on carbohydrate counts
- Temporary insulin resistance / exercise component
- SMBG testing with manual correction boluses
- Rescue carbohydrates
- Full range of options to model a person in-the-loop
Click here to view example scenarios for which T1DMS has been applied …
T1DMS provides a variety of analysis tools and metrics to understand the simulation data. Such analyses include:
- Blood glucose excursion analysis
- Control Variability Grid Analysis
- Error Grid – Continuous Glucose Analysis
- Number of subjects experiencing low and high glucose episodes
- Full range of individual (per subject) responses and population responses
[/wptabcontent]
[/wptabs]
The T1DMS system was developed through research efforts at the Universities of Padova and Virginia.
Specifically, using quantitative knowledge of human glucose-insulin metabolism, a kinetic simulation model thereof was created and refined. The parameters for this kinetic model were determined via triple tracer NMR and PET experiments using a large population of human subjects under a standardized meal protocol.
Such studies provided quasi-model-independent estimates of major glucose and insulin fluxes (e.g., meal rate of appearance, endogenous glucose production, utilization of glucose, insulin secretion).
The T1DMS system employs these parameters to simulate glucose metabolism – and its regulation by insulin – during a meal at both the organ-tissue and whole-body levels (e.g., hepatic glucose production, muscle utilization, renal extraction).
Implemented in Simulink/MATLAB, T1DMS uses a well-defined set of interfaces for testing of closed loop, user-defined treatment scenarios in the Simulink controller block with prescribed meal profiles. Additionally, insulin pump injection parameters and accurate sensor noise profiles for several devices are incorporated into the simulator.
When combined with the FDA-accepted in silico human subject population, T1DMS comprises a well-defined, validated tool to assess results of a proposed closed loop system early in the development phase of treatment strategies and protocols. As such, it has been accepted by the FDA as a substitute for pre-clinical animal studies for various glucose control system scenarios.
The Epsilon Group provides simulation services to support examination of your proprietary algorithm, device, or compound under extensive and rigorous testing. Such simulations employ a large in silico subject population which spans the variability of persons with Type 1 Diabetes.
Simulation scenarios can be designed to model self-management/lifestyle patterns and/or clinical protocols. Additionally, scenarios which would be potentially dangerous and/or unethical in a clinical setting can be simulated to examine possible limitations of the device or intervention under study.
Specific simulation details include the following:
In Silico population
- Two distinct sets of 300 in silico subjects
- Set 1 is used for iterative testing
- Set 2 is the “Gold Standard” set used in final simulation studies to be submitted to the FDA for IDE approval
- Each such population set consists of
- 100 Adults
- 100 Adolescents (ages 13-18)
- 100 Children (ages 2-12)
Basic Simulation Input (adjustable for specific requirements of study)
- Meal profiles: amount, timing, and duration of meal(s)
- Insulin treatment: amount and timing of basal/bolus insulin dose(s)
- Time of simulation and regulation (time of day and night)
- Control Law definition* in control algorithms
* Several standard controller blocks are provided. Specific simulation scenarios may involve simple to complex closed-loop controls each of which may include manual insulin injections in addition to insulin dosing controlled by the specific simulation algorithm
In Silico Subject-Specific Data Available to Examine and Tune Treatment Options
- Age
- Body weight (kg)
- ‘Optimal’ subject-specific basal insulin dose rate (U/hr)
- ‘Optimal’ subject-specific carbohydrate ratio (CR, g/U)
- ‘Optimal’ subject-specific maximum drop (MD, mg/dl per Unit insulin)
- Total daily insulin and a measure of insulin sensitivity (TDI, U/day)
- Metabolic testing results may be simulated for individual subjects and incorporated into treatment plans prior to the regulated model run
Simulation Output (per Subject)*
- Blood glucose (BG) values (mg/dl per minute)
- Simulated sensor BG readings (mg/dl per minute)
- Simulated time
- Basal/Bolus insulin injections (pmol/minute)
- CHO – carbohydrate meal dose and timing
- Individual subject identification
- Subject system states: optimal control vs. poor control with hypo/hyper-glycemia
* Additional model output may be collected per specific requirements of study
Population and “Per Subject” Analysis
- BG trace: traditional plot of BG data vs. simulation time
- BG density: probability distribution of BG values with calculated probabilities of values below/within/above a preset target range
- Glucose risk trace: includes fluctuations of low blood glucose indexes (LBGI) and high blood glucose indexes (HBGI), computed hourly; emphasizes large glucose excursions and suppresses fluctuations within target to highlight essential BG variances into hypo/hyper-glycemic levels
- Aggregated BG trace: corresponds to time spent below/within/above a reset target range
- Histogram of BG rate of change: represents the spread and range of glucose transitions (mg/dl per minute)
- Poincaré plot: spread of data indicates system (subject) stability; more widespread data points are associated with unstable diabetes and rapid glucose fluctuations
- Control variability grid analysis (CVGA): event-based analysis representing the effectiveness of glycemic control
- Clark continuous glucose error grid analysis (CG-EGA) and standard Clarke error grid analysis (SG-EGA)
Additionally, the following “per subject” outcomes relevant to glucose control can be imported into Excel to create a Safety & Efficacy Table for the population:
- Mean blood glucose reading
- Mean pre/post-meal blood glucose values
- Percent time in severe hypoglycemia (BG ≤ 50 mg/dl)
- Percent time in hypoglycemia (BG ≤ 70 mg/dl or alternate low target BG value per specific study requirements)
- Percent time in euglycemia (70 mg/dl < BG ≤ 180 mg/dl or alternate BG target zone per specific study requirements)
- Percent time in hyperglycemia (BG > 180 mg/dl or alternate high target BG value per specific study requirements)
- Percent time in severe hyperglycemia (BG > 300 mg/dl)
- Low BG index (LBGI)
- High BG index (HGBI)
- BG risk index (BGRI)
- Standard deviation of BG rate of change
The UVA/Padova T1DMS distributed software is available from The Epsilon Group. It provides members of the wider diabetes research community with a T1DMS system for independent research and teaching. Furthermore, if your final goal is FDA IDE submission for a novel device, protocol, or treatment strategy, you can conduct early stage R&D using the distributed version of the software prior to initiating a final T1DMS simulation via TEG’s Simulation Services.
T1DMS is an integrated dynamic computer model of human insulin-glucose metabolism in the presence of Type 1 Diabetes. In the Distributed Version, each individual of an in silico population of 30 subjects responds to insulin dosing, dietary factors and exercise factors to simulate daily lifestyle or clinical testing of diabetes management strategies. The distributed version is identical in functionality to the full version. The study population is 30 subjects (10 adults, 10 adolescents and 10 children) – hence a smaller sample size for testing. They both require Matlab/Simulink and the curve-fitting toolbox. There are no differences in the functionality. The results of such simulations can be used to: (i) test the efficacy, safety, and limitations of new therapies under varying conditions, (ii) compare existing treatments, and (iii) bolster research and product development strategies.
Specific simulation details include the following:
In Silico population
- 30 in silico subjects consisting of
- 10 Adults
- 10 Adolescents (ages 13-18)
- 10 Children (ages 2-12)
Basic Simulation Input (adjustable for specific requirements of study)
- Meal profiles: amount, timing, and duration of meal(s)
- Insulin treatment: amount and timing of basal/bolus insulin dose(s)
- Time of simulation and regulation (time of day and night)
- Control Law definition* in control algorithms
* Several standard controller blocks are provided. Specific simulation scenarios may involve simple to complex closed-loop controls each of which may include manual insulin injections in addition to insulin dosing controlled by the specific simulation algorithm
In Silico Subject-Specific Data Available to Examine and Tune Treatment Options
- Age
- Body weight (kg)
- ‘Optimal’ subject-specific basal insulin dose rate (U/hr)
- ‘Optimal’ subject-specific carbohydrate ratio (CR, g/U)
- ‘Optimal’ subject-specific maximum drop (MD, mg/dl per Unit insulin)
- Total daily insulin and a measure of insulin sensitivity (TDI, U/day)
- Metabolic testing results may be simulated for individual subjects and incorporated into treatment plans prior to the regulated model run
Simulation Output (per Subject)*
- Blood glucose (BG) values (mg/dl per minute)
- Simulated sensor BG readings (mg/dl per minute)
- Simulated time
- Basal/Bolus insulin injections (pmol/minute)
- CHO – carbohydrate meal dose and timing
- Individual subject identification
- Subject system states: optimal control vs. poor control with hypo/hyper-glycemia
* Additional model output may be collected per specific requirements of study
Population and “Per Subject” Analysis
- BG trace: traditional plot of BG data vs. simulation time
- BG density: probability distribution of BG values with calculated probabilities of values below/within/above a preset target range
- Glucose risk trace: includes fluctuations of low blood glucose indexes (LBGI) and high blood glucose indexes (HBGI), computed hourly; emphasizes large glucose excursions and suppresses fluctuations within target to highlight essential BG variances into hypo/hyper-glycemic levels
- Aggregated BG trace: corresponds to time spent below/within/above a reset target range
- Histogram of BG rate of change: represents the spread and range of glucose transitions (mg/dl per minute)
- Poincaré plot: spread of data indicates system (subject) stability; more widespread data points are associated with unstable diabetes and rapid glucose fluctuations
- Control variability grid analysis (CVGA): event-based analysis representing the effectiveness of glycemic control
- Clark continuous glucose error grid analysis (CG-EGA) and standard Clarke error grid analysis (SG-EGA)
Additionally, the following “per subject” outcomes relevant to glucose control can be imported into Excel to create a Safety & Efficacy Table for the population:
- Mean blood glucose reading
- Mean pre/post-meal blood glucose values
- Percent time in severe hypoglycemia (BG ≤ 50 mg/dl)
- Percent time in hypoglycemia (BG ≤ 70 mg/dl or alternate low target BG value per specific study requirements)
- Percent time in euglycemia (70 mg/dl < BG ≤ 180 mg/dl or alternate BG target zone per specific study requirements)
- Percent time in hyperglycemia (BG > 180 mg/dl or alternate high target BG value per specific study requirements)
- Percent time in severe hyperglycemia (BG > 300 mg/dl)
- Low BG index (LBGI)
- High BG index (HGBI)
- BG risk index (BGRI)
- Standard deviation of BG rate of change
System and Software Requirements
- PC with Windows 7® or Windows XP® (SP2 or later)
- 1 GB of RAM
- 10 MB free disk storage space
- Matlab® 2009b (or later) (32- or 64-bit) with Simulink® and the CurveFitting ToolboxTM
- Parallel Computing ToolboxTM required for multi-core processing (not required to run model)
| Real-Life Scenario | Clinical Scenario | Meal Compostion |
|---|---|---|
Ideal situation, optimal control: a well-controlled person with Type 1 Diabetes. | Baseline:
|
|
Insulin stacking: an overestimation of the insulin bolus size occurs, causing an incorrect insulin:carbohydrate ratio (high); or, the subject has a high insulin sensitivity factor and gives themselves "a little more to cover dessert". | Hyperinsulization - Risk of Hypoglycemia:
|
|
Missed meal bolus or Underestimated Bolus Dose resulting in incorrect insulin:carbohydrate ratio (low) or insulin sensitivity factor (low) | Hypoinsulization - Risk of Hyperglycemia:
|
|
Insulin stacking and exercise: teenager eats an after-school snack, then administers incorrect bolus (high); shortly thereafter, teenager goes to sports practice with no temporary basal adjustment or suspension to compensate for exercise then participates in a long period of exercise. | Exercise/Teenage challenge - Risk of Hypoglycemia:
|
|
Mis-timed insulin dosing: insulin given in anticipation of a meal, but, the meal is delayed; alternatively, the insulin bolus is given much after meal is consumed; not using combo bolus for high carbs / high fat meals. | Mismatch timing challenge - Risk of both Hypo-/Hyper- glycemia:
|
|
Hypoglycemia avoidance: insulin given appropriately for meals, but given at a consistently (slightly) lower dose than needed for blood glucose target range (or snack is not covered). | Hypoglycemia avoidance - Risk of Hyperglycemia:
|
|
Variable insulin sensitivity due to illness: insulin given appropriately for meals - based on measured blood glucose levels, but illness is present resulting in fever, vomiting, inability to eat. | Illness - Risk of both Hypo-/Hyper- glycemia:
|
|
SMBG meter error: meter reading variances of +/- 5/10/15/20 % from actual IV blood glucose levels. | SMGB meter error variability - Risk of both Hypo-/Hyper- glycemia:
|
|
Evening or weekend binge eating: overeating on occasion. | Evening/Weekend/Holiday binge eating - Risk of both Hypo-/Hyper- glycemia:
|
|
Elite athlete with Type 1 Diabetes: intense regular exercise and carb loading | High insulin sensitivity - Risk of both Hypo-/Hyper- glycemia:
|
|