Systems and methods for titrating basal insulin doses

Automatic adjustment of basal insulin dose through CGM technology and processor modules solves the problems of cumbersome and low blood sugar risks in the traditional insulin titration process, and achieves safer and more efficient insulin treatment.

CN118974840APending Publication Date: 2024-11-15DEXCOM INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202380030869.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-04-26
Filing Date
2023-04-26
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

When using basal insulin therapy in patients with type 2 diabetes, the traditional insulin titration process is cumbersome and accompanied by the risk of hypoglycemia, resulting in fear and discomfort in patients, and insufficient self-monitoring and hypoglycemia events are reported, affecting the accurate adjustment of insulin dose.

Method used

Continuous glucose monitoring (CGM) technology is used in combination with processor modules to generate daily blood glucose status parameters through historical data, automatically adjust basal insulin doses, reduce dependence on manual blood glucose measurements and patient reports, and provide safer and more effective dose recommendations.

Benefits of technology

Improves the safety and efficiency of insulin titration, reduces the risk of hypoglycemia, simplifies the patient monitoring and reporting process, and achieves faster and more accurate dose adjustments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118974840A_ABST
    Figure CN118974840A_ABST
Patent Text Reader

Abstract

A continuous glucose monitor (CGM) driven basal insulin titration system and method for type 2 diabetic patients can adapt to the needs and attention of subjects who just begin to receive basal insulin therapy. The method uses historical CGM data, base insulin dose information, hypoglycemia reports, and past recommendations as inputs and generates adjusted insulin doses along with reports suggested whether to continue or stop the titration process. The method enables new recommendations to be generated periodically (e.g., daily) until a sufficient, consistent dose size is determined.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Application Serial No. 63 / 335,176, filed on April 26, 2022, the contents of which are incorporated herein by reference. Background Art

[0003] Long-acting (basal) insulin therapy for patients with type 2 diabetes can be a daunting process for both patients and their caregivers. For some patients, insulin may be the first medication a patient injects, which can cause fear and discomfort. The titration process also traditionally involves daily finger-stick measurements of fasting blood glucose (BG), causing additional discomfort. Finally, and importantly, insulin is associated with the risk of hypoglycemia, which makes some people fear this necessary and beneficial treatment.

[0004] Clinically, providers must rely on patient compliance with fingerstick BG measurements to guide patients through the insulin titration process. Providers also rely on patients to self-report hypoglycemia. Analysis of data from observational studies suggests that some patients receiving basal insulin therapy may be hypoglycemic for prolonged periods of time. Patient self-monitoring and / or inadequate reporting of hypoglycemic events may also be a barrier to finding appropriate insulin dosing.

[0005] Conceptually, combining CGM with continuous glucose monitoring could address several unmet needs in the standard process of initiating insulin therapy for type 2 diabetes. CGM provides a robust and comprehensive measure of glucose control, which can be combined with computational technology to recommend faster and safer titration paths. CGM-based basal insulin titration also has the potential to reduce the burden on healthcare providers through automated deployment to patients.

[0006] Automated basal insulin titration based on CGM improves conventional manual titration based on self-monitoring of blood glucose (SMBG) in several ways. In contrast to relying on one or only a few finger stick samples, CGM provides a holistic view of the patient's blood glucose changes throughout the day, accurately characterizing the patient's exposure to both hypoglycemia and hyperglycemia. Conventional titration relies on the assumption that the patient will sample blood glucose in the fasting state, which cannot be independently verified and will lead to several potential dangers: (i) the patient's actual fasting blood glucose level may be lower than the sample value obtained by the patient, (ii) the patient's lowest blood glucose of the day (obtained at different times of the day) may be significantly lower than the sample value obtained by the patient, and (iii) misleading samples collected by the patient cause the conventional titration procedure to stay on an artificially high basal insulin dose, exposing the patient to unnecessary hypoglycemia risk. Automated basal insulin titration based on CGM does not rely on a single fasting blood sample reported by the patient every day, so all of these dangers can be avoided. In addition, CGM provides information about the nature of a patient's exposure to hyperglycemia, e.g., whether hyperglycemia is (i) persistent throughout the day (where increasing the patient's basal dose is an obvious choice) or (ii) fluctuates between two extremes (where increasing the patient's dose must be considered with caution). Indeed, if a patient is already experiencing significant hyperglycemia (e.g., as measured by A1c, mean glucose, or time above target range), as well as perceptible hypoglycemia (e.g., as measured by the number, duration, or depth of hypoglycemic events), then adjusting the patient's basal insulin may yield only marginal benefit, and it may be prudent to counsel the patient / HCP to consider other, more obvious modifications to therapy, such as the addition of prandial (rapid-acting) insulin to one or more meals per day. CGM provides a sufficiently rich signal to tease apart and decompose these different considerations, and automation removes the subjectivity of data analysis. Given the clear understanding of the statistical nature of a patient's glycemic variability, automated CGM-based insulin titration has the potential to achieve a safer and more effective final dose than conventional methods, and to reach this conclusion more quickly than using manual SMBG-based basal insulin titration. From a human factors perspective, automated basal insulin titration based on CGM minimizes the opportunity for user error by, for example, not requiring the patient to manually record SMBG samples or manually calculate dose changes. In addition, if the patient is using a connected or “smart” insulin pen, then the patient does not even have to record / report the basal insulin dose. Summary of the invention

[0007] In the first aspect, a basic titration adjustment device for titrating the basic insulin dose of a subject is proposed. The device includes one or more processors, a continuous glucose monitor (CGM) module, an insulin module, at least a first dose-daily blood glucose state response modeler, a target dose adjuster, and a dose finalizer. When the CGM module is executed by one or more processors, it receives CGM data obtained in at least a series of days, thereby generating a historical estimated glucose value (EGV), and then further generates each of the daily blood glucose state parameters from these EGV values, each of which reflects an aspect of the subject's blood glucose state. When the insulin module is executed by one or more processors, it receives the daily basic insulin dose value applied to the subject in this series of days, thereby generating a daily effective dose, each of which represents the total plasma insulin acting on the subject on the corresponding day, and explains the insulin accumulation produced by all previously applied doses in the subject. When the first dose-daily blood glucose state response modeler is executed by one or more processors, it receives the first daily blood glucose state parameter and the daily effective dose in the multiple daily blood glucose state parameters, thereby generating a first regularized dose-response model that associates the first daily blood glucose state parameter with the effective basic insulin dose. The target dose adjuster, when executed by one or more processors, receives a first regularized dose-response model, thereby generating a target basal insulin dose that will produce a value of a first daily glycemic state parameter predicted by the first regularized dose-response model at least with a specified confidence level. The dose finalizer, when executed by one or more processors, receives the target basal insulin dose and the value of one or more daily glycemic state parameters, thereby generating an adjusted basal insulin dose that modifies the target basal insulin dose according to one or more pre-established safety rules.

[0008] In one embodiment of the first aspect, the plurality of daily blood glucose status parameters are selected from the group consisting of an effective fasting blood glucose (EFBG) parameter, a hypoglycemia risk parameter representing the daily risk of experiencing low EGV, a low daily EGV parameter specifying whether the subject experiences an EGV below a specified level on a given day, and a daily reference blood glucose parameter reflecting a comparison of the daily EGV with a treatment target EGV.

[0009] In one embodiment of the first aspect, the first regularized dose-response model is a regularized linear regression model.

[0010] In one embodiment of the first aspect, the first regularized dose-response model is regularized using population data to bias at least the initial target dose to a safe dose at the population level.

[0011] In one embodiment of the first aspect, the first daily glycemic state parameter is a daily estimated fasting blood glucose (EFBG) parameter.

[0012] In one embodiment of the first aspect, the first daily glycemic status parameter is a hypoglycemia risk parameter.

[0013] In one embodiment of the first aspect, the one or more pre-established safety rules reduce the risk of hypoglycemia.

[0014] In one embodiment of the first aspect, the basal titration adjustment device also includes a reporting hypoglycemia module, which, when executed by one or more processors, receives hypoglycemia event data reported by the subject and generates a hypoglycemia report history from the hypoglycemia event data and the historical EGV, the hypoglycemia report history specifying the time and severity of credible hypoglycemia events reported by the subject, wherein the dose finalizer, when executed by the one or more processors, uses the hypoglycemia report history as an additional input to generate an adjusted basal insulin dose.

[0015] In one embodiment of the first aspect, the basal titration adjustment device also includes a termination checker module, which, when executed by one or more processors, receives values ​​of some or all of a plurality of daily blood glucose state parameters, a history of hypoglycemia reports, and adjusted basal insulin doses generated in previous days, thereby generating a termination report indicating whether the titration of basal insulin should be terminated or continued.

[0016] In one embodiment of the first aspect, at least the first dose-daily blood glucose state response modeler includes a second dose-daily blood glucose state response modeler, which, when executed by one or more processors, receives a second daily blood glucose state parameter among multiple daily blood glucose state parameters and a daily effective dose, thereby generating a second regularized dose-response model that associates the second daily blood glucose state parameter with the effective basal insulin dose, and the target dose adjuster, when executed by one or more processors, receives the first regularized dose-response model and the second regularized dose-response model, thereby generating a target basal insulin dose.

[0017] In one embodiment of the first aspect, the first daily glycemic state parameter is a daily estimated fasting blood glucose parameter, and the second daily glycemic state parameter is a hypoglycemia risk parameter specifying the risk of the subject's EGV falling below a threshold one or more times in a day.

[0018] In an embodiment of the first aspect, the first regularized linear regression model comprises values ​​for a slope, an intercept, and an uncertainty measure.

[0019] In an embodiment of the first aspect, the value of the EFBG parameter is based on CGM data obtained at a specified time of day.

[0020] In one embodiment of the first aspect, the value of the EFBG parameter is based on a CGM value that is lower than a particular percentile of the subject's blood glucose distribution.

[0021] In one embodiment of the first aspect, a hypoglycemic event is considered unrelated if it occurred more than a specified period of time before the current time, or if the hypoglycemic event occurred when the current adjusted basal insulin dose at that time differed from the current adjusted basal insulin dose by more than a specified amount.

[0022] In one embodiment of the first aspect, the first dose-daily glycemic state response modeler generates a basal insulin dose-hypoglycemia risk regression model that predicts the risk of hypoglycemia corresponding to an effective basal insulin dose.

[0023] In one embodiment of the first aspect, the basal insulin dose-hypoglycemia risk regression model is a linear regularized linear regression model.

[0024] In one embodiment of the first aspect, the linear regularized linear regression model is regularized using population data to at least bias the initial target dose to a safe dose at the population level.

[0025] In one embodiment of the first aspect, the hypoglycemia risk parameter indicates the risk of the subject's EGV falling below a threshold one or more times in a day.

[0026] In one embodiment of the first aspect, the hypoglycemia risk parameter represents a daily EGV that is below a specified percentile of all EGVs of the subject in a day.

[0027] In an embodiment of the first aspect, the dose finalizer, when executed by the one or more processors, periodically generates an adjusted basal insulin dose.

[0028] In one embodiment of the first aspect, the periodic period is daily.

[0029] In an embodiment of the first aspect, the target dose adjuster, when executed by one or more processors, determines a first preliminary target dose from a first regularized dose-response model and a second preliminary target dose from a second regularized dose-response model, and generates a target dose by balancing the first preliminary target dose with the second preliminary target dose.

[0030] In one embodiment of the first aspect, the generated target dose is the minimum of the first preliminary target dose and the second preliminary target dose, the maximum of the first preliminary target dose and the second preliminary target dose, or a weighted average of the first preliminary target dose and the second preliminary target dose.

[0031] In one embodiment of the first aspect, the one or more pre-established safety rules prevent the value of the newly generated adjusted basal insulin dose from exceeding a previously generated adjusted basal insulin dose of a previous day.

[0032] In one embodiment of the first aspect, one or more pre-established safety rules establish values ​​for design parameters defining an EGV target deadband around a target EGV such that for values ​​of a daily reference blood glucose parameter within the target deadband, a newly generated adjusted basal insulin dose will not be adjusted from a previously generated adjusted basal insulin dose.

[0033] In one embodiment of the first aspect, one or more pre-established safety rules generate a warning predicting that excessive basal insulinization will occur.

[0034] In an embodiment of the first aspect, one or more pre-established safety rules constrain modification of the target basal insulin dose if excessive hyperinsulinization is predicted to occur.

[0035] In one embodiment of the first aspect, the termination report indicates that titration of the basal insulin dose should be terminated if the target value of a daily reference blood glucose parameter cannot be achieved without introducing prandial insulin, and the daily reference blood glucose parameter reflects the comparison results of the daily EGV with the treatment target EGV.

[0036] In a second aspect, a method for titrating a basal insulin dose of a subject is provided. The method includes: receiving CGM data obtained over at least a series of days, thereby generating a historical estimated glucose value (EGV); generating each of a plurality of blood glucose state parameters from the EGV values, each of which reflects an aspect of the subject's blood glucose state; receiving the values ​​of the basal insulin doses administered to the subject over the series of days, thereby generating an effective dose, each of which represents the total plasma insulin acting on the subject on the corresponding day and accounts for the accumulation of insulin produced by all previously administered doses in the subject; receiving a first blood glucose state parameter and the effective dose from a plurality of blood glucose state parameters, thereby generating a first regularized dose-response model associating the first blood glucose state parameter with the effective basal insulin dose; receiving the first regularized dose-response model, thereby generating a target basal insulin dose, the target basal insulin dose will produce a value of the first blood glucose state parameter predicted by the first regularized dose-response model at least with a specified confidence level; and receiving the target basal insulin dose and the values ​​of one or more blood glucose state parameters from the plurality of blood glucose state parameters, thereby generating an adjusted basal insulin dose, the adjusted basal insulin dose modifying the target basal insulin dose according to one or more pre-established safety rules.

[0037] In one embodiment of the second aspect, the basal insulin dose is administered daily.

[0038] In one embodiment of the second aspect, the adjusted basal insulin dose is produced daily.

[0039] This summary is provided to introduce a series of concepts in a simplified form, which are further described in the detailed description below. The summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to assist in determining the scope of the claimed subject matter. Furthermore, the claimed subject matter is not limited to implementations that solve any or all of the disadvantages mentioned in any part of this disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a functional block diagram of an example of a basal insulin titration system.

[0041] Figure 2 Shows Figure 1 An example of a CGM module is shown.

[0042] Figure 3 Shows Figure 1 An example of an insulin module is shown.

[0043] Figure 4 Shows Figure 1An example of a module reporting hypoglycemia is shown.

[0044] Figure 5 Shows Figure 1 An example of the Dose-EFBG modeler is shown.

[0045] Figure 6 Shows Figure 1 An example of the Dose-LBGR modeler is shown.

[0046] Figure 7 Shows Figure 1 An example of a target dose adjuster is shown.

[0047] Figure 8 Shows Figure 1 An example of a dose finalizer is shown.

[0048] Fig. 9 Shows Figure 1 An example of a termination checker is shown.

[0049] Fig.10 One example of an illustrative operating environment in which the basic titration adjustment systems and methods described herein may operate is shown.

[0050] Fig.11 An example of a computing device is shown in which example embodiments and aspects may be implemented. DETAILED DESCRIPTION

[0051] This paper describes a CGM-driven basal insulin titration system and method for type 2 diabetes patients. The system and method can be adapted to the needs and concerns of a subject who has just begun to receive basal insulin therapy. As explained in more detail below, the method uses historical CGM, basal insulin dosage information, hypoglycemia report and past recommendations as input, and generates adjusted insulin dosage, together with a report of whether to continue the titration process or stop the titration process. The method can regularly (e.g., every day) generate new recommendations until sufficient and consistent dosage sizes are determined. Some exemplary features and advantages of the system and method are as follows.

[0052] The system and method can make model-based personalized dosage recommendations by estimating a dose-response model for the individual from CGM-derived glucose control measures (including glucose value percentiles and estimated fasting glucose). Regularization functions are added to the model for the individual to bias the fit toward safe recommendations in the first few days, when CGM history is still sparse. CGM-based glucose variability estimates are incorporated into the algorithm design to guard against dose sizes that may increase the risk of hypoglycemia. Finally, the system takes into account the time-delayed response of long-acting insulin, referred to herein as the effective dose of insulin. The effective dose model is population-based (not individual-based).

[0053] The system and method also incorporate checks to ensure that the dosage recommendation is both safe and meaningful. For example, a check can ensure that the dose is not increased after a recent non-severe hypoglycemia. The system can issue an alert when clinical guidelines indicate that the maximum per kilogram dose has been reached. In addition, the reference glucose BG is estimated daily. ref , forcing dose changes to be consistent with estimates relative to the target value (e.g., when BG ref Dose reduction is not allowed above the target range, and when BG ref The system can also force dose reductions in the presence of severe hypoglycemia detected by the CGM and reported by the patient.

[0054] System Overview

[0055] Figure 1 It is a functional block diagram of an example of the basal insulin titration system 100 described herein. In general, the system is used to read historical data, including CGM history, insulin administration history, reported hypoglycemia (Hypo) history and recommended history, and generate both adjusted doses and termination reports. CGM history refers to the estimated glucose value with a time stamp collected since the start of the titration process. Insulin administration history refers to the basal (long-acting) insulin administration injection with a time stamp since the start of the titration process, wherein the time stamp indicates when it is administered (e.g., injected). The reported hypoglycemia history refers to the reported experience of symptomatic hypoglycemia reported by patients, healthcare practitioners (HCP), informal care providers and / or other related individuals, wherein each report indicates the level of event severity. Recommended history refers to the history of previously adjusted doses during the titration process.

[0056] At a high level, the system is used to (i) interpret CGM history and insulin dosing history data in the form of one or more dose-response models from which to derive a target dose, and (ii) consider daily fasting glucose values, hypoglycemia exposure, and past recommendations to calculate an adjusted dose and recommend next steps in the form of a discontinuation report.

[0057] The flow of data through the various components or modules of the system is as follows.

[0058] See also Figure 1 The CGM history data is read into the CGM module 110 (see CGM module section below), which outputs a sequence of the following items: (i) daily estimated fasting blood glucose (EFBG), {EFBG d} d=1,…,t , where d = 1 corresponds to the first day of the titration process, d = t corresponds to the EFBG value of the day, (ii) daily hypoglycemia risk (LBGR), {LBGR d} d=1,…,t , (iii) daily targets for non-severe hypoglycemia (e.g., less than 70 mg / dl), {I <70,d} d=1,…,t , where I d = "True" means experiencing non-severe hypoglycemia on day d, (iv) daily indicator of severe hypoglycemia (e.g., less than 54 mg / dl), {I <54,d} d=1,…,t , where I d = "True" means experiencing severe hypoglycemia on day d, and (v) reference blood glucose value BG ref , the dose finalizer is used to generate glucose statistics for the adjusted dose.

[0059] The insulin administration history data is read into the insulin module 120 (see Insulin Module section below), which generates a sequence of effective doses (ED) {ED d} d=1,…,t-1 , where t-1 corresponds to the most recent effective dose measured by CGM, i.e. the dose acting during this day of the patient's life, and is presented as EFBG t LBGR t ,I <70,t ,I <54,t and BG ref .

[0060] The reported hypoglycemia history and CGM history data are read into the reported hypoglycemia module 130 (see the reported hypoglycemia module section below), which generates a sequence of daily reported hypoglycemia markers {H d} d=1,…,t , where H d Takes different values ​​depending on whether the patient or caregiver reports severe or non-severe symptomatic hypoglycemia on day d.

[0061] Daily estimated fasting blood glucose data {EFBG d} d=1,…,tand effective dose data {ED d} d=1,…,t-1 is read into the dose-EFBG modeler 140 (see the dose-EFBG modeler section below), which estimates the response of the ED to EFBG. In one implementation of the system, the dose response is approximated as with a slope C efbg , intercept I efbg , residual standard deviation s efbg and the fitting quality r 2 efbg linear relationship.

[0062] Daily hypoglycemia risk data {LBGR d} d=1,…,t and effective dose data {ED d} d=1,…,t-1 is read into the Dose-LBGR Modeler 150 (see Dose-EFBG Modeler section below), which estimates the response of the ED to the LBGR. In one implementation of the system, the dose response is approximated as with a slope C lbgr , intercept I lbgr , residual standard deviation s lbgr and the fitting quality r 2 lbgr linear relationship.

[0063] The models produced by the Dose-EFBG Modeler 140 and the Dose-LBGR Modeler 150 are read into the Target Dose Adjuster 160 (see Target Dose Adjuster section below), which produces an adjusted target dose that balances the need for fasting glucose within the target range with the risk of hypoglycemia.

[0064] The adjusted dose is generated by the dose finalizer 170 (see Dose Finalizer section below), which works based on: the target dose (from the target dose adjuster 160), together with (i) a reference glucose BG ref , (ii) non-severe hypoglycemia indicators {I <70,d} d=1,…,t and severe hypoglycemia index {I <54,d} d=1,…,t , (iii) daily reported hypoglycemia {H d} d=1,…,t , (iv) past dose recommendations recorded in the recommendation history, and possibly (v) other algorithm state parameters from previous titration recommendations. The dose finalizer 170 imposes constraints to ensure that the final adjusted dose is safe and matches the patient's expectations based on their previous day's blood glucose experience.

[0065] Finally, the outputs of the CGM module, insulin module, and report hypoglycemia module, along with the recommendation history, are read into the termination checker 180 (see the termination checker section below), which generates a termination report that explains whether the titration process should be terminated or continued, and also makes other subsequent recommendations, such as, when appropriate, suggesting that prandial (rapid-acting) insulin should be added to the patient's treatment regimen.

[0066] In summary, the basal insulin titration system described herein reads in CGM history, insulin dosing history, reported hypoglycemia history, and recommendation history (eg, daily) to generate dosage recommendations and termination reports.

[0067] We will now describe Figure 1 Several alternative embodiments of the high-level system are shown.

[0068] For example, different implementations may handle historical data in different ways. For example, in some cases, the entire history may be entered into the system each day, or alternatively, only differential inputs describing new data since the last recommendation may be entered, with most of the history represented as an evolving "state" of the system. Similarly, CGM data can be obtained in different ways, such as by reading the CGM data directly from a CGM device, or from a database query from a server. Similarly, insulin dosing data can be obtained in different ways, such as by reading in from a smart pen device. Alternatively, insulin dosing data can be read in from the dosage reported by the patient / caregiver in a smart phone or web application ("app").

[0069] Different specific implementations of the system can also be used to titrate different types of basal (long-acting) insulins. For example, the basal insulin that is titrated can be insulin glargine. Alternatively, the basal insulin can be other preparations, including preparations with longer pharmacokinetic profiles (e.g., insulin degludec, insulin icodec) or preparations with shorter pharmacokinetic profiles (e.g., NPH).

[0070] The reported hypoglycemia data can be obtained from the patient and / or caregiver in different ways. For example, reported hypoglycemia events can be obtained from data collected in a smartphone "app" or collected via a web "app" or "portal". The adjusted dose data can be stored in different ways, such as in a smartphone "app" or in a database accessible via a data network.

[0071] The titration method described herein can be turned into a product in different ways. For example, the method can be implemented in different places. For example, the adjusted dosage value and termination report can be calculated by a server accessing the CGM history, insulin administration history, reported hypoglycemia history, and recommendation history. Alternatively, the adjusted dosage value and termination report can be calculated by a smart phone or other portable device accessing the CGM history, insulin administration history, reported hypoglycemia history, and recommendation history. Similarly, the method can be implemented at different times, such as on demand, or can be pre-calculated and transmitted to the patient and / or caregiver according to a predetermined schedule. The following will be combined with Fig.10 An example of an environment in which a titration method may be implemented is described below.

[0072] The outputs provided by the system and method may also be provided to different parties. For example, adjusted dosage values ​​and termination reports may be viewed and implemented directly by the patient, while the caregiver has the ability to monitor the treatment remotely. Alternatively, adjusted dosage values ​​and termination reports may first be provided to the caregiver (HCP, GP, CDE), who has the ability to approve or reject the recommendations before they are sent to the patient.

[0073] Different termination criteria may be used by the termination checker 180. For example, the algorithm may be configured to minimize the time required to reach the optimal abstract daily dose or to conclude that basal insulin alone is insufficient to meet the patient's needs. Alternatively, the termination checker 180 may be configured to operate in a continuous manner, continuously adjusting the patient's insulin dose in response to slowly changing insulin needs, and terminate only when it concludes that basal insulin alone is insufficient to meet the patient's needs.

[0074] The specific output from the CGM module 110 (generally referred to herein as the daily blood glucose state parameter) may be different in different implementations and can be determined in a variety of different ways. For example, hypoglycemia risk (LBGR) can be quantified in different ways (see the CGM module section below). In an illustrative example, the LBGR can be quantified as the percentage of time below a threshold glucose level, such as the percentage of time below 70 mg / dl (or 3.9 mmol / l), or can be quantified as the glucose level associated with the Xth percentile of the daily CGM value (expressed in mg / dl or ml / l or otherwise), where X may include 0.5, 1, 2, 10 or other percentiles. In other examples, the LBGR can be calculated as a risk space value as described in U.S. Patent No. 11,355,238, or as a hypoglycemic index (LBGI) as described in U.S. Patent No. 9,317,657. Alternatively, the LBGR can be calculated as an operational hypoglycemia risk value as described in U.S. Patent Publication No. 20200178905A1.

[0075] Reference blood glucose (BG ref ) is another daily blood glucose status parameter that can also be quantified in different ways (see CGM Modules section below). For example, BG ref This may be quantified as the estimated fasting blood glucose (EFBG) mentioned above or as the time-averaged BG calculated from the CGM, or any other BG statistic that characterizes the overall quality of glycemic control.

[0076] In some cases, the calculation of the target dose can use different dose response models (see the target dose adjuster section below). For example, the target dose can be calculated using only the dose-LBGR modeler without the dose-EFBG modeler, or vice versa. In other embodiments, different dose-response models can be used, including nonlinear dose-response models and dose-response models that characterize model uncertainty in different ways (except for residual standard deviation and fit quality).

[0077] The systems and methods described herein can be adapted to different titration scenarios without changing their basic structure. For example, if the patient is unfamiliar with basal insulin (never used it before or is not currently using it), the parameters of the titrator can be configured to (i) more quickly learn the dose response from an initial dose that is otherwise the same as but lower than that situation, (ii) particularly avoid hypoglycemia, which is often a concern for new insulin users, and (iii) appropriately cautiously conclude that basal insulin adjustments alone are insufficient to meet the patient's needs. On the other hand, if the patient is already using basal insulin and is seeking to identify a better daily dose, the titrator can utilize previously available blood glucose and insulin dosing data, and the parameters of the algorithm can be adjusted so that (i) the dose response near the patient's current prescribed dose is appropriately understood, (ii) hypoglycemia is not too concerned (because insulin is the patient's daily medication by now), and (iii) the conclusion that basal insulin alone is no longer sufficient can be explored more actively. In other embodiments, the titrator can be configured to continuously adjust the patient's basal insulin dose, tracking the patient's changing need for basal insulin in response to: (i) seasonal effects, (ii) changes in diet / lifestyle, and (iii) changes in underlying physiological function.

[0078] The following will describe in more detail Figure 1 Exemplary embodiments of the various components and modules shown, along with some alternatives thereof.

[0079] CGM Module

[0080] like Figure 2As shown, the CGM module 110 reads in the CGM history and then presents the values ​​of a series of daily blood glucose status parameters. In one specific embodiment described herein, the values ​​of five such parameters are output:

[0081] 1.{EFBG d} : Estimated fasting blood glucose (EFBG) history, a sequence of positive values ​​every d-th day.

[0082] 2. {LBGR d} : Hypoglycemia risk (LBGR) history, a sequence with a non-negative value every d-th day.

[0083] 3. {I <70,d}: Below 70 time indicator history, a sequence of true or false values ​​for each dth day, indicating that the patient's blood sugar was below 70 mg / dL (true) or not below 70 mg / dL for most of the time on the dth day

[0084] (Fake).

[0085] 4. {I <54,d}: Below 54 time indicator history, a sequence of true or false values ​​for each dth day, indicating that the patient's blood sugar was below 54 mg / dL (true) or not below 54 mg / dL for most of the time on the dth day

[0086] (Fake).

[0087] 5. {BG ref}: Reference blood glucose value, which drives the adjusted dose toward achieving a BG within a specified target range ref changes in dosage.

[0088] like Figure 2 As shown, the CGM module 110 includes reading the CGM history and generating the EGV history {EGV1, EGV2, ..., EGV n}EGV parser 200. Each element of the EGV history EGV i Each includes a timestamp and an estimated blood glucose value (mg / dL). The elements of the EGV history are sorted by timestamp and do not repeat each other. The EGV history is provided as input to each of the five submodules of the CGM module corresponding to the five outputs. The main steps performed by each submodule are described below.

[0089] Fasting blood glucose estimator 210 (output {EFBG d}) Use EGV history to estimate the patient's daily fasting blood glucose. Fasting blood glucose corresponds to a physiological state without short-term glucose disturbances caused by meals, exercise, illness, and other events. This physiological state is intended to be similar to the state formed by a standard fast before a laboratory blood test. EFBGd Values ​​are intended to reflect the treatment effect of the previous adjusted dose. The previous adjusted high dose was associated with a smaller EFBG relative to the previous adjusted low dose. d Value associated.

[0090] Hypoglycemia risk estimator 212 (output {LBGR d}) Use EGV history to estimate the patient's daily risk of hypoglycemia. High (vs. low) LBGR d A value of d indicates that a confluence of factors (eg, insulin sensitivity, meals and exercise, and doses of basal insulin and / or other medications) results in an increased risk for the patient to experience low blood sugar (eg, hypoglycemia) on day d.

[0091] Below 70 time identifier 214 (output {I <70,d}) Use the EGV history to identify if the patient has had blood glucose below 70 mg / dL for most of the day on day d. A single EGV below 70 mg / dL i May be the result of a transient CGM error or may be isolated and not be therapy related. Below 70 Time Identifier isolates therapy related EGV below 70 mg / dL i value, and use I <70,d All other days accept I <70,d The false value.

[0092] Below 54 time identifier 216 (output {I <54,d}) Use the EGV history to identify if the patient has had blood glucose below 54 mg / dL for most of the day on day d. A single EGV below 54 mg / dL i May be the result of a transient CGM error or may be isolated and not be therapy related. Below 54 time identifier isolates therapy related EGV below 54 mg / dL i value, and use I <54,d All other days accept I <54,d The false value.

[0093] Reference blood glucose calculator 218 (output BG ref ) uses the EGV history to calculate a reference blood glucose value. This reference value characterizes the quality of the previous adjusted dose compared to the treatment target. A reference blood glucose value above the treatment target on day d indicates that the adjusted dose may need to be increased on day d+1 (see Dose Finalizer for details).

[0094] Some alternative implementations of the various submodules of the CGM module 110 are described below.

[0095] In some cases, the fasting blood glucose estimator 210 may be configured to d The output is:

[0096] a. EGV historical statistics on day d regarding the patient's specific circadian rhythm of sleep and eating. For example, EFBG d It can be the average EGV of the patient around the time of waking up and before the first meal of the day on day d. i .

[0097] b. Historical statistical values ​​of EGV on day d as specified percentiles known to be associated with the fasting physiological state.

[0098] c. Combining several statistical values ​​(eg, 1a and 1b) with a weighting scheme (eg, a linear function) to generate an estimate of a statistical value known to be associated with a fasting physiological state.

[0099] In some cases, the hypoglycemia risk estimator 212 may be configured to d The output is:

[0100] a. A percentile (e.g., 1st percentile) of the EGV history on day d. In other words, the percentile characterizes the fraction of time that the patient's blood glucose is below a certain level (mg / dL). For example, if the 1st percentile of the EGV history on day d is 68 mg / dL, then the patient has 1% of their time with blood glucose below 68 mg / dL on day d. Their LBGR 1st percentile is 68. A larger or smaller percentile can be selected to modify the severity of the hypoglycemia risk assessed by the algorithm.

[0101] b. Low Blood Glycemic Index (LBGI) using the risk space approach described in US Pat. No. 11,355,238 or US Pat. No. 9,317,657.

[0102] c. An actionable hypoglycemia risk value as described in U.S. Patent Application No. 20200178905.

[0103] d. The left side (negative value) of the average daily risk range recorded by Patton and Clements (see Patton SR, Clements MA. Average daily risk range as a measure for clinical research and routine care. J Diabetes Sci Technol. 2013 Sep 1; 7(5): 1370-5. doi:

[0104] 10.1177 / 193229681300700529.PMID: 24124966; PMCID: PMC3876383).

[0105] In some cases, the time identifier 214 may be less than 70 seconds. <70,d The output is:

[0106] a. True for any d-th day where several (e.g., 3) consecutive elements of the EGV history are below 70 mg / dL. False for all other days.

[0107] b. Alternatively, I <70,d Some other quantitative assessment of exposure to non-severe hypoglycemia may be calculated.

[0108] In some cases, the time identifier 216 below 54 may be {I <54,d The output is:

[0109] a. True for any d-th day where several (e.g., 3) consecutive elements of the EGV history are below 54 mg / dL. False for all other days.

[0110] b. Alternatively, I <54,d Some other quantitative assessment of exposure to non-severe hypoglycemia may be calculated.

[0111] In some cases, the reference blood glucose calculator 218 may be configured to calculate BG ref The output is:

[0112] a. Will show BG ref EFBG of day d d This alternative will drive the adjusted dose toward the dose that achieves the specified target fasting glucose.

[0113] b. BG will be displayed ref LBGR of day d d This alternative will drive the adjusted dose toward the dose that achieves the specified hypoglycemia risk.

[0114] Insulin module

[0115] like Figure 3 As shown, the insulin module 120 receives or reads in the recorded dose history during the titration process and then presents a sequence of daily effective dose history values ​​that illustrate the pharmacokinetic cumulative effect of available insulin in plasma over time. The daily effective dose history is then used as an independent variable in the effective dose-response generated by the dose-EFBG modeler and the dose-LBGR modeler.

[0116] The insulin module 120 performs two main steps. First, the daily dose parser 230 interprets the recorded dose history {RD1, RD2, ..., RD tr} and generate daily dose history {DD1,DD2,…,DD t-1},like Figure 3 Here, the recorded dose history RD r Each element of (for r = 1, 2, ..., tr) is the rth recorded (long-acting) insulin dose, described by the recorded dose value and timestamp. For a daily dose history, each element at a specific timestamp is associated with an abstract titration day, such that DD d (for d=1, 2, ..., t-1) is the historical total basal dose received on day d of the titration process, where t-1 refers to the most recent injection before the injection that the system will recommend.

[0117] Next, the dose accumulator 240 processes the daily dose history to generate an associated daily effective dose history {ED1, ED2, ..., ED t-1}, where ED d (for d = 1, 2, ..., t-1) is the effective dose for the patient on the d-th titration day, representing the total available (effective) plasma insulin acting on the patient on that day. The effective dose (i) takes into account the accumulation of insulin produced in the patient's body by all previously administered doses, (ii) is suitable for long-acting insulins with pharmacokinetic profiles measured in days (from half a day to 30 days), (iii) is used to quantify the relative effects of past insulin doses within a specific time window (e.g., once a day), and (iv) can be used in various patient-specific dose-response models.

[0118] Some alternative implementations of the insulin module 120 are described below.

[0119] In some cases, elements of the recorded dose history may be collected via a "connected" or "smart" insulin pen, bolus doses reported by the user via an "app," or an insulin injection detection algorithm (e.g., based on a CGM, or any combination of devices).

[0120] The time window in which the insulin administration history can be obtained can vary. For example, in the case of 2 injections per day, the time window can be 12 hours, the time window for daily injections can be 1 day (as described in the above examples), or the time window for weekly injections can be 1 week.

[0121] The total basal dose can be calculated by the daily dose parser 230 as the result of a single injection of basal (long-acting) insulin on day d, the sum of multiple doses on day d, or the weighted sum of multiple doses on day d (weighted by a time degradation coefficient - so that the earlier in the past a dose is injected, the lower the weight).

[0122] In some cases, calculation of the effective dose can be based on a linear model of the daily dose effect on day d starting from the most recent N days, which is derived from a PK / PD model, derived from patient-specific or population data, or learned online or determined in advance.

[0123] Report Hypoglycemia (Hypo) Module

[0124] like Figure 4 As shown, the report hypoglycemia module 130 receives or otherwise reads the CGM history and the reported hypoglycemia history, and then presents the hypoglycemia event history. The patient generates each record in the reported hypoglycemia history, so these records may contain certain defects. For example, these records may involve hypoglycemia events in the distant past, or they may report hypoglycemia events when hypoglycemia events do not actually occur physiologically. The report hypoglycemia module 130 filters the reported hypoglycemia history down into the hypoglycemia event history, wherein each hypoglycemia event is both relevant and credible for the next adjusted dose. Hypoglycemia is the main risk of basal insulin therapy, and these hypoglycemia events help other modules of the system to mitigate adverse hypoglycemia attacks.

[0125] like Figure 4 As shown, the report hypoglycemia module 130 includes: (1) a CGM history parser 250, which reads the CGM history and generates an EGV history {EGV1, EGV2, ..., EGV k}, (2) a hypoglycemia report parser 260, which reads the reported hypoglycemia history and generates a hypoglycemia report history {HR1, HR2, ..., HR n}, and (3) a hypoglycemia report filter 270 that combines the EGV history and the hypoglycemia report history to present a hypoglycemia event history {H1, H2, ..., H m}. Each element of the hypoglycemia report history HR iThe timestamps of hypoglycemic events, timestamps of hypoglycemic report submissions, and indicators of the severity of hypoglycemic events are included. The elements of the hypoglycemic report history are sorted by the timestamps of the hypoglycemic events and do not repeat each other. The hypoglycemic report history is provided as input to the hypoglycemic report filter 270, which determines whether each HR is sufficiently relevant and credible. All hypoglycemic reports that are sufficiently relevant and credible pass through the hypoglycemic report filter 270 and become part of the hypoglycemic event history, and its length can vary in the range of from empty (i.e., no hypoglycemic events) to the length of the hypoglycemic report history (i.e., all hypoglycemic reports are retained as hypoglycemic events).

[0126] In some embodiments, it may be found that the relevance of the hypoglycemia report is irrelevant. For example, if the age of the hypoglycemia report (relative to the current time) is greater than a pre-specified threshold (e.g., 7 days), the hypoglycemia report may be considered irrelevant. The threshold will determine when the hypoglycemia event loses its relevance to the next adjusted dose. If a hypoglycemia event is experienced when the adjusted dose is significantly different from the current adjusted dose, the hypoglycemia report may also be considered irrelevant. For example, if a hypoglycemia event is experienced when the adjusted dose is 37U / day, and the current dose is only 10U / day, the hypoglycemia report may be considered irrelevant to the next adjusted dose. If the severity of the hypoglycemia report is low enough, it may also be considered irrelevant.

[0127] In some embodiments, hypoglycemia report may be considered as unreliable. For example, if the age of hypoglycemia report (relative to current time) is greater than a pre-specified threshold value (e.g., 2 days), then hypoglycemia report may be considered as unreliable. For example, this threshold value will determine when hypoglycemia report loses credibility due to memory recall limitations. If EGV history is credible and continuously above a specific threshold value (e.g., 130mg / dL) when hypoglycemia event occurs, then hypoglycemia report may also be considered as unreliable, so hypoglycemia event is unlikely to be effective. For example, the patient may enter the wrong date for hypoglycemia event, and EGV history can pick out such an error.

[0128] Dose-EFBG Modeler

[0129] like Figure 5 As shown, the dose-EFBG modeler 140 receives the estimated fasting blood glucose value history and the insulin effective dose history. The modeler outputs a dose-EFBG regression model that predicts the estimated fasting blood glucose corresponding to the effective dose. In addition to predicting the estimated fasting blood glucose, the dose-EFBG model can also include an assessment of how the prediction can change.

[0130] Figure 5The input and output of the dose-EFBG modeler 140 generated during day t are shown. The dose-EFBG modeler 140 receives the following two input streams on day t:

[0131] 1. Daily history of estimated fasting blood glucose values ​​generated by the CGM module, for days d=2,…,t, labeled EFBG d .

[0132] 2. Daily history of estimated effective insulin dose generated by the insulin module, for day d, where day d = 1, 2, ..., t-1, labeled ED d .

[0133] In one embodiment, based on these inputs, the effective dose-EFBG response modeler 280 generates a linear regression model with an intercept I efbg and coefficient C efbg , used to relate estimated fasting blood glucose to effective insulin dose.

[0134] To bias the regression model towards a safe population-averaged dose response early in the titration process, the linear regression objective was modified by adding a regularization term as follows:

[0135]

[0136] This equation is based on the coefficient C efbg The mean value is μ C And the standard deviation is σ C The normal distribution population model is based on C efbg The effective dose-EFBG response modeler 140 also generates a goodness of fit r for the regression model. 2 efbg assessment.

[0137] The effective dose-EFBG response uncertainty modeler 290 calculates the variance of the residuals produced by the effective dose-EFBG response modeler. It also applies a constant learning rule to update its efbg Estimate the variance of .

[0138] In one embodiment, an "estimated fasting glucose" is estimated from CGM data by looking at the estimated glucose value at a particular time of day (e.g., about 5:30 AM). An alternative embodiment can take into account potential early meals by selecting the last CGM reading before a detected meal. In some implementations, an EFBG alternative that is more descriptive of the patient's steady state can be used. These alternatives can include the median or another percentile of the patient's glucose profile or a glucose profile model.

[0139] Various embodiments of the effective dose-EFBG response modeler 280 may employ nonlinear regression models, machine learning methods (including neural networks), and may incorporate additional predictors in addition to the effective dose. Additionally, the population model used in the regularization may be multivariate and include all regression model parameters. Likewise, historical CGM data may be used to generate a digital twin or clone of the patient and simulate the model as an alternative method to generate an estimated fasting glucose value corresponding to the effective dose.

[0140] Various embodiments of the effective dose-EFBG response uncertainty modeler 290 may employ variance models that vary with the magnitude of the effective dose and / or variance models that depend on other characteristics.

[0141] Dose-LBGR Modeler

[0142] like Figure 6 As shown, the dose-LBGR modeler 150 receives the history of hypoglycemia risk values ​​and the history of insulin effective doses. The modeler outputs a dose-LBGR regression model that predicts the hypoglycemia risk corresponding to the effective dose. In addition to predicting the hypoglycemia risk, the dose-LBGR model also includes an assessment of how the prediction will be able to change, and an estimate of the insulin dose that will achieve the maximum hypoglycemia threshold (called the dose at the low BG guardrail).

[0143] Figure 6 The inputs and outputs generated by the dose-LBGR modeler 150 during day t are shown. The dose-LBGR modeler 150 receives two input streams on day t. One input stream represents the daily history of hypoglycemia risk values ​​generated by the CGM module, for days d=1, ..., t-1, labeled LBGR d In one embodiment, the risk of hypoglycemia is defined as the 0.5 percentile of the patient's CGM values. Another input stream represents the daily history of estimated effective insulin doses generated by the insulin module for day d, where day d = 1, 2, ..., t-1, labeled ED d .

[0144] Based on these inputs, the effective dose-LBGR response modeler 310 generates a linear regression model with an intercept I lbgr and coefficient C lbgr , used to relate hypoglycemia risk to effective insulin dose.

[0145] To bias the regression model towards a safe population-averaged dose response early in the titration process, the linear regression objective was modified by adding a regularization term as follows:

[0146]

[0147] This equation is based on the coefficient C lbgr The mean value is μ C And the standard deviation is σ C The normal distribution population model is based on C lbgr The effective dose-LBGR response modeler 310 also generates a goodness of fit r for the regression model. 2 lbgr and an assessment of the dose at the low BG guardrail, which is the maximum dose that does not result in a greater than 5% likelihood of hypoglycemia.

[0148] The effective dose-LBGR response uncertainty modeler 320 then calculates the variance of the residuals produced by the effective dose-LBGR response modeler 310. It also applies a constant learning rule to update its lbgr Estimate the variance of .

[0149] In some cases, alternative definitions of hypoglycemia risk may be used, which may include the low blood glucose index (LBGI) or other percentiles.

[0150] In some alternative embodiments, the effective dose-LBGR response modeler 310 can employ nonlinear regression models, machine learning methods (including neural networks), and / or additional predictors in addition to the effective dose. In addition, a population model can be used in the regularization, which is multivariate and includes all regression model parameters. In addition, historical CGM data can be used to generate a digital twin or clone of the patient, and simulate the model as an alternative method to generate a hypoglycemia risk value corresponding to the effective dose.

[0151] Alternative embodiments of the effective dose-LBGR response uncertainty modeler 320 may employ variance models that vary with the magnitude of the effective dose and / or variance models that vary with other characteristics.

[0152] Target dose adjuster

[0153] like Figure 7 As shown, the target dose adjuster 160 receives the dose-EFBG model from the dose-EFBG modeler 140 and the dose-LBGR model from the dose-EFBG modeler 150 as inputs, and outputs a target dose obtained after a trade-off between efficiency and safety. The target dose is provided to the dose finalizer 170.

[0154] The submodules of the target dose adjuster 160 perform three main steps. The first and second steps described below can be performed in parallel.

[0155] In a first step, the effective target dose calculator 330 takes the dose-EFBG model as input and generates an effective dose from the inversion of the model. The dose-EFBG model contains the elements necessary to: (i) predict the estimated fasting blood glucose value and its surrounding error relative to the basal insulin dose; and (ii) evaluate the quality of the prediction. In one specific implementation of the dose-EFBG modeler 140, a dose-EFBG model with coefficients C is used. efbg and the intercept I efbg The linear model is used to perform predictions, and the error around the prediction is described as having a 68% chance of falling within the predicted s efbg and using the r-squared value (r 2 efbg ) to evaluate the quality of the prediction. If the quality of the prediction is considered too poor (r 2 efbg If the value is too low, no effective dose is produced. Otherwise, the effective dose is the dose that satisfies the expected estimated fasting blood glucose with a certain confidence level (taking into account the prediction error). In one embodiment, the effective dose is based on s efbg The patient's actual response is equally likely to be above or below the target fasting glucose.

[0156] In the second step, the safe target dose calculator 340 takes the dose-LBGR model as input and generates a safe dose from the inversion of the model. The dose-LBGR model contains the elements necessary to: (i) predict hypoglycemia risk values ​​and their surrounding errors relative to the basal insulin dose; and (ii) evaluate the quality of the prediction. In one specific implementation of the dose-LBGR modeler 150, a safe target dose calculator 340 is used with coefficients C lbgr and the intercept I lbgr The linear model is used to perform predictions, and the error around the prediction is described as having a 68% chance of falling within the predicted s lbgr and using the r-squared value (r 2 lbgr ) to evaluate the quality of the prediction. If the quality of the prediction is considered too poor (r 2 lbgrIf the risk of hypoglycemia is too low, no safe dose is generated. Otherwise, the safe dose is a dose that meets the expected risk of hypoglycemia with a certain confidence level (taking into account the prediction error). In a specific implementation, the safe dose is based on s lbgr A dose such that there is a 95% chance that the actual LBGR response will correspond to a hypoglycemic exposure below the defined threshold.

[0157] In the final step performed by the target dose adjuster 160, the dose coordinator 350 takes the effective dose and the safe dose as input and generates a target dose. The target dose is the result of a trade-off between the effective dose and the safe dose. In one specific implementation, the target dose is equal to the safe dose.

[0158] In some implementations, the target dose adjuster 160 can use any of the various embodiments of the dose-EFBG model and the dose-LBGR model described above. In addition, the confidence level used can be set to any suitable value.

[0159] In some embodiments, the dose coordinator 350 may use the minimum of the effective dose and the safe dose, the maximum of the effective dose and the safe dose, or a weighted average of the effective dose and the safe dose, where the weight is calculated based on the model accuracy and / or prediction error.

[0160] Dose Finalizer

[0161] like Figure 8 As shown, the inputs of the dose finalizer 170 include the target dose, the output from the CGM module 110 ({I <70,d}、{I <54,d} and BG ref ), hypoglycemia event history {H1, H2, …, H m}, most recently adjusted dose (RD t-1 ) and the dose at the lower BG guardrail DLBGG t-1 The module applies a set of safeguards and outputs a new adjusted dose.

[0162] The dose finalizer 170 applies a set of safeguards to the recommended target dose to mitigate the risk of hypoglycemia and prevent hypoglycemia relative to the current adjusted dose RD. t-1 In this regard, the dose finalizer 170 may define a target deadband around a target glucose value, which is a design parameter of the algorithm. The target deadband is a range of glucose values ​​within which the adjusted dose will not change. The various submodules of the dose finalizer 170 operate as follows.

[0163] The consistency validator 360 applies a set of rules to ensure that the proposed target dose is responsive to the patient's clinical state. For example, if BG Ref In the target dead zone, the consistency verifier can set the adjusted dose to the current dose RD t-1 If BG Ref Below the target dead zone and the target dose is greater than the current dose RD t-1 , the consistency verifier can also set the adjusted dose as the current dose RD t-1 If BG ref If the target deadband is higher than the target deadband, the consistency verifier 360 can also set the adjusted dose to be greater than RD t-1 The value of .

[0164] Hypoglycemia-based protection 370 will be achieved by limiting adjusted dose = min (adjusted dose, DLBGG t-1 ) to ensure that the resulting adjusted dose does not produce hypoglycemia, for example 95% of the time. The module will further prevent <70,d} or {I <54,d} or hypoglycemia event history {H1,H2,…,H m}Initiate any increase in dosage in the presence of recent indications.

[0165] Maximum variation protection 380 ensures a smooth dose change by setting a maximum variation. If the recommended target dose will make the current adjusted dose RD t-1 If the change exceeds the maximum change, then, instead, the adjusted dose = RD t-1 +Maximum Change to apply maximum change.

[0166] Finally, the over-basal insulinization check 390 generates a warning if the target dose exceeds the clinically acceptable limits for the patient's basal insulin therapy. This is a prediction of possible future over-basal insulinization. A warning is also generated if the adjusted dose produced by the maximum variation protection exceeds the clinically acceptable limits for the patient's basal insulin therapy. This is an assertion that over-basal insulinization currently exists.

[0167] In some alternative embodiments, the dose finalizer 170 may use any of a variety of different rules to establish the maximum allowable variation. Additionally, in some cases, the hypoglycemia-based protection 370 may be based on risk or other hypoglycemia-related measures. Finally, in some implementations, excessive basal insulination may introduce a constraint on the dose, rather than simply serving as a warning.

[0168] Termination Checker

[0169] like Fig. 9As shown, the termination checker 180 includes a termination checker submodule 410, which receives the estimated fasting blood glucose history, hypoglycemia event history, hypoglycemia risk history, time index history below 70, time index history below 54, reference blood glucose, daily effective dose history and recommendation history as input, and outputs a termination report. Figure 1 As shown:

[0170] ●Estimated fasting glucose history refers to the output of the CGM module

[0171] ●The reported hypoglycemia event history refers to the output of the reporting hypoglycemia module

[0172] ●Hypoglycemia risk history refers to the output of the CGM module

[0173] ●CGM below 70 indicator history refers to the output of the CGM module

[0174] ●CGM below 54 indicator history refers to the output of the CGM module

[0175] ● Reference blood sugar BG ref Refers to the output of the CGM module

[0176] ● Daily effective dose history refers to the output of the insulin module

[0177] The recommendation history used as input (where the recommendation history R d (each element for d = 1, 2, ..., t is the recommended element for day d) is described by both the adjusted dose (output of the dose finalizer) and the termination report (output of the termination checker). Likewise, the termination report, which is also used as input, contains termination information such as whether termination occurred, and if so, the timestamp and reason for termination.

[0178] The termination checker 180 performs two main steps. In the first step, the termination checker submodule 410 receives all inputs and outputs a termination report. In one implementation, the termination checker submodule 410 checks which elements of a given list of termination criteria are verified. If any criteria are verified, the termination report returns "yes", and if it returns "yes", it specifies which criteria are verified. In one particular implementation, three different criteria can terminate the titration process:

[0179] a. All seven past elements of the estimated fasting glucose history are within the desired target range, all seven past elements of the time indicator history are below 70 are below the desired target, all seven past elements of the time indicator history are below 54 are below the desired target, and no elements in the hypoglycemic event history have timestamps in the past seven days.

[0180] b. The seven past adjusted doses obtained from the last element of the recommendation history are equal to zero.

[0181] c. The termination deadline has been reached: This is the 30-day scheduled termination deadline.

[0182] In the second major step performed by the termination checker 180, the termination reporter 420 receives as input the termination report and the recommendation history from the termination checker submodule 410. If the termination reporter 420 indicates that termination occurred, the termination reporter returns the last termination report from the recommendation history, otherwise it returns the termination report from the termination checker 180.

[0183] In some alternative embodiments, the termination checker submodule 410 may use different inputs. For example, the termination decision may be based on alternative CGM or insulin dose statistics that can be calculated by the CGM module or insulin module, respectively. The termination checker submodule 410 may also use alternative blood glucose or insulin measurements that more broadly reflect diabetes management criteria (e.g., time in range, estimated HbA1c).

[0184] Likewise, different embodiments may use different termination criteria to make termination decisions. For example, termination decisions may be based on alternative statistical values ​​associated with the patient's hypoglycemia experience, which may be calculated by the report hypoglycemia module 130. Alternatively, termination decisions may be based on patient satisfaction or caregiver satisfaction. In addition, termination criteria may use different evaluation windows or any measure that reaches a desired target range in any amount of time.

[0185] In some cases, termination reporting may be based on the system not being able to reach target BG within the expected number of titration days. ref In this case, the report may indicate that variability in the CGM trace (whether caused by the patient's physiological state or behavior) limits achievement of target BG without the introduction of prandial insulin (e.g., multiple daily injection therapy) ref ability.

[0186] Fig.10 An example of an exemplary operating environment in which the basal titration adjustment system and method described herein can be run is shown. Of course, those of ordinary skill in the art will recognize that the basal insulin titration adjustment system and method can also be used in other environments, and in some cases can even be run as a stand-alone device, or alternatively, the method can be implemented on any suitable computing device in conjunction with additional functions. As shown in the figure, the environment includes an insulin delivery device 110, a glucose monitor 120, a processor 130, a subject 140, an activity monitor 150, and a smart phone 160.

[0187] One or more of the insulin delivery device 110, glucose monitor 120, processor 130, activity monitor 150, and smartphone 160 may communicate via a network. The network may be a variety of network types, including a public switched telephone network (PSTN), a cellular telephone network, and a packet switched network (e.g., the Internet). Figure 1 Only an insulin delivery device 110, one glucose monitor 120, one processor 130, one subject 140, one activity monitor 150, and one smart phone 160 are shown, but there is no limit to the number of insulin delivery devices 110, glucose monitors 120, processors 130, subjects 140, activity monitors 150, and smart phones 160 that can be supported.

[0188] The insulin delivery device 110 may be any device that dispenses insulin, such as a syringe, a pump (e.g., external, mechanical, patch, or implanted), and an inhaler. The insulin delivery device 110 may also include a device that dispenses other medications that help control glucose levels, such as glucagon (dual-hormone artificial pancreas), GLP-1, etc. Depending on the specific type of device employed, the insulin delivery device may or may not be a connected device that communicates over a network.

[0189] Depending on the specific implementation, the glucose monitor 120 can be any type of CGM or SMBG (self-monitoring of blood glucose) device. The glucose monitor 120 can be a connected device that provides glucose readings continuously, or provides a set of glucose readings when the device is scanned or downloaded. In addition to providing glucose readings, the glucose monitor 120 can also record user interactions, such as when and how users (e.g., subjects, patients, caregivers, medical professionals, etc.) view their glucose traces, and how they respond to warnings and alarms. User interactions can provide insights into the timing and motivations of treatment decisions, including why and when they consider the impact of eating or administering insulin.

[0190] Processor 130 collects data from insulin delivery device 110 and glucose monitor 120 and subject 140, and in some embodiments, may run the basal insulin titration method described herein. Processor 130 (as well as insulin delivery device 110, glucose monitor 120, activity monitor 150, and / or smartphone 160) may be implemented using a variety of computing devices, such as smartphones, desktop computers, laptop computers, and tablet computers. In this regard, processor 130 may be implemented in a computer system such as a computer system, ... Figure 1The processor may be used in one of the other devices shown in the operating environment of the present invention, such as a smart phone 160, an activity monitor 150, or, depending on the type employed (e.g., a smart pen), an insulin device 110. Other types of computing devices in which the processor 130 may be incorporated may also be supported. Suitable computing devices are Fig.11 5 is shown as computing device 500.

[0191] To provide a robust system, these calculations performed by processor 130 may be dynamic and distributed, depending on the connected devices and processors. For example, cloud computing may be used when there is a connection site, a smartphone processor may be used when there is no connection site, and then a transmitter or smartwatch may be used when the smartphone is not connected. Complex calculations, such as model optimization, can only be run when a more powerful processor is available. When a powerful processor is not available, the algorithm may use the latest parameters or simpler approximations.

[0192] Subject 140 can provide input to the system, including information about meals, activities, and diabetes treatment (e.g., basal insulin injections), using, for example, any computing device that communicates with the system, such as subject 140's smartphone 160 or other computing device. These inputs can be initiated by the user or prompted by the system. These inputs can describe current, previous, and / or upcoming events.

[0193] Activity monitor 150 can be any device that monitors the physiological and mental state of a user. An example is a fitness tracker that uses accelerometers, gyroscopes, heart rate, and oxygen sensors to monitor activity, exercise, and sleep. This can also include smartphones, as they can detect location and user activity / interaction, or smart home devices (e.g., Amazon Alexa). In some implementations, activity monitor 150 can include a device that detects meals.

[0194] Smartphone 160 can be used as an activity and context monitor, data entry device, data collection device (talking to devices with Bluetooth, NFC, Wi-Fi, etc.), and run applications (e.g., apps) that estimate nutritional information through manual input or automatic input (e.g., photos).

[0195] Fig.11An exemplary computing device in which the exemplary embodiments and aspects may be implemented is shown. The computing device is only one example of a suitable computing device and is not intended to imply any limitation on the scope of use or functionality. Many other general or special computing device environments or configurations may be used. Examples of known computing devices, environments, and / or configurations suitable for use include, but are not limited to, personal computers, server computers, handheld or laptop devices, multiprocessor systems, microprocessor-based systems, network personal computers (PCs), minicomputers, mainframe computers, embedded systems, distributed computing environments including any of the above systems or devices, and the like.

[0196] Computer executable instructions such as program modules that are executed by a computer may be used. Typically, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. Distributed computing environments may be used in which tasks are performed by remote processing devices that are linked through a communication network or other data transmission medium. In a distributed computing environment, program modules and other data may be located in local and remote computer storage media including memory storage devices.

[0197] refer to Fig.11 , an exemplary system for implementing various aspects described herein includes a computing device, such as computing device 500. In its most basic configuration, computing device 1500 typically includes at least one processing unit 502 and memory 504. Depending on the exact configuration and type of computing device, memory 504 can be volatile memory (such as random access memory (RAM)), non-volatile memory (such as read-only memory (ROM), flash memory, etc.), or some combination of the two. Fig.11 This most basic configuration is shown by dashed line 506 in FIG.

[0198] The computing device 500 may have additional features / functionality. For example, the computing device 500 may include additional storage devices (removable and / or non-removable), including but not limited to magnetic or optical disks, or tapes. Such additional storage devices may be stored in the computer system. Fig.11 , shown by removable storage device 508 and non-removable storage device 510.

[0199] Computing device 500 typically includes a variety of computer-readable media. Computer-readable media can be any available media that can be accessed by device 500 and includes both volatile and nonvolatile media, and both removable and non-removable media.

[0200] Computer storage media includes volatile and nonvolatile media, as well as removable and non-removable media implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules or other data. Memory 504, removable storage 508, and non-removable storage 510 are all examples of computer storage media. Computer storage media include, but are not limited to, RAM, ROM, electrically erasable programmable read-only memory (EEPROM), flash memory or other storage technology, CD-ROM, digital versatile disk (DVD) or other optical storage, cassettes, magnetic tape, magnetic disk storage or other magnetic storage, or any other medium that can be used to store the desired information and can be accessed by computing device 500. Any such computer storage media can be part of computing device 500.

[0201] The computing device 500 may include a communication connection 512 that allows the device to communicate with other devices. The computing device 500 may also have an input device 514, such as a keyboard, a mouse, a pen, a voice input device, a touch input device, etc. It may also include an output device 516, such as a display, a speaker, a printer, etc. All of these devices are well known in the art and do not need to be discussed in detail here.

[0202] It should be understood that the various techniques described herein may be implemented in conjunction with hardware components or software components, or, where appropriate, in conjunction with a combination of the two. Exemplary types of hardware components that may be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), etc. The methods and systems of the disclosed subject matter, or specific aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in a tangible medium (such as a floppy disk, CD-ROM, hard drive, or any other machine-readable storage medium), wherein when the program code is loaded into and executed by a machine (such as a computer), the machine becomes a device for performing the disclosed subject matter.

[0203] Although exemplary implementations may refer to utilizing aspects of the disclosed subject matter in the context of one or more stand-alone computer systems, the subject matter is not limited thereto, but may be implemented in conjunction with any computing environment, such as a network or distributed computing environment. In addition, aspects of the presently disclosed subject matter may be implemented in or across multiple processing chips or devices, and storage devices may similarly be implemented across multiple devices. For example, such devices may include personal computers, network servers, and handheld devices.

[0204] Although the subject matter has been described in language specific to structural features and / or methodological acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as exemplary forms of implementing the claims.

Claims

1. A basal titration adjustment device for titrating a basal insulin dose of a subject, comprising: one or more processors; a continuous glucose monitor (CGM) module which, when executed by the one or more processors, receives CGM data obtained over at least a series of days, generates historical estimated glucose values ​​(EGVs) therefrom, and then further generates each of a plurality of daily blood glucose state parameters from the EGV values, each of the daily blood glucose state parameters reflecting an aspect of the subject's blood glucose state; an insulin module that, when executed by the one or more processors, receives daily basal insulin dosage values ​​administered to the subject over the series of days and generates daily effective dosages therefrom, each of the daily effective dosages representing the total plasma insulin acting on the subject on the corresponding day and accounting for the accumulation of insulin in the subject resulting from all previously administered dosages; at least a first dose-daily blood glucose state response modeler which, when executed by the one or more processors, receives a first daily blood glucose state parameter of the plurality of daily blood glucose state parameters and the daily effective dose, thereby generating a first regularized dose-response model relating the first daily blood glucose state parameter to an effective basal insulin dose; a target dose adjuster that, when executed by the one or more processors, receives the first regularized dose-response model and generates therefrom a target basal insulin dose that will produce a value of the first daily glycemic state parameter predicted by the first regularized dose-response model at least with a specified confidence level; and a dose finalizer which, when executed by the one or more processors, receives the target basal insulin dose and the value of one or more of the daily glycemic state parameters, thereby generating an adjusted basal insulin dose which modifies the target basal insulin dose according to one or more pre-established safety rules.

2. The basal titration adjustment device according to claim 1, wherein the plurality of daily blood glucose state parameters are selected from the group consisting of: an effective fasting blood glucose (EFBG) parameter, a hypoglycemia risk parameter representing a daily risk of experiencing a low EGV, a low daily EGV parameter specifying whether the subject experiences an EGV below a specified level on a given day, and a daily reference blood glucose parameter reflecting a comparison result of the daily EGV with a treatment target EGV.

3. The basal titration adjustment device according to claim 1, wherein the plurality of daily blood glucose state parameters include an effective fasting blood glucose (EFBG) parameter, a hypoglycemia risk parameter representing a daily risk of experiencing a low EGV, a low daily EGV parameter specifying whether the subject experiences an EGV below a specified level on a given day, and a daily reference blood glucose parameter reflecting a comparison result of the daily EGV with a treatment target EGV.

4. The basic titration adjustment device according to claim 1, wherein the first regularized dose-response model is a regularized linear regression model.

5. The basic titration adjustment device according to claim 1, wherein the first regularized dose-response model is regularized using population data to at least bias the initial target dose to a safe dose at the population level.

6. The basal titration adjustment device according to claim 1, wherein the first daily blood glucose state parameter is a daily estimated fasting blood glucose (EFBG) parameter.

7. The basal titration adjustment device according to claim 1, wherein the first daily blood glucose status parameter is a hypoglycemia risk parameter.

8. The basal titration device of claim 1, wherein the one or more pre-established safety rules reduce the risk of hypoglycemia.

9. The basal titration adjustment device according to claim 1 further includes a reporting hypoglycemia module, which, when executed by the one or more processors, receives hypoglycemia event data reported by the subject and generates a hypoglycemia reporting history from the hypoglycemia event data and the historical EGV, wherein the hypoglycemia reporting history specifies the time and severity of credible hypoglycemia events reported by the subject, wherein the dose finalizer, when executed by the one or more processors, uses the hypoglycemia reporting history as an additional input to generate the adjusted basal insulin dose.

10. The basal titration adjustment device according to claim 9 further comprises a termination checker module, which, when executed by the one or more processors, receives values ​​of some or all of the plurality of daily blood glucose state parameters, the hypoglycemia report history, and the adjusted basal insulin doses generated in previous days, thereby generating a termination report indicating whether the titration of basal insulin should be terminated or continued.

11. A basal titration adjustment device according to claim 1, wherein the at least first dose-daily blood glucose state response modeler includes a second dose-daily blood glucose state response modeler, which, when executed by the one or more processors, receives a second daily blood glucose state parameter among the multiple daily blood glucose state parameters and the daily effective dose, thereby generating a second regularized dose-response model that associates the second daily blood glucose state parameter with the effective basal insulin dose, and the target dose adjuster, when executed by the one or more processors, receives the first regularized dose-response model and the second regularized dose-response model, thereby generating the target basal insulin dose.

12. The basal titration adjustment device according to claim 11, wherein the first daily glycemic state parameter is a daily estimated fasting blood glucose parameter, and the second daily glycemic state parameter is a hypoglycemia risk parameter, which specifies the risk of the subject's EGV being below a threshold one or more times in a day.

13. The basic titration adjustment device according to claim 4, wherein the first regularized linear regression model includes values ​​for slope, intercept and uncertainty measure.

14. The basal titration device of claim 6, wherein the value of the EFBG parameter is based on CGM data obtained at a specified time of day.

15. The basal titration device of claim 6, wherein the value of the EFBG parameter is based on a CGM value that is lower than a particular percentile of the subject's glycemic distribution.

16. The basal titration adjustment device of claim 9, wherein a hypoglycemic event is deemed irrelevant if it occurred more than a specified period of time before the current time, or if the hypoglycemic event occurred when the current adjusted basal insulin dose at that time differed from the current adjusted basal insulin dose by more than a specified amount.

17. The basal titration adjustment device according to claim 7, wherein the first dose-daily blood glucose state response modeler generates a basal insulin dose-hypoglycemia risk regression model, which predicts the hypoglycemia risk corresponding to the effective basal insulin dose.

18. The basal titration adjustment device according to claim 17, wherein the basal insulin dose-hypoglycemia risk regression model is a linear regularized linear regression model.

19. The basal titration adjustment device according to claim 18, wherein the linear regularized linear regression model is regularized using population data to at least bias the initial target dose to a safe dose at the population level.

20. The basal titration device of claim 7, wherein the hypoglycemia risk parameter represents the risk of the subject's EGV falling below a threshold one or more times in a day.

21. The basal titration device of claim 7, wherein the hypoglycemia risk parameter represents a daily EGV that is below a specified percentile of all EGVs of the subject in a day.

22. The basal titration adjustment device of claim 1, wherein the dose finalizer, when executed by the one or more processors, periodically generates the adjusted basal insulin dose.

23. The basal titration adjustment device according to claim 22, wherein the periodic period is daily.

24. The basic titration adjustment device according to claim 11, wherein the target dose adjuster, when executed by the one or more processors, determines a first preliminary target dose from the first regularized dose-response model and determines a second preliminary target dose from the second regularized dose-response model, and generates the target dose by balancing the first preliminary target dose with the second preliminary target dose.

25. The basic titration adjustment device according to claim 24, wherein the generated target dose is the minimum value of the first preliminary target dose and the second preliminary target dose, the maximum value of the first preliminary target dose and the second preliminary target dose, or the weighted average value of the first preliminary target dose and the second preliminary target dose.

26. The basal titration adjustment device of claim 1, wherein the one or more pre-established safety rules prevent a newly generated adjusted basal insulin dose from having a value exceeding a previously generated adjusted basal insulin dose of a previous day.

27. A basal titration adjustment device according to claim 1, wherein the one or more pre-established safety rules establish values ​​for design parameters defining an EGV target deadband around a target EGV, such that for values ​​of a daily reference blood glucose parameter within the target deadband, a newly generated adjusted basal insulin dose will not be adjusted from a previously generated adjusted basal insulin dose.

28. The basal titration device of claim 1, wherein the one or more pre-established safety rules generate a warning predicting that excessive basal insulinization will occur.

29. The basic titration adjustment device according to claim 1, wherein: The one or more pre-established safety rules constrain modification of the target basal insulin dose if excessive hyperinsulinization is predicted to occur.

30. A basal titration adjustment device according to claim 10, wherein the termination report indicates that if the target value of a daily reference blood glucose parameter cannot be achieved without introducing prandial insulin, the titration of the basal insulin dose should be terminated, and the daily reference blood glucose parameter reflects the comparison result of daily EGV with the treatment target EGV.

31. A method for titrating a subject's basal insulin dose, comprising: receiving CGM data obtained over at least a series of days to generate a historical estimated glucose value (EGV); generating each of a plurality of blood glucose state parameters from the EGV value, each of the blood glucose state parameters reflecting an aspect of the blood glucose state of the subject; accepting the basal insulin dose values ​​administered to the subject on the series of days, thereby generating effective doses, each of which represents the total plasma insulin acting on the subject on the corresponding day and accounts for the accumulation of insulin in the subject resulting from all previously administered doses; receiving a first blood glucose state parameter of the plurality of blood glucose state parameters and the effective dose, thereby generating a first regularized dose-response model relating the first blood glucose state parameter to an effective basal insulin dose; receiving the first regularized dose-response model, thereby generating a target basal insulin dose that will produce a value of the first glycemic state parameter predicted by the first regularized dose-response model at least with a specified confidence level; as well as The target basal insulin dose and a value of one or more of the plurality of glycemic state parameters are received, thereby generating an adjusted basal insulin dose that modifies the target basal insulin dose according to one or more pre-established safety rules.

32. The method of claim 31 , wherein the basal insulin dose is administered daily.

33. The method of claim 31 , wherein an adjusted basal insulin dose is produced daily.

34. A non-transitory computer readable medium comprising instructions for causing a computing device to perform a method, the method comprising the method of claim 31.