In an embodiment of the second aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the insulin dose is calculated by adding to a basal insulin rate insulin bolus, and wherein the insulin dose is corrected, using a correction for exercise module, by using information about physical exercise.

In a third aspect, a system or method for providing a real-time therapeutic recommendation for use in type 1 diabetes management in a patient is provided, comprising: inputting, to a decision support module, data including continuous glucose monitoring data obtained from a glucose sensor, insulin pump data, and other diabetes management data; generating, based on the input data, a first therapeutic recommendation; displaying the therapeutic recommendation to a patient; inputting, to the decision support model, data indicative of an action taken by the patient in response to receiving the displayed therapeutic recommendation, the data including continuous glucose monitoring data obtained from the glucose sensor and insulin pump data; and generating, based on the input data and the data indicative of an action taken by the patient, a second therapeutic recommendation.

In an embodiment of the third aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the continuous glucose monitoring data is selected from the group consisting of a glucose concentration trend, a predicted glucose concentration, a number of past hyperglycemic events, and a number of past hypoglycemic events.

In an embodiment of the third aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the other diabetes management data is selected from the group consisting of meal data and physical exercise data.

In an embodiment of the third aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the first therapeutic recommendation or the second therapeutic recommendation is selected from the group consisting of a recommended insulin dose and a basal insulin change.

In an embodiment of the third aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the first or second therapeutic recommendation is delivered through a smartphone notification.

In an embodiment of the third aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the decision support module compares the data indicative of an action taken by the patient responsive to the first therapeutic recommendation to determine if the patient has taken the therapeutic recommendation, has ignored the therapeutic recommendation, or has taken an action similar to the therapeutic recommendation.

In an embodiment of the third aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the action taken by the patient is employed to modify a decision-making module configured to simulate real-life situations and everyday patient behaviors, wherein the decision-making module utilizes self-monitored blood glucose data measurement errors, continuous glucose monitoring data measurement errors, miscalculations of meal carbohydrate content, early insulin administration errors, delayed insulin administration errors, and missed insulin bolus errors.

In an embodiment of the third aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the decision-making module utilizes patient data to assess glycemic control or efficacy of the decision support module, or to suggest modifications to the decision support module to improve its performance.

In a fourth aspect, a system or method for delivering a therapeutic recommendation for insulin delivery is provided, comprising: collecting data from a patient over a period of time, the data including continuous glucose monitoring data obtained from a glucose sensor and insulin pump data; generating, based on the collected data, a model comprising estimated parameters of the patient's physiology that reproduces the patient's glucose-insulin dynamics; inputting the estimated parameters into a type 1 diabetes patient glucose-insulin model, whereby an individualized patient model is obtained; utilizing the individualized patient model in patient-specific simulations to estimate optimal insulin therapy parameters or individualized guidelines for continuous glucose monitoring-driven insulin dosing; and outputting the optimal insulin therapy parameters.

In an embodiment of the fourth aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), outputting the optimal insulin therapy parameters comprises displaying the therapeutic recommendation to a patient on a display.

In an embodiment of the fourth aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), the display is a smartphone display.

In an embodiment of the fourth aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), outputting the optimal insulin therapy parameters comprises issuing instructions to an insulin pump for delivering an insulin dose.

In an embodiment of the fourth aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein), utilizing the individualized patient model in patient-specific simulations comprises iteratively running simulations, wherein at the first iteration, the patient's current insulin therapy is implemented, then, at each subsequent iteration, the therapy parameters are updated until it is determined that the therapy parameters satisfy predetermined criteria.

In an embodiment of the fourth aspect, which is generally applicable (i.e., independently combinable with any of the aspects or embodiments identified herein),it is determined if the therapy parameters need to be updated by assessing a level of glycemic control.

Any of the features of an embodiment of the first through fourth aspects is applicable to all aspects and embodiments identified herein. Moreover, any of the features of an embodiment of the first through fourth aspects is independently combinable, partly or wholly with other embodiments described herein in any way, e.g., one, two, or three or more embodiments may be combinable in whole or in part. Further, any of the features of an embodiment of the first through fourth aspects may be made optional to other aspects or embodiments. Any aspect or embodiment of a method can be performed by a system or apparatus of any aspect or embodiment, and any aspect or embodiment of a system can be configured to perform a method of any aspect or embodiment.

These and other aspects of the invention will become apparent from the following description of the preferred embodiments taken in conjunction with the following drawings. As would be obvious to one skilled in the art, many variations and modifications of the invention may be effected without departing from the spirit and scope of the novel concepts of the disclosure.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram of a general structure of the DM model whose main components are labelled with letters A, B, C and D.

FIG. 2 is a block diagram of an application of the DM model to test and compare in silico SMBG- and CGM-driven T1D therapies.

FIG. 3 is a block diagram of a possible embodiment of the T1D therapy model. White blocks are activated in both SMBG- and CGM-driven therapies, while blocks highlighted in yellow are activated only in CGM-driven therapies.

FIG. 4 includes several graphs showing comparison of results obtained by embodiments A and B in a representative subject. Top panel: SMBG samples (black circles) and CGM profile (blue line), used for insulin dosing respectively in embodiments A and B. Middle panel: insulin doses obtained by embodiments A (black circles) and B (blue circles). Bottom panel: BG profile obtained from the therapy of embodiment A (black line) compared with that obtained as result of the therapy of embodiment B (blue line).

FIG. 5 is a block diagram of an application of the DM model to test and compare candidate guidelines for CGM-driven insulin therapy.

FIG. 6 is a block diagram of an application of the DM to off-line optimization of individualized insulin therapies.

FIG. 7 is a block diagram of an application of the DM model to test and compare CGM-driven decision-support systems.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

An embodiment is now described in detail. Referring to the drawings, like numbers indicate like parts throughout the views. Unless otherwise specifically indicated in the disclosure that follows, the drawings are not necessarily drawn to scale. As used in the description herein and throughout the claims, the following terms take the meanings explicitly associated herein, unless the context clearly dictates otherwise: the meaning of “a,” “an,” and “the” includes plural reference, the meaning of “in” includes “in” and “on.”

In the four subsections, A to D below, the four components of the DM model are described in detail.

T1D Patient Submodel

This submodel (block A, FIG. 1) simulate the BG and IG concentration profiles in a T1D patient resulting from carbohydrates intake and insulin pump infusions. In a preferred embodiment, a model of T1D patient glucose-insulin system can be used, which received FDA approval to substitute preclinical trials for certain insulin treatments. In particular, the T1D patient glucose-insulin model describes physiological events related to BG dynamics like gastrointestinal absorption of carbohydrates, insulin and glucagon kinetics and their action in regulating glucose endogenous production and utilization. The model can reproduce the dynamics of a large population of virtual patients (each one represented by a set of model parameters) and run multi-day simulations thanks to the description of inter- and intra-day variability of insulin sensitivity.

Device for Glucose Monitoring Submodel

The device for glucose monitoring submodel (block B of FIG. 1) generates the glucose measurements used by patients in their diabetes management to adjust insulin doses or carbohydrate intake. The model receives the glucose concentration (BG or IG) in input and can be configured to simulate SMBG spot measurements or CGM profile. When a SMBG driven therapy is simulated, the device for glucose monitoring model produces sparse BG measurements obtained by down sampling the BG profile and corrupting samples by using a model of SMBG measurement error. In a preferred embodiment, a stochastic model of general applicability can be employed, whose parameters can be tuned to reproduce measurements collected by different commercial SMBG devices. When a CGM driven therapy is simulated, the device for glucose monitoring model generates a continuous-time signal obtained by corrupting the IG profile by using a model of CGM measurement error. In a preferred embodiment, a multi-component model of general applicability can be employed, whose parameters can be tuned to reproduce measurements collected by different commercial CGM devices.

T1D Therapy Submodel

The T1D therapy submodel (block C in FIG. 1) simulates the therapeutic decisions that T1D patients make in their everyday life on the basis of meal composition and BG monitoring. To simulate realistic patient behavior, the model includes the principal common errors that patients are used to doing in their diabetes management. The model receives in input information about meal carbohydrate content and physical exercise, BG measurements (either CGM or SMBG) and patient-specific therapy parameters. The model returns in output carbohydrate intake (both meals and other carbohydrate consumptions) and insulin doses. The T1D therapy model can be configured to simulate either a SMBG-driven insulin therapy or a CGM-driven insulin therapy.

In a preferred embodiment, when an SMBG driven therapy is simulated, carbohydrate intake is calculated by adding to the meal carbohydrate content hypotreatments, i.e. adjunctive carbohydrate intakes (e.g. snacks of 15-20 grams) necessary to resolve hypoglycemic events; pre-meal insulin boluses are computed before meals to cover meal carbohydrate intake on the basis of an estimate of meal carbohydrate content and current BG concentration.

When a CGM driven therapy is simulated, CGM data are used to generate, besides pre-meal boluses, post-meal correction boluses in response to hyperglycemic alarms, to generate hypo-treatments in response to hypoglycemic alarms, to correct insulin boluses for CGM trend/prediction and to correct basal insulin for current BG value/trend/prediction.

Insulin Pump Submodel

The insulin pump model (block D in FIG. 1) is simply an actuator that receives in input insulin doses generated by the T1D therapy model and returns in output the insulin pump infusion rate (e.g. in mmol/min). A parameter of the model is the duration of insulin boluses infusion.

Aspects of Methods and Systems

The methods and systems described herein provide various advantages and improvements over state-of-art methods and systems.

First, the methods or systems include a model of patient decision-making (DM), i.e. the complex process in which diabetic patients in their every-day life analyze glucose monitoring data and take therapeutic decisions (e.g. insulin dosing). In particular, the DM model takes into account possible errors the patients are used to doing in their diabetes management e.g. miscalculation of meal carbohydrate content, early/delayed insulin doses administrations and missed boluses occurrence. Such components, linked to patient's behavior and not considered by previous simulation approaches, are key to obtaining a realistic simulation scenario and test SMBG/CGM driven therapy strategies in a nearly real-life context.

Second, the embodiments are based on a realistic complex non-linear model of T1D patient glucose-insulin dynamics, which allows a better description of patient physiology than the simplified models used in conventional methods. In addition, the used glucose-insulin model allows running multi-day simulations (not possible with conventional methods) thanks to the description of inter- and intra-day variability of insulin sensitivity.

Finally, the modular structure of the embodiments allows a great number of configurations that can be used for many applications involving in silico testing of T1D therapies using realistic models of commercial SMBG and CGM in a wide population of virtual subject in which intra- and inter-day variability is properly considered.

Applications

In the following subsections, four possible applications of the comprehensive DM model are presented i.e. the use of the DM model to assess in silico if/when CGM can safely substitute SMBG in T1D management, to develop and test candidate guidelines for CGM driven insulin dosing, to optimize and individualize off-line insulin therapies and to develop and test decision support systems.

1. Assessment of Non-Adjunct Use of CGM

A direct application of the DM model is its use to test in silico if CGM can safely substitute SMBG in diabetes management. For such a purpose, the DM model can be employed as represented in FIG. 2. In particular, several days simulations can be run in a large set of virtual patients by using two versions of the DM model: a first version configured to simulate SMBG driven insulin therapy (top model in FIG. 2) and a second version configured to simulate CGM driven insulin therapy (bottom model in FIG. 2). After simulation, virtual patient BG profiles can be used to assess the level of glycemic control achieved by the computation of a number of metrics. Finally, by comparing the metrics values obtained by the SMBG and the CGM configuration it can be determined if/when the use of CGM is as safe as SMBG for insulin dosing and if/when the use of CGM can even improve glycemic control in T1D management.

Simulations comparing the efficacy of SMBG and CGM driven T1D therapies were run by two embodiments of the DM model: embodiment A in which a SMBG driven therapy is simulated and embodiment B in which a CGM driven therapy is simulated. For both the embodiments in block A a model as described elsewhere herein is used. Concerning block B, in embodiment A a model of SMBG error derived for the One Touch Ultra 2 (Lifescan Inc., Milpitas, CA) is used, in embodiment B a model of the CGM sensor derived for the Dexcom G4 Platinum (Dexcom Inc., San Diego, CA) is used. The detailed structure for the T1D therapy model in embodiments A and B is reported in FIG. 3. Specifically, a single scheme is reported in which white blocks are activated both in embodiments A and B, while yellow blocks are activated in embodiment B only.

In both the embodiments, carbohydrate intake is calculated by adding to the meal carbohydrate content hypo-treatments, generated by the hypo-treatment block in FIG. 3.

Meal boluses are computed before meals to cover meal carbohydrate intake. The carb-counting block in FIG. 3 simulates patient estimate of meal carbohydrate content by exploiting a model of patient carbohydrate counting error. In this embodiment carb-counting errors (in %) are simulated by a normal distribution with zero mean and 20% standard deviation. The estimate of meal carbohydrate content is then used to calculate the insulin dose required to cover the meal. The computation is performed by the meal bolus block of FIG. 3 by using patient carb ratio (CR) according to the guidelines.

A correction insulin bolus is added to meal bolus in order to take into account patient current BG concentration. The correction bolus block in FIG. 3 calculates correction boluses by using patient correction factor (CF) and current BG measurement (collected by either SMBG or CGM).

Variability in the pre-meal bolus administration time is generated by the bolus time variability block in FIG. 3 to simulate the occurrence of early/delayed bolus administrations due to unexpected setbacks that may happen in real life. In this specific embodiment, pre-meal boluses time is sampled from a uniform distribution in the interval between 10 min before and 10 min after the meal beginning.

From time to time, the occurrence of missed boluses, regarded the major cause of suboptimal glycemic control, is simulated by the missed bolus block of FIG. 3. Despite the missed bolus block has been included in the general scheme of the embodiment, to obtain the preliminary results presented in the current file the occurrence of missed boluses has not been simulated.

The final insulin dose is calculated by adding to basal insulin rate (mmol/min) insulin boluses.

Final insulin doses are corrected by using information about physical exercise by the correction for exercise block of FIG. 3. Despite the correction for exercise block has been included in the general scheme of the embodiment, to obtain the preliminary results presented in the current file such a correction has not been simulated.

Insulin therapy parameters, CR and CF, as well as the basal insulin infusion rate used in the pump model are calculated from patient body weight and total daily insulin according to guidelines.

In embodiment A: 16-gram hypo-treatments are generated every 15 min whenever BG drops below 40 mg/dl, which is assumed to be the threshold at which hypoglycemia symptoms start appearing, until the BG returns inside the euglycemic range.

In embodiment B: 16-gram hypo-treatments are generated every 30 minutes when BG drops below 70 mg/dl, which is assumed to be the CGM hypoglycemic alarm threshold, until the CGM profile returns inside the euglycemic range.

Post-meal correction boluses are generated, besides pre-meal boluses, when CGM increases over 180 mg/dl, that is assumed to be the CGM hyperglycemic alarm threshold, and at least 2 hours passed since last insulin bolus. Post-meal correction boluses are generated as pre-meal correction boluses dividing the dose by two as an empirical way to account for insulin on board.

Both pre-meal and post-meal insulin boluses are adjusted to account for glucose trend/prediction calculated in real-time from the CGM profile by the glucose trend/prediction block in FIG. 3. The block labelled as correction for glucose trend/prediction in FIG. 3 adjusts the total insulin bolus (sum of meal and correction bolus) according to CGM trend/prediction. Since CGM has not been approved for insulin dosing yet, currently there are no officially approved guidelines to adjust insulin doses on the basis of CGM trend/prediction. In the literature, only two CGM trials are available in which specific instructions were given to patients on how to adjust their insulin doses on the basis of CGM trend, e.g., the DirectNet and the Juvenile Diabetes Research Foundation studies. Both these studies recommended to increase/decrease the insulin dose by 10% in response to a moderately increasing/decreasing CGM trend and to increase/decrease the insulin dose by 20% in response to a more rapidly increasing/decreasing CGM trend. These guidelines are adopted to implement the correction for glucose trend/prediction block in this embodiment.

Basal insulin infusion rate is corrected to account for current CGM value/trend/prediction by the correction of basal insulin for CGM block, whose function is to suspend basal insulin delivery when patient CGM profile is (or is going to be) below a certain level. Despite the correction of basal insulin for CGM block has been included in the general scheme of the embodiment, to obtain the preliminary results presented in the current file such a correction has not been simulated.

Finally, concerning the insulin pump model implementation, the duration of insulin boluses has been set to 1 minute.

Examples of Results

Simulations were run in 20 adult virtual subjects for a 7-day period by using the DM model with both embodiment A and B. Each virtual subject has been associated to a specific scenario, which is the same used for embodiment A and B, containing three meals per day sampled from normal distributions with mean equal respectively to 60 grams for breakfast, 50 grams for lunch, and 70 grams for dinner, and CV equal to 20%. After simulation, the BG profiles obtained in each virtual subject for embodiments A and B have been compared. In FIG. 4, an example in which a better a glycemic control was achieved by the use of CGM (embodiment B) is reported. In particular, the post-meal correction boluses, that were administered at 09:00 and 15:00 as reaction to the CGM hyperglycemic alarms generated by CGM profile hyperglycemic threshold crossing, allow to reduce the time spent in hyperglycemia in embodiment B compared with embodiment A. For each virtual subject, the simulated BG profile was used to calculate, both in embodiments A and B, time-in-target, time-in-hypo and time-in-hyper, i.e. the percentage of samples falling respectively between 70 mg/dl and 180 mg/dl, under 70 mg/dl and over 180 mg/dl. In Table 1, the median values calculated for time-in-target, time-in-hypo and time-in-range in the 20 subjects are reported for embodiment A (second column) and B (third column). In embodiment B, compared with embodiment A, time-in-target significantly increases, from 60.04% to 69.05%, time-in-hyper significantly decreases, from 39.32% to 30.65%, while time-in-hypo slightly increases from 0.39% to 0.63%. To assess differences between metrics median values in embodiment A and B, a non-parametric paired sign test, whose p-values are reported in the fourth column of Table 1, was performed. With a 5% significance level, the improvement obtained in embodiment B for time-in-target and time-in-hyper is statistically significant, while the difference observed for time-in-hypo is not statistically significant.

TABLE 1

Metric
Embodiment A
Embodiment B
P-value

Time-in-target
60.04%
69.05%
0.0026

Time-in-hypo
0.39%
0.63%
0.7905

Time-in-hyper
39.32%
30.65%
0.0118

Table 1: Median values of time-in-target, time-in-hypo and time-in-hyper obtained for the glucose profiles simulated by model embodiments A and B are reported respectively in the second and third columns. P-values of a non-parametric paired sign test performed to compare metrics median values in embodiments A and B are reported in the fourth column.

2. Optimization of Guidelines for CGM-Driven Insulin Dosing

Currently, since CGM sensor has not been approved yet by regulatory agencies for insulin dosing, there are no officially approved guidelines for CGM driven insulin dosing. The DirectNet and the Juvenile Diabetes Research Foundation studies provide some non-officially approved guidelines to adjust insulin doses according to CGM trend. However, a recent survey revealed that patients use CGM trend to make significantly larger adjustment to insulin doses. The DM model can be an important tool to test and compare in silico candidate guidelines for CGM driven insulin dosing. For such a purpose, the DM model can be employed as represented in FIG. 5. Specifically, after defining a specific scenario, simulations can be run in a wide set of virtual patients by using a CGM configuration of the DM model in which block B simulates a certain CGM sensor device, and the T1D therapy model implements a set of candidate guidelines for CGM driven insulin dosing. These guidelines may include rules to adjust both insulin boluses and basal infusion rate according to CGM trend or prediction. The glycemic control achieved by each candidate guidelines can be assessed by a number of performance metrics in order to determine the safest rules that patients can use to calculate insulin doses on the basis of CGM readings. Candidate guidelines can be tested in the whole population of virtual patients or in a specific class of subjects (e.g. patients with low/high insulin sensitivity) to determine which are the best guidelines for each class of subjects.

3. Off-Line Insulin Therapy Optimization

The use of the DM model for off-line insulin therapy optimization is represented in FIG. 6. CGM and insulin pump data collected (e.g. in a week) for a specific patient can be employed to clone the specific patient's physiology i.e. identify the parameters of the T1D patient model that allow reproducing that patient glucose-insulin dynamics. The estimated parameters of patient's physiology are then put into the T1D patient glucose-insulin model to obtain an individualized patient model. Such a model is included in the DM model and used to run patient-specific simulations in order to optimize patient insulin therapy, thus to estimate optimal insulin therapy parameters (e.g. CR, CF and basal insulin) and individualize guidelines for CGM driven insulin dosing. More specifically, after defining a specific scenario, simulations are run iteratively in order to optimize block C parameters. At the first iteration, the patient current insulin therapy is implemented in block C. Then, at each iteration, therapy parameters are updated until optimal parameters are found and returned in output. At each iteration, to determine if therapy parameters need to be updated or not, the level of glycemic control is assessed by some metrics and some glycemic control criteria are checked. If the performance of the current iteration therapy parameters satisfies the desired criteria, then the algorithm stops and returns in output optimal therapy parameters. Conversely, if the performance of the current iteration therapy parameters does not satisfy the desired criteria, then therapy parameters are updated and another iteration is performed.

4. Development of Algorithms for On-Line Decision Support

The DM model can be used to develop and test algorithms for on-line CGM-driven decision support i.e. smartphone applications which exploit insulin pump and CGM information (e.g. CGM trend, predicted glucose, number of past hyper/hypoglycemic events etc.) and other information (e.g. meals and physical exercise) to provide in real-time to the patient therapeutic recommendations (e.g. recommended insulin dose, basal insulin changes etc.). A possible use of the DM model for such a purpose is represented in FIG. 7. A candidate decision support system (DSS), highlighted in yellow, is given in input information about the current scenario (e.g. meals and physical exercise) and CGM and insulin pump data simulated by the DM model. The DSS analyzes data and returns in output some recommendations about changes/adjustments to patient's therapy. When the patient receives the recommendations, e.g. through a smartphone notification if the decision support algorithm is implemented in a smartphone, he/she can decide to accept the advice and take the recommended action, ignore the advice or accept the advice and take an action similar to the recommended one (e.g. stronger/weaker). Patient's decision is modelled by the block patient decision in FIG. 7 that receives in input the DSS recommendations and returns in output the actual actions the patient takes which are then used to modify the therapy implemented in the block C of the DM model. Simulations with the whole model of FIG. 7 allow generation of BG profiles of patients who take their therapeutic decisions taking advantage of a particular DSS recommendations. A final evaluation of the level of glycemic control achieved permits evaluating the DSS performance and suggests if/how the decision support algorithm can be modified to improve its performance.

Methods and Systems

System 1. A system for modeling decision-making in managing type 1 diabetes, comprising: a type 1 diabetes therapy model; a type 1 diabetes patient glucose-insulin system model; a device for glucose model; wherein, when a continuous glucose monitoring driven therapy is simulated, continuous glucose monitoring data are used to generate, in addition to pre-meal boluses, post-meal correction boluses in response to hyperglycemic alarms, to generate hypo-treatments in response to hypoglycemic alarms, to correct insulin boluses for continuous glucose monitoring trend/prediction, and to correct basal insulin for current blood glucose value/trend/prediction.

System 2. The system of System 1, further comprising an insulin pump model, wherein the insulin pump model is configured to receive as input an insulin dose generated by the type 1 diabetes therapy model and configured to return as output an insulin pump infusion rate.

System 3. The system any one of Systems 1 to 2, wherein the type 1 diabetes patient glucose-insulin system model is configured to simulate the blood glucose and interstitial glucose concentration profiles in a type 1 diabetes patient resulting from carbohydrates intake and insulin pump infusions.

System 4. The system any one of Systems 1 to 3, wherein the type 1 diabetes patient glucose-insulin system model is configured to describes physiological events related to BG dynamics including gastrointestinal absorption of carbohydrates, insulin and glucagon kinetics and their action in regulating glucose endogenous production and utilization.

System 5. The system any one of Systems 1 to 4, wherein the type 1 diabetes patient glucose-insulin system model is configured to reproduce dynamics of a population of virtual patients, each patient represented by a set of model parameters, and to run multi-day simulations utilizing a description of inter-day and intra-day variability of insulin sensitivity.

System 6. The system any one of Systems 1 to 5, wherein the type 1 diabetes therapy model is configured to simulate therapeutic decisions that type 1 diabetes patients make on the basis of meal composition and blood glucose monitoring, including errors that patients make in diabetes management. Wherein the type 1 diabetes therapy model is configured to receive in input information about meal carbohydrate content and physical exercise, blood glucose measurements, and patient-specific therapy parameters, and wherein the type 1 diabetes therapy model is configured to return as output carbohydrate intake and insulin doses.

System 7. The system any one of Systems 1 to 6, wherein the type 1 diabetes therapy model is configured to simulate either a self-monitored blood glucose driven insulin therapy or a continuous glucose monitoring driven insulin therapy.

System 8. The system any one of Systems 1 to 7, wherein the device for glucose monitoring model is configured to generate glucose measurements used by patients in their diabetes management to adjust insulin doses or carbohydrate intake, and wherein the device for glucose monitoring model is configured to receive a glucose concentration as input and is configured to simulate self-monitored blood glucose spot measurements or a continuous glucose monitoring profile.

Method 9. A method for assessing whether a continuous glucose monitor is safe to employ in dosing insulin, the method accounting for intra-individual patient variability, comprising: running, using a first decision-making module, simulations of insulin therapy driven by self-monitored blood glucose data and thereafter computing metrics values using virtual patient blood glucose profiles to assess a level of glycemic control; running, using a second decision-making module, simulations of insulin therapy driven by continuous glucose monitoring data and thereafter computing metrics values using virtual patient blood glucose profiles to assess a level of glycemic control; comparing the metrics values to determine if the use of the continuous glucose monitor is at least as safe as the use of self-monitored blood glucose for insulin dosing.

Method 10. The method of Method 9, wherein the decision-making module utilizes a nonlinearized glucose-insulin model.

Method 11. The method of any one of Methods 9 to 10, wherein the decision-making module is configured to run multi-day simulations, whereby time-variability of insulin sensitivity is accounted for.

Method 12. The method of any one of Methods 9 to 11, wherein the decision-making module is configured to model a real patient's decision-making actions, taking into account possible errors made by patients in their diabetes management, wherein the errors include miscalculation of meal carbohydrate content, early insulin dose administration, delayed insulin dose administrations, and missed boluses occurrence.

Method 13. The method of any one of Methods 9 to 12, wherein the decision-making module receives a simulation scenario as model inputs, the model inputs comprising a sequence of meals carbohydrate content, information about physical exercise, and patient-specific parameters required for the calculation of a therapy recommendation, and wherein the decision making module outputs a simulator output comprising a patient glucose concentration.

Method 14. The method of any one of Methods 9 to 13, wherein the decision-making module utilizes models including: a glucose-insulin model of a type 1 diabetes patient, configured to receive as input carbohydrate intake and insulin pump infusion rate, and configured to return as output patient blood glucose and interstitial glucose profiles; a glucose monitoring model configured to simulate self-monitored blood glucose and continuous glucose monitor measurements utilizing the output of the glucose-insulin model; a type 1 diabetes therapy model configured to use meal carbohydrate content and self-monitored blood glucose and continuous glucose monitor measurements to simulate therapeutic decisions a patient makes in diabetes management to determine carbohydrate intake and insulin boluses; and an insulin pump model configured to simulate an insulin pump infusion rate.

Method 15. The method of any one of Methods 9 to 14, further comprising, if it is determined that the use of the continuous glucose monitor is at least as safe as the use of self-monitored blood glucose for insulin dosing, issuing insulin dosing instructions to an insulin pump by the continuous glucose monitor, whereby a patient receives an insulin dose.

Method 16. The method of any one of Methods 9 to 15, wherein the patient has type 1 diabetes, and wherein improved glycemic control in type 1 diabetes management for the patient is obtained.

Method 17. The method of any one of Methods 9 to 16, wherein the insulin dosing instructions are generated by: calculating, using a hypotreatment module, carbohydrate intake by adding to meal carbohydrate content hypotreatments, generated by a hypotreatment module; calculating, using a carb-counting module, a meal bolus before a meal to cover a meal carbohydrate intake; correcting, by the carb-counting module, a patient estimate of meal carbohydrate content by exploiting a model of patient carbohydrate counting error; calculating, using a meal bolus module, an insulin dose required to cover the meal using a patient carb ratio; and adding a correction insulin bolus to the meal bolus in order to take into account the patient's current glucose concentration.

Method 18. The method of any one of Methods 9 to 17, wherein a bolus time variability module calculates a variability in the pre-meal bolus administration time, whereby an occurrence of early/delayed bolus administration is simulated.

Method 19. The method of any one of Methods 9 to 18, wherein an occurrence of missed boluses is simulated by a missed bolus block.

Method 20. The method of any one of Methods 9 to 19, wherein the insulin dose is calculated by adding to a basal insulin rate insulin bolus, and wherein the insulin dose is corrected, using a correction for exercise module, by using information about physical exercise.

Method 21. A method for providing a real-time therapeutic recommendation for use in type 1 diabetes management in a patient, comprising: inputting, to a decision support module, data including continuous glucose monitoring data obtained from a glucose sensor, insulin pump data, and other diabetes management data; generating, based on the input data, a first therapeutic recommendation; displaying the therapeutic recommendation to a patient; inputting, to the decision support model, data indicative of an action taken by the patient in response to receiving the displayed therapeutic recommendation, the data including continuous glucose monitoring data obtained from the glucose sensor and insulin pump data; and generating, based on the input data and the data indicative of an action taken by the patient, a second therapeutic recommendation.

Method 22. The method of Method 21, wherein the continuous glucose monitoring data is selected from the group consisting of a glucose concentration trend, a predicted glucose concentration, a number of past hyperglycemic events, and a number of past hypoglycemic events.

Method 23. The method of any one of Methods 21 to 22, wherein the other diabetes management data is selected from the group consisting of meal data and physical exercise data.

Method 24. The method of any one of Methods 21 to 23, wherein the first therapeutic recommendation or the second therapeutic recommendation is selected from the group consisting of a recommended insulin dose and a basal insulin change.

Method 25. The method of any one of Methods 21 to 24, wherein the first or second therapeutic recommendation is delivered through a smartphone notification.

Method 26. The method of any one of Methods 21 to 25, wherein the decision support module compares the data indicative of an action taken by the patient responsive to the first therapeutic recommendation to determine if the patient has taken the therapeutic recommendation, has ignored the therapeutic recommendation, or has taken an action similar to the therapeutic recommendation.

Method 27. The method of any one of Methods 21 to 26, wherein the action taken by the patient is employed to modify a decision-making module configured to simulate real-life situations and everyday patient behaviors, wherein the decision-making module utilizes self-monitored blood glucose data measurement errors, continuous glucose monitoring data measurement errors, miscalculations of meal carbohydrate content, early insulin administration errors, delayed insulin administration errors, and missed insulin bolus errors.

Method 28. The method of any one of Methods 21 to 27, wherein the decision-making module utilizes patient data to assess glycemic control or efficacy of the decision support module, or to suggest modifications to the decision support module to improve its performance.

Method 29. A method for delivering a therapeutic recommendation for insulin delivery, comprising: collecting data from a patient over a period of time, the data including continuous glucose monitoring data obtained from a glucose sensor and insulin pump data; generating, based on the collected data, a model comprising estimated parameters of the patient's physiology that reproduces the patient's glucose-insulin dynamics; inputting the estimated parameters into a type 1 diabetes patient glucose-insulin model, whereby an individualized patient model is obtained; utilizing the individualized patient model in patient-specific simulations to estimate optimal insulin therapy parameters or individualized guidelines for continuous glucose monitoring-driven insulin dosing; and outputting the optimal insulin therapy parameters.

Method 30. The method of Method 29, wherein outputting the optimal insulin therapy parameters comprises displaying the therapeutic recommendation to a patient on a display.

Method 31. The method of any one of Methods 29 to 30, wherein the display is a smartphone display.

Method 32. The method of any one of Methods 29 to 31, wherein outputting the optimal insulin therapy parameters comprises issuing instructions to an insulin pump for delivering an insulin dose.

Method 33. The method of any one of Methods 29 to 32, wherein utilizing the individualized patient model in patient-specific simulations comprises iteratively running simulations, wherein at the first iteration, the patient's current insulin therapy is implemented, then, at each subsequent iteration, the therapy parameters are updated until it is determined that the therapy parameters satisfy predetermined criteria.

Method 34. The method of any one of Methods 29 to 33, wherein it is determined if the therapy parameters need to be updated by assessing a level of glycemic control.

Any combination of one or more computer readable medium(s) may be utilized in the methods, systems, aspects, and/or embodiments herein. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++, or the like, and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The program code may execute entirely on the computing unit.

It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.