Reference blood glucose based algorithmic titration of diabetes medications using continuous glucose monitoring

Abstract

Systems and methods are described for determining a reference blood glucose (RefBG) metric and for determining a medicament dose or dose adjustment for a subject based at least in part the reference blood glucose metric. The RefBG metric is based on continuous and / or intermittent glucose monitoring samples and can be derived from a specified time window of the samples. One or more additional metrics of a therapeutic performance of the subject may be modeled and used to establish a target value for the RefBG metric, which may be compared to historical RefBG values to determine dose values to be avoided in future updates of medicament doses.

Description

Dexcom Docket No. 0991-PCT01Reference Blood Glucose Based Algorithmic Titration of Diabetes Medications using Continuous Glucose MonitoringCross Reference to Related Application

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 717,696 filed November 7, 2024 entitled “REFERENCE BLOOD GLUCOSE BASED ALGORITHMIC TITRATION OF DIABETES MEDICATIONS USING CONTINUOUS GLUCOUSE MONITORING”, the contents of which are incorporated herein by reference.Field

[0002] The present disclosure relates generally to systems and methods for measuring and analyzing data obtained from a continuous analyte sensor. The present disclosure also relates to the determination of medicament dose adjustments or recommendations based at least in part on reference blood glucose metrics.Background

[0003] Despite years of continuous glucose monitoring (CGM) replacing fingerstick sampling for medical decision-making in diabetes, point of care clinical assessments remain the basis for interpreting, judging, and formulating medical treatment. Among these is “fasting Blood Glucose,” which is usually defined as the patient’s blood glucose after a fast of several hours (e.g., eight or more) usually overnight and can be obtained from venous or capillary blood. When obtained in a true fasting state, the fasting measure can be correlated to the patient’s underlying disease state.

[0004] Fasting Blood Glucose (FBG) measurements taken at home are often corrupted by the patient not actually being in a normal fasting state. Depending on what was eaten and when, the sample value may read (i) artificially high due to ongoing absorption of carbohydrates in the meal as glucose rate of appearance or (ii) artificially low as the result of endogenous overcompensation with secreted insulin. Similarly, recent physical activity can have a corrupting effect on the assessment of fasting BG, causing the sample to not be representative of the patient’s underlying medical condition, and this, in turn, can lead to erroneous medical decision making.Summary

[0005] In a first aspect, a method of improving glycemic outcomes for a person with diabetes comprises measuring or determining a first value of a reference parameter associated with the person, wherein the reference parameter is based at least in part on a respective plurality ofDexcom Docket No. 0991-PCT01 continuous or intermittent glucose monitoring samples, deriving a target dose for a medicament based at least in part on a history of continuous or intermittent glucose monitoring, providing a first therapeutic recommendation based on the first value of the reference parameter, the first therapeutic recommendation comprising an adjusted dose of the medicament, measuring or determining a second value of the reference parameter associated with the person, and providing a second therapeutic recommendation based on the first therapeutic recommendation and the second value of the reference parameter, the second therapeutic recommendation comprising an update to the adjusted dose of the medicament.

[0006] In a second aspect, a method of determining a medicament dose comprises receiving Continuous Glucose Monitoring (CGM) data for a subject over a period of time, defining a reference blood glucose (RefBG) metric based at least in part on the CGM data, wherein the RefBG metric is a statistic derived from a specified time window of the CGM data, modeling one or more additional metrics of a therapeutic performance of the subject using the RefBG metric to produce one or more models of therapeutic performance, establishing (i) a medicament dose and (ii) a RefBG target value using the one or more models of therapeutic performance, evaluating whether one or more values of the RefBG metric are above the RefBG target value, and, when the one or more values of the RefBG metric are above the RefBG target, avoiding use of the corresponding medicament dose in future updates of the medicament dose, and adjusting the medicament dose to provide a subsequent medicament dose based, at least in part, on the evaluation of the one or more values of the RefBG metric.

[0007] In a third aspect, a system for determining a medicament dose comprises a processor and computer-readable storage media storing instructions executable by the processor to: receive Continuous Glucose Monitoring (CGM) data for a subject over a period of time, define a reference blood glucose (RefBG) metric based at least in part on the CGM data, wherein the RefBG metric is a statistic derived from a specified time window of the CGM data, model one or more additional metrics of a therapeutic performance of the subject using the RefBG metric to produce one or more models of therapeutic performance, establish a RefBG target value using the one or more models of therapeutic performance, for a plurality of historical RefBG values, evaluate whether each respective historical RefBG value is above the RefBG target value, and, when the respective historical RefBG value is above the RefBG target, mark a dose value or values that produced the respective historical RefBG value as “to-be-avoided” in future updates of the medicament dose, and adjust a current medicament dose to provide fixed-size or proportional dose adjustment away from the dose value or values marked as “to-be-avoided.”Dexcom Docket No. 0991-PCT01

[0008] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. Furthermore, the claimed subject matter is not limited to implementations that solve any or all disadvantages noted in any part of this disclosure.Brief Description of the Drawings

[0009] FIG. 1 is a functional block diagram of one example of a basal insulin titration system according to one or more examples of the present disclosure.

[0010] FIG. 2 shows one example of the CGM Module shown in FIG. 1.

[0011] FIG. 3 shows one example of the Insulin Module shown in FIG. 1.

[0012] FIG. 4 shows one example of the Reported Hypo Module shown in FIG. 1 .

[0013] FIG. 5 shows one example of the Dose-EFBG Modeler shown in FIG. 1.

[0014] FIG. 6 shows one example of the Dose-LBGR Modeler shown in FIG. 1.

[0015] FIG. 7 shows one example of the Target Dose Adjuster shown in FIG. 1

[0016] FIG. 8 shows one example of the Dose Finalizer shown in FIG. 1.

[0017] FIG. 9 shows one example of the Termination Checker shown in FIG. 1.

[0018] FIG. 10 is a functional block diagram of another example of a basal insulin titration system according to one or more examples of the present disclosure.

[0019] FIG. 11 shows one example of a Dose Adapter System.

[0020] FIG. 12 shows an example of an illustrative environment in which a basal titration adjustment system and method according to one or more examples of the present disclosure may be operated.

[0021] FIG. 13 shows one example of a computing device in which examples and aspects of the present disclosure may be implemented.Detailed Description

[0022] In one aspect, the subject matter described herein addresses the problem of erroneous medical decision-making due to corrupted fasting blood glucose samples by (i) replacing the “fasting blood glucose” sample with a new metric that is referred to herein as reference bloodDexcom Docket No. 0991-PCT01 glucose (“RefBG”) computed from continuous or intermittent glucose monitoring, and (ii) introducing medication titration algorithms with dose-adjustment gains and thresholds that are tuned to yield optimal outcomes in terms of standard-of-care metrics, including Ale and time-in- range, either at an individual or population level.

[0023] The RefBG metric is any statistic (e.g., a reference parameter) derived from a time window (e.g., a 24-hour window) of continuous glucose monitoring having one or more of the following properties:• statistical stability I robustness (minimal day-to-day variation),• freedom from behavioral influences (absent post prandial effects and physical activity),• correlation to FBG.Thus, the RefBG metric may be defined in a number of different ways. Some illustrative concrete examples of an RefBG metric include:1. The Xthpercentile of an individual’s or population’s daily CGM profile, where X is chosen as the percentile that: a. Minimize inter-day variance. i. For example, the 12lhpercentile computed from a population of basal insulin users, not on MDI. b. Minimize a cost function that penalizes both inter-day variance and proximity to true FBG samples (if available) with weights of the cost function chosen to reflect user preferences.2. Median of CGM samples that are: a. Not within X hours (e.g. t_spike - .25 hr, t_spike + X-.25 hr) of a meal, acknowledged or detected (e.g. with glucose spikes), and b. Not within Y hours of physical activity (e.g. t_activity - .25 hr, t_activity + X-.25 hr), acknowledges or detected (e.g. physical activity trackers or transient responses in metabolic analytes (e.g. glucose dips that are unrelated to medicaments or lactate spikes marking intense physical activity)), and c. Not associated with high rates of change.

[0024] In some cases the RefBG metric may be used to titrate any of a variety of different medicaments. One example of such a method may be described by the following steps:Dexcom Docket No. 0991-PCT01Step 0 - Input individual- or population-level historical data• Daily CGM traces (e.g. all values between consecutive midnights)• True FBG samples from the same timeframe as the daily CGM traces, if availableStep 1 - Define a RefBG metric for an individual or a population using the data from Step 0.Step 2 - Relate the RefBG metric to Other Metrics of interestUse individual- or population-level data to model (e.g. via regression analysis, machine learning (ML), artificial intelligence (Al), statistical methods) key metrics of therapeutic performance to the RefBG of Step 1. Illustrative metrics of interest include:• Ale• Time in Range• True FBG (if available)Step 3 - Algorithmically titrate one or more medicaments with the RefBG metric.1. Initialize: a. Set an initial dose value, which can be: i. Determined by a health care practitioner (HCP), or ii. Determined from individual biomarkers b. Set a RefBG target value or range using the models of Step 2 to achieve desired therapeutic performance2. When it is time to update the medicament dose, a. [Evaluate] For all historical RefBG values collected during the titration process, evaluate whether the RefBG value is above the RefBG target or range, then mark the dose value or values that produced that RefBG as “to-be-avoided” in future titration updates. b. [Update] Update the current dose value to achieve a fixed-size or proportional dose adjustment away from to-be-avoided dose values.

[0025] Specific examples of applications in which the RefBG metric may be employed include titration of basal insulin, basal / incretin / SGLT2i co-therapy titration, multiple daily insulin (MDI) parameter titration (e.g., carb ratios, correction factors, insulin-on-board (IOB) durations, basal doses and fixed doses (per meal type or time-of-day). The RefBG metric also may be employed inDexcom Docket No. 0991-PCT01 systems such as automated insulin delivery (AID) systems and other closed-loop, open-loop and partially open-loop insulin delivery systems and other medicament delivery systems.

[0026] One exemplary application in which the RefBG metric may be employed is a CGM-driven basal insulin titration system and method for patients with Type 2 Diabetes. Viewed at a high level, the system operates to (i) interpret CGM history and dose insulin history data in the form of one or more dose-response models from which a target dose is derived and (ii) account for RefBG values, fasting blood glucose values, exposure to hypoglycemia, and / or past recommendations to provide a therapeutic recommendation, which may include computing an adjusted dose and recommending next steps in the form of a termination report. More specifically, the system uses the CGM history to generate historical estimated glucose values (EGVs) therefrom and further generates from the EGVs values a plurality of daily blood glucose state parameters each reflective of an aspect of a blood glucose state of the subject. One such daily blood glucose state parameter that may be employed is the RefBG metric described herein. Other daily blood glucose state parameters that may be employed include, by way of example, an effective fasting blood glucose (EFBG) parameter, a low blood glucose risk parameter representing a daily risk of experiencing a low EGV and a low daily EGV parameter specifying whether or not the subject experienced an EGV below a specified level on a given day. In some cases the EFBG parameter may be eliminated, effectively being replaced by the RefBG metric.

[0027] In the context of this exemplary CGM-driven basal insulin titration system and method, the RefBG metric can serve as an index of the quality of hyperglycemia management via basal insulin therapy that can be computed without the knowledge or assumption that the patient is in a steady fasting state at any particular time. In part, this embodiment avoids the use of EFBG because it has been found that, in practice, due in part to patient behavior (e.g. the patient may not actually be in a fasting state when BG is stable), it is highly variable from day to day and often does not represent the lowest blood glucose value of the day.Examples

[0028] The following description provides a number of particular implementations of the CGM- driven basal insulin titration system and method for patients with Type 2 Diabetes described above in which the RefBG metric may be employed. The system and method can be adapted to the needs and concerns of subjects just starting on basal insulin therapy. As explained in more detail herein, the method uses as inputs historical CGM, basal insulin dose information, reports of hypoglycemia,Dexcom Docket No. 0991-PCT01 and past recommendations and generates an adjusted insulin dose (an indication of which may be transmitted to a medicament delivery device) along with a report advising whether to continue the titration process, or to stop. The method can generate a new recommendation on a regular basis (e.g., each day) until it determines an adequate, consistent dose size. Some of the illustrative features and advantages of the system and method are as follows.

[0029] The system and method can make individualized, model-based dose recommendations by estimating personalized dose-response models from CGM-derived glucose control metrics including glucose value percentiles and estimated fasting glucose. Regularization is added to the personalized model to bias fits in the first few days toward safe recommendations, when the CGM history is still sparse. CGM-based estimates of glucose variability are incorporated into the algorithm design to guard against dose sizes that may increase hypoglycemic risk. Finally, the system accounts for the time-delayed response of long-acting insulin, referred to herein as an insulin Effective Dose. The Effective Dose model is population-based (not personalized).

[0030] The system and method also incorporate checks that ensure safe and meaningful dose recommendations (e.g., for transmission to a medicament delivery device for recommending insulin amounts to be delivered by the medicament delivery device). For instance, one check may ensure that the dose does not increase following recent non-severe hypoglycemia. In some implementations, the system can issue an alert when clinical guidelines indicate that a maximum per-kilogram dose has been reached. In some implementations, a RefBG can be estimated each day, and dose changes can be forced to be coherent with the estimated value relative to a target value (e.g., dose decreases are not permitted when RefBG is above a target range, and dose increases are not permitted when RefBG is below a target range). The system may also force dose reductions in the presence of CGM-detected and patient-reported severe hypoglycemia.System Overview

[0031] FIG. 1 is a functional block diagram of one example of the basal insulin titration system 100 described herein. As a general matter, the system serves to read in historical data including CGM History, Dosed Insulin History, Reported Hypoglycemia (Hypo) History, and / or a Recommendation History, and can produce an Adjusted Dose and a Termination Report. CGM history generally refers to estimated glucose values with timestamps collected since the beginning of the titration process. Dosed Insulin History generally refers to dosed basal (long-acting) insulin injections with timestamps indicating when administered (e.g., injected) since the beginning of theDexcom Docket No. 0991-PCT01 titration process. Reported Hypo History generally refers to reported experiences of symptomatic hypoglycemic as reported by the patient, a health care practitioner (HCP), an informal care provider, and / or other concerned individuals, with each report indicating the level of severity of the event. Recommendation History generally refers to the history of previous adjusted dose amounts during the titration process.

[0032] In some implementations, the system can operate to (i) interpret CGM History and Dose Insulin History data in the form of one or more dose-response models from which a Target dose is derived and (ii) account for RefBG values, fasting blood glucose values, exposure to hypoglycemia, and / or past recommendations to compute an Adjusted Dose and to recommend next steps in the form of a Termination Report. In some examples, the Adjusted Dose may be transmitted to a medicament delivery device.

[0033] In some implementations, the flow of data through the various components or modules of the system is as follows.

[0034] Referring to FIG. 1, CGM History data can be read into a CGM Module 110 (see the CGM Module section for further detail), which outputs sequences of, for example, (i) daily Estimated Fasting Blood Glucose (EFBG), {EFBGd}d=i, .... t , where d = 1 corresponds to the first day of the titration process and d = t corresponds to the EFBG value for the current day, (ii) daily Low Blood Glucose Risk (LBGR), {LBGRd}d=i, ...,t . (hi) daily indicators of non-severe low blood glucose (e.g. less than 70 mg / dl), { I<Thresh(L), d}d=i, .... t , where Id = True means that there was an experience of non-severe low blood glucose on day d, (iv) daily indicators of severe low blood glucose (e.g. less than 54 mg / dl), {I <Thresh(LL), <i } <1=1. .... t , where l — True means that there was an experience of severe low blood glucose on day d, (v) a determination as to when EGVs constitute a low-glucose event based, for example, on an HCP-specified threshold (the threshold can be specified as part of a larger “treatment plan” that also includes values such as the initial dose, maximum dose, and maximum daily increase of the dose.), and / or (vi) a reference blood glucose value, RefBG, the glucose statistic used by the Dose Finalizer to produce the Adjusted Dose.

[0035] Dosed Insulin History data can be read into an Insulin Module 120 (see the Insulin Module section for further detail), which produces a sequence of Effective Doses (ED), {EDd)d=i, ..., t-i , where t-1 corresponds to the most recent effective dose- the one that was active during the day of the patient’s life measured by CGM and rendered as EFBGt, LBGRt, I<Thresh(L), t, I<Thresh(LL), t, and RefBG.Dexcom Docket No. 0991-PCT01

[0036] Reported Hypo History and CGM History data can be read into a Reported Hypo Module 130 (see the Reported Hypo Module section for further detail), which produces a sequence of daily report hypoglycemic flags, {Hd]d=i, . t , where Hd takes different values depending on whether severe or non-severe symptomatic hypoglycemia where reported by the patient or caregivers on day d.

[0037] Daily Estimated Fasting Blood Glucose data {EFBGd]d=i, . , t and Effective Dose data {EDd]d=i, ... t-i , can be read into a Dose-EFBG Modeler 140 (see the Dose-EFBG Modeler section for further detail) that estimates the ED-to-EFBG response. In one implementation of the system, the dose response is approximated as a linear relationship with slope Cefbg, intercept Iefbg, residual standard deviation sefbg, and quality -of-fit reii>2.

[0038] Daily Low Blood Glucose Risk data {LBGRd]d=i, ..., t and Effective Dose data {EDd]d=i, ..., t- i , can be read into a Dose-LBGR Modeler 150 (see the Dose-EFBG Modeler section for further detail) that estimates the ED-to-LBGR response. In one implementation of the system, the dose response is approximated as a linear relationship with slope Cibgr, intercept Iibgr, residual standard deviation sibgr, and quality -of-fit Abgr.

[0039] The models produced by the Dose-EFBG Modeler 140 and Dose-LBGR Modeler 150 can be read into a Target Dose Adjuster 160 (see the Target Dose Adjuster section for further detail) that produces an adjusted Target dose that balances the desire for fasting glucose in a target range with low blood glucose risk.

[0040] An Adjusted Dose can be produced by a Dose Finalizer 170 (see the Dose Finalizer section for further detail), which operates on the Target dose (from the Target Dose Adjuster 160) along with (i) the reference glucose RefBG, (ii) the indicators of non-severe hypoglycemia { I<Thresh(L), d }a=i, ... t and severe hypoglycemia {I<Tiiresh(LD, d }d=i, ..., t , (iii) daily reported hypoglycemia [ Hd}d=i, ....t, (iv) past dose recommendations as recorded in the Recommendation History, and possibly (v) other algorithm status parameters from previous titration recommendations. The Dose Finalizer 170 can impose constraints that ensure that the final Adjusted Dose is safe and matches the patient’s expectations based on its experience with blood glucose in the preceding day.

[0041] Finally, the outputs of the CGM, Insulin, and Reported Hypo Modules, along with the Recommendation History, can be read into a Termination Checker 180 (see the Termination Checker section for further detail), producing a Termination Report articulating whether the titration process should terminate or continue and also making other follow-on recommendations,Dexcom Docket No. 0991-PCT01 e.g. when appropriate, suggesting that prandial (rapid-acting) insulin should be added to the patient’s therapy regime.

[0042] In summary, in some implementations, the basal insulin titration system described herein can read in CGM History, Dose Insulin History, Reported Hypo History, and the Recommendation History e.g., each day, to produce a Dose Recommendation and a Termination Report.

[0043] Exemplary implementations of the system shown in FIG. 1 are now described.

[0044] For example, different implementations may handle historical data in different ways. For instance, in some cases whole histories can be input to the system each day or just differential inputs describing new data since the last recommendation can be input, with most of the history expressed as an evolving “state” of system. Eikewise, CGM data can be obtained in different ways, such as by reading the CGM data directly from a CGM device or obtaining it from a database query from a server. Similarly, dosed insulin data can be obtained in different ways such as by reading it in from a smart pen device. In some implementations, dosed insulin data can read in from patient / caregiver-reported doses in a smartphone or web application (“app”).

[0045] Different implementations of the system can also be used to titrate different types of basal (e.g., short or long-acting) insulin. For instance, the basal insulin being titrated could be insulin glargine. In some implementations, the basal insulin could be other formulations including ones with longer pharmacokinetic profiles (e.g. insulin degludec, insulin icodec) or with shorter pharmacokintetic profiles (e.g. NPH).

[0046] The Reported Hypo data can be obtained from patients and / or caregivers in different ways. For example, reported hypo events can be obtained from data collected in a smartphone “app” or it could be collected via a web “app” or “portal”. Adjusted dose data can be stored in different ways such as in a smartphone “app” or in a database accessible via a data network.

[0047] The titration method described herein can be implemented in different ways. For instance, the method could be implemented in different devices and / or locations. For example, Adjusted Dose values and Termination Reports could be computed by a server having access to the CGM, Dosed Insulin, Reported Hypo, and Recommendation Histories. In some implementations, Adjusted Dose values and Termination Reports could be computed by a smart phone or other portable device having access to the CGM, Dosed Insulin, Reported Hypo, and Recommendation Histories. Similarly, the method could be implemented at different times, such as on demand, or it can be pre-computed and transmitted to patients and / or caregivers on a scheduled basis. AnDexcom Docket No. 0991-PCT01 example of an environment in which the titration method may be implemented are described below in connection with FIG. 12.

[0048] The output provided by the system and method can also be provided to different parties. For instance, Adjusted Dose values and Termination Reports could be viewed directly and implemented by patients, with caregivers having the ability to monitor therapy remotely. In some implementations, Adjusted Dose values and Termination Reports could be made available first to caregivers (HCPs, GPs, CDEs), who have the ability to approve or reject recommendations before they are sent to their patients.

[0049] The termination checker 180 can use different criteria for termination. For instance, the algorithm may be configured to minimize the time required to either reach an optimal abstract daily dose or reach a conclusion that basal insulin alone is insufficient to meet the patient's needs. In some implementations, the termination checker 180 may be configured to run in an ongoing fashion, continuously adapting the patient’s insulin dose in response to slowly changing insulin needs, and terminating only when it reaches the conclusion that basal insulin alone is insufficient to meet the patient’s needs.

[0050] The outputs from the CGM Module 110, referred to herein generally as daily blood glucose state parameters, may differ in different implementations and can be determined in a variety of different ways. For instance, the Low Blood Glucose Risk (LBGR) can be quantified in different ways (see the CGM Module section for further detail). In illustrative examples, the LBGR can be quantified as a percent time below a threshold glucose level, e.g. percent time below 70 mg / dl (or 3.9 mmol 11), or it can be quantified as the glucose level (in mg / dl or ml / 1 or other ways) associated with the Xth percentile of daily CGM values, where X could include .5, 1, 2, 10, or other percentiles. In other examples the LBGR can be computed as a risk space value as described in U.S. Pat No. 11,355,238, as a Low Blood Glucose Index (LBGI) or as described in U.S. Pat No. 9,317,657. In some implementations, the LBGR can be computed as an actionable low blood glucose risk value as described in U.S. Patent Publ. No. 2020 / 0178905A1.

[0051] The reference blood glucose (RefBG), which is another of the daily blood glucose state parameters, also can be quantified in different ways (see the CGM Module section for further detail). For instance, the RefBG can be quantified as the Estimated Fasting Blood Glucose (EFBG) referred to above or as time-average BG computed from CGM, or any other BG statistic that characterizes the overall quality of glycemic control. The RefBG may be a value of blood glucose or interstitial glucose.Dexcom Docket No. 0991-PCT01

[0052] In some implementations the computation of the Target dose may use different dose response models (see the Target Dose Adjuster section below). For instance, the target dose could be computed using just the Dose-LBGR Modeler, without the Dose-EFBG Modeler, or vice versa. In other embodiments different dose-response models could be used, including nonlinear doseresponse models and dose response models that characterize model uncertainty in different ways (other than residual standard deviation and quality -of-fit).

[0053] The systems and methods described herein can be adapted to different titration scenarios without changing their basic structure. For example, if the patient is new to basal insulin (either having never used it before or not currently using it), then the parameters of the titrator can be configured to (i) learn dose responses more rapidly and from lower initial doses than would otherwise be the case, (ii) avoid hypoglycemia which is often a concern of new insulin users, and (iii) be appropriately circumspect about reaching the conclusion that basal adjustments alone are insufficient. Alternatively, if the patient is already on basal insulin and is seeking to identify a better daily dose, then the titrator can take advantage of previously available blood glucose and insulin dosing data, and the parameters of the algorithm can be adapted to (i) learn dose responses near the patient’s currently prescribed dose, (ii) be less concerned about hypoglycemia (since insulin is by now a routine for the patient), and (iii) may more aggressively explore the conclusion that basal-insulin by itself is no longer adequate. In other embodiments, the titrator may be configured for continuous ongoing adaptation of the patient’ s basal insulin dose, tracking the patient’s ever-changing need for basal insulin in response to (i) seasonal effects, (ii) changes in diet / lifestyle, and / or (iii) changes in underlying physiology.

[0054] Illustrative embodiments of the individual components and modules shown in FIG. 1 and some implementations thereof are described in more detail below.CGM Module

[0055] As illustrated in FIG. 2, the CGM Module 110 can read in the CGM History and render values for a series (e.g., a time series) of daily blood glucose state parameters. In one exemplary embodiment described herein, values for one or more of such parameters can be output. For example:1. {EFBGd}: Estimated Fasting Blood Glucose (EFBG) History, a sequence with one positive numerical value per day d.Dexcom Docket No. 0991-PCT012. {LBGRd} : Low Blood Glucose Risk (LBGR) History, a sequence with one nonnegative value per day d.3. { I<Thresh(L).d} : Time Below L Indicator History, a sequence with one True or False value per day d indicating that the patient did (True) or did not (False) spend a substantial fraction of day d with their blood glucose below a threshold (e.g., 70 mg / dL).4. {I<Thresh(LL),d}: Time Below LL Indicator History, a sequence with one True or False value per day d indicating that the patient did (True) or did not (False) spend a substantial fraction of day d with their blood glucose below a threshold (e.g., 54 mg / dL).5. A singular series of indicators for EGVs below an HCP-specified threshold to mitigate subsequent risk of hypoglycemia by adjusting the dose according to the indicators.6. {RefBG}: Reference Blood Glucose value, which drives the Adjusted Dose toward one that achieves RefBGrefin a specified target range.

[0056] As shown in FIG. 2, the CGM Module 110 can contain an EGV Parser 200 that reads the CGM History and produces an EGV History {EGVi, EGV2, . . ., EGVn}. Each element EGVi of the EGV History can comprise a timestamp and an estimated blood glucose value (mg / dL). The elements of EGV History can be ordered by timestamp and nonduplicative with each other. The EGV History can be provided as input to each of the CGM Module’s five submodules, which correspond to the five outputs. Steps performed by each submodule are described in the examples below.

[0057] The Fasting Blood Glucose Estimator 210 (that can output {EFBGd}) can use the EGV History to estimate the patient’s Fasting Blood Glucose (e.g., for each day). The Fasting Blood Glucose corresponds to a physiological state that is free of short-term glucose disturbances caused by meals, exercise, illness, and other events. This physiological state is intended to be similar to the state created by a standard fast prior to laboratory blood work. The EFBGd value is intended to reflect the therapeutic effect of prior Adjusted Doses. Large prior Adjusted Doses are associated with smaller EFBGd values relative to small prior Adjusted Doses.

[0058] The Low Blood Glucose Risk Estimator 212 (that can output {LBGRd}) can use the EGV History to estimate the patient’s Low Blood Glucose Risk (e.g., for each day). A high (versus low) LBGRd value indicates that a confluence factors (e.g., insulin sensitivity, meals and exercise, andDexcom Docket No. 0991-PCT01 doses of basal insulin and / or other medications) caused the patient to experience elevated risks from low blood glucose (e.g., hypoglycemia) on day d.

[0059] The Time Below L Identifier 214 (that can output {I<thresh(Dd}) can use the EGV History to identify whether the patient spent a substantial fraction of day d with their blood glucose below a threshold L (e.g., 70 mg / dL). A single EGVi below the threshold could be the result of a transient CGM error, or it might be so isolated as to be therapeutically irrelevant. The Time Below L Identifier 214 can isolate therapeutically relevant EGVi values below the threshold, and flag the corresponding days with True values of I< threshed. All other days receive False values for I< thresh(T.),d.

[0060] The Time Below LL Identifier 216 (that can output {!< thresh(LL),d}) can use the EGV History to identify whether the patient spent a substantial fraction of day d with their blood glucose below a second threshold LL (e.g., 54 mg / dL). A single EGVi below the second threshold could be the result of a transient CGM error, or it might be so isolated as to be therapeutically irrelevant. The Time Below LL Identifier 216 can isolate therapeutically relevant EGVi values below the second threshold, and flag the corresponding days with True values of L tiiresh(LL),d. All other days receive False values for L thresh(LL),d.

[0061] The Reference Blood Glucose Calculator 218 (that can output RefBG) can use the EGV History to calculate a Reference Blood Glucose (RefBG) value. For example, the RefBG value may be a value of interstitial glucose (e.g., a reference interstitial glucose value), and the reference interstitial glucose value may be derived from a time series of CGM values. The reference value (RefBG value) characterizes the quality of prior Adjusted Doses vis-a-vis a therapeutic target. A Reference Blood Glucose value (RefBG) above the therapeutic target on day d suggests that the Adjusted Dose might need to increase for day d+1 (see Dose Finalizer for details).

[0062] Some implementations of the various submodules of the CGM Module 110 are described below.

[0063] In some cases, the Fasting Blood Glucose Estimator 210 may output {EFBGd} as: a. A statistic of the EGV History for day d with respect to the patient’s particular circadian rhythm of sleeping and eating. For example, EFBGd might be the average EGVi around the time the patient awakes on day d and prior to their first meal of the day. b. A statistic of the EGV History for day d that is a specified percentile known to correlate with the physiological state of fasting.Dexcom Docket No. 0991-PCT01 c. An estimation that combines several statistics (e.g., la and lb) with a weighting scheme (e.g., a linear function) to generate a statistic known to correlate with the physiological state of fasting.

[0064] In some cases, the Low Blood Glucose Risk Estimator 212 may output {LBGRd} as: a. A percentile (e.g., the 1stpercentile) of the EGV History on day d. In words, this percentile characterizes the fraction of time that the patient spends with their blood glucose below a certain level of mg / dL. For example, if the 1stpercentile of EGV History on day d is 68 mg / dL, then the patient spent 1% of day d with their blood glucose below 68 mg / dL. Their lst-percentile LBGR is 68. A larger or smaller percentile may be chosen to modify the severity of the Low Blood Glucose Risk assessed by the algorithm. b. Low Blood Glucose Index (LBGI), using the risk space approach described in U.S. Pat. No. 11,355,238 or U.S. Pat. No. 9,317,657. c. Actionable low blood glucose risk value as described in U.S. Pat. Appl. No. 20200178905. d. The left (negative) side of the Average Daily Risk Range value documented 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 Sep1 ;7(5): 1370-5. doi: 10.1177 / 193229681300700529. PMID: 24124966; PMCID: PMC3876383).In some cases, the Time Below L Identifier 214 may output {I< thresh(L),d} as: a. True for any day d in which several (e.g., 3) consecutive elements of EGV History are below a threshold (e.g., 70 mg / dL). False for all other days. b. In some implementations, I< thresh(L),d could be computed as an indicator of some other quantitative assessment of exposure to non-severe hypoglycemia.

[0065] In some cases, the Time Below LL Identifier 216 may output {I<thresh(LL),d} as:Dexcom Docket No. 0991-PCT01 a. True for any day d in which several (e.g., 3) consecutive elements of EGV History are below a second threshold (e.g., 54 mg / dL). False for all other days. b. In some implementations, I< thresh(LL),d could be computed as an indicator of some other quantitative assessment of exposure to non-severe hypoglycemia.

[0066] In some cases, the Reference Blood Glucose Calculator 218 may output RefBG as: a. The value of EFBGd on the day d for which RefBG is to be rendered. This implementation would drive the Adjusted Dose toward one that achieves a specified target fasting blood glucose. b. The value of LBGRd on the day d for which RefBG is to be rendered. This implementation would drive the Adjusted Dose toward one that achieves a specified Low Blood Glucose Risk.Insulin Module

[0067] As illustrated in FIG. 3, the Insulin Module 120 can receive or read in the Recorded Dose History during the titration process and then renders a sequence of Daily Effective Dose History values that account for the pharmacokinetic accumulation of available insulin in blood plasma over time. The Daily Effective Dose History is used subsequently as independent variables in the effective dose-response produced by the Dose-EFBG Modeler and the Dose-LBGR Modeler.

[0068] The Insulin Module 120 can perform two main steps. First, a Daily Dose Parser 230 can interpret the Recorded Dose History {RDi, RD2, ..., RDtr] and produce a Daily Dose History {DDi, DD2, . . ., DDM }, as shown in FIG. 3. Here, each element of Recorded Dose history RDr(for r = 1 , 2, . . ., tr) is the r-th recorded (long-acting) insulin dose, described by a recorded dose value and timestamp. For the Daily Dose History, each element at a specific timestamp can be associated to an abstract titration day, such that DDd (for d = 1, 2, . . ., t-1) is the historical total basal dose received on the d-th day of the titration process and where t-1 refers to the most recent injection prior to the one that the system is about to recommend.

[0069] Next, a Dose Accumulator 240 can process the Daily Dose History to produce an associated Daily Effective Dose History {EDi, ED2, . . ., EDt-i }, where EDd (for d = 1, 2, . . ., t-1) is the effective dose for the patient on the d-th day of titration representing the total available (effective)Dexcom Docket No. 0991-PCT01 plasma insulin serving the patient on that day. The effective dose can (i) account for accumulation of insulin in the patient’s body from all previously administered doses, (ii) apply to long-acting insulin with pharmacokinetic profiles measured in days (from half a day to 30 days), (iii) be used for the purpose of quantifying the relative effect of past insulin doses on a specific time window (e.g., once a day), and / or (iv) be used for various patient-specific dose-response models.

[0070] Some implementations of the Insulin Module 120 are described below.

[0071] In some implementations, the elements of the Recorded Dose History may be collected via a “Connected” or “smart” insulin pen, user reported boluses through an “app,” or an insulin injection detection algorithm (e.g. CGM based, or any combination of devices).

[0072] The time window over which the dosed insulin history may be obtained can vary. For instance, it can be a time window of 12 hours in the case of 2 injections per day, a time window of 1 day for daily injections (as described in the example above), or a time window of 1 week for weekly injections.

[0073] The total basal dose may be calculated by the Daily Dose Parser 230 as the result of a single injection of basal (long-acting) insulin on day d, as the sum of multiple doses on day d, or as the weighted sum of multiple doses on day d (weighted with a time degradation coefficient, such that the further in past the dose was injected the lower the weight).

[0074] In some implementations the computation of the effective dose may be based on a linear model of the effect of daily doses on day d from the last N days, derived from a PK / PD model, derived from patient-specific or population data, or learned online or predetermined.Reported Hypoglycemia (Hypo) Module

[0075] As illustrated in FIG. 4, the Reported Hypo Module 130 can receive or otherwise read in the CGM History and Reported Hypo History and render the Hypo Event History. The patient can generate each record in the Reported Hypo History, and thus these records might contain certain deficiencies. For example, these records might refer to a Hypo Event in the distant past, or they might report a Hypo Event when none, in fact, occurred physiologically. The Reported Hypo Module 130 filters the Reported Hypo History down to a Hypo Event History in which each Hypo Event is both relevant and credible for the next Adjusted Dose. Hypoglycemia is a primary risk of basal insulin therapy, and these Hypo Events assist other modules of the system in mitigating adverse hypoglycemic episodes.Dexcom Docket No. 0991-PCT01

[0076] As shown in FIG. 4, the Reported Hypo Module 130 can include (1) a CGM History Parser 250 that can read the CGM History and produce an EGV History { EGV i, EGV2, . . . , EGVk } , (2) a Hypo Report Parser 260 that can read the Reported Hypo History and produce a Hypo Report History {HRi, HR2, HRn}, and (3) a Hypo Report Filter 270 that can combine the EGV History and the Hypo Report History to render a Hypo Event History {Hi, H2, . . Hm}. Each element HRi of the Hypo Report History can comprise a timestamp of the Hypo Event experience, a timestamp of the Hypo Report submission, and / or an indicator of the Hypo Event’s severity. The elements of the Hypo Report History can be ordered by timestamp of the Hypo Event and nonduplicati ve with each other. The Hypo Report History can be provided as input to the Hypo Report Filter 270, which can determine whether each HR is sufficiently relevant and credible. The Hypo Reports that are sufficiently relevant and credible pass through the Hypo Report Filter 270 and can become part of the Hypo Event History, whose length may vary from empty (i.e., no Hypo Events) to the length of the Hypo Report History (i.e., all Hypo Reports are retained as Hypo Events).

[0077] In some embodiments, the Hypo Report relevance may be found to be irrelevant. For example, the Hypo Report may be deemed irrelevant if its age (relative to the current time) is greater than a prespecified threshold (e.g., 7 days). This threshold determines when Hypo Events lose relevance for the next Adjusted Dose. The Hypo Report may also be deemed irrelevant if the Hypo Event was experienced at a time when the Adjusted Dose was substantially different from the current Adjusted Dose. For example, if the Hypo Event was experienced when the Adjusted dose was 37 U / day, whereas the current dose is only lOU / day, then the Hypo Report might be deemed irrelevant for the next Adjusted Dose. The Hypo Report may also be deemed irrelevant if its severity is sufficiently low.

[0078] In some embodiments, the Hypo Report may be deemed noncredible. For instance, it may be deemed noncredible if its age (relative to the current time) is greater than a prespecified threshold (e.g., 2 days). This threshold determines when Hypo Reports lose credibility due to memory recall limitations, for example. A Hypo Report also may be deemed noncredible if the EGV History is credible and consistently above a certain threshold (e.g., 130 mg / dL) at the time of the Hypo Event, such that the Hypo Event is unlikely to be valid. For example, the patient might key in the wrong date for a Hypo Event, and the EGV History may pick out such errors.Dose-EFBG ModelerDexcom Docket No. 0991-PCT01

[0079] As shown in FIG. 5, the Dose-EFBG Modeler 140 can receive a history of Estimated Fasting Blood Glucose values as well as a history of Effective Doses of insulin. The modeler can output a Dose-EFBG regression model that can predict the Estimated Fasting Blood Glucose that corresponds to an Effective Dose. The Dose-EFBG model may also include an assessment of how variable this prediction may be.

[0080] FIG. 5 shows the inputs and outputs of the Dose-EFBG Modeler 140 produced during day t. In some implementations, the Dose-EFBG Modeler 140 can receive two streams of inputs at day t:1. A daily history of Estimated Fasting Blood Glucose values (EFBGa) for days d=2,. . ,,t, as generated by the CGM Module.2. A daily history of Estimated Effective Insulin Doses (EDa) for day d for days d=l,2,...,t-l, as generated by the Insulin Module.

[0081] Based on these inputs, in one embodiment, the Effective Dose-EFBG Response Modeler 280 can produce a linear regression model with intercept Iefbgand coefficient Cefbgrelating Estimated Fasting Blood Glucose to an Effective Dose of insulin.

[0082] To bias the regression model toward a safe population average dose response in early stages of the titration process, the linear regression objective can be modified by adding a regularization term as follows:based on a normally distributed population model for coefficient Cefbg with mean gcand standard deviation crc. The population model can be based on historical observed values for Cefbgand guides model fitting when the histories of EFBGs and EDs are too short to produce reliable regression models. The Effective Dose - EFBG Response Modeler 140 can also produce an assessment of fitness Fefbg of the regression model.

[0083] The Effective Dose - EFBG Response Uncertainty Modeler 290 can calculate the variance of the residuals produced by the Effective Dose-EFBG Response Modeler. It also can apply a constant learning rule to update its estimate of variance (sefbg).

[0084] In one embodiment, the Estimated Fasting Blood Glucose can be estimated from CGM data by considering the estimated glucose value at a particular time of day (e.g., about 5:30 AM). Some implementations may account for potential early meals by selecting the last CGM reading before aDexcom Docket No. 0991-PCT01 detected meal. In some implementations alternatives to the EFBG may be used that are more descriptive of the steady-state of the patients. These implementations may include the median or another percentile of the patient’s glucose distribution or a model of the glucose distribution.

[0085] Various embodiments of the Effective Dose-EFBG Response Modeler 280 may employ non-linear regression models, machine-learning methods, including neural-networks and may incorporate additional predictors in addition to Effective Doses. Also, the population model used in regularization can be multi-variate and include all regression model parameters. Likewise, historical CGM data may be used to produce a Digital Twin or Clone of the patient and simulate this model as an alternate method to produce an Estimated Fasting Blood Glucose value corresponding to an Effective Dose.

[0086] Various embodiments of the Effective Dose - EFBG Response Uncertainty Modeler 290 may employ a variance model that changes with the size of Effective Dose and / or a variance model that is dependent on other features.Dose-LBGR Modeler

[0087] As shown in FIG 6, the Dose-LBGR Modeler 150 can receive a history of Low Blood Glucose Risk values as well as a history of Effective Doses of insulin. The modeler can output a Dose-LBGR regression model that can predict the Low Blood Glucose Risk that corresponds to an Effective Dose. The Dose-LBGR model can also include an assessment of how variable this prediction may be, as well as an estimate of the insulin dose that would achieve the maximum hypoglycemia threshold, called Dose at Low BG Guardrail

[0088] FIG. 6 shows inputs and outputs produced during day t by the Dose-LBGR Modeler 150. The Dose-LBGR Modeler 150 can receive several streams of inputs at day t. One input stream can represent a daily history of Low Blood Glucose Risk values (LBGRd) for days d=l,. . .,t-L as generated by the CGM Module. In one embodiment, Low Blood Glucose Risk can be defined as the 0.5 percentile of the patient’s CGM values. Another input stream can represent a daily history of Estimated Effective Insulin Doses (EDa) for day d for days d=l,2,. . ,,t-l, as generated by the Insulin Module.

[0089] Based on these inputs, the Effective Dose-LBGR Response Modeler 310 can produce a linear regression model with intercept Iibgr and coefficient Cibgr relating Low Blood Glucose Risk to an Effective Dose of insulin.Dexcom Docket No. 0991-PCT01

[0090] To bias the regression model toward a safe population average dose response in early stages of the titration process, the linear regression objective can be modified by adding a regularization term as follows:based on a normally distributed population model for coefficient Cibgr with mean icand standard deviation crc. The population model can be based on historical observed values for Cibgrand guides model fitting when the histories of LBGRs and EDs are too short to produce reliable regression models. The Effective Dose-LBGR Response Modeler 310 can also produce an assessment of fitness i'2n,2rof the regression model as well as Dose at Low BG Guardrail. The latter is the maximum dose that may not result in the likelihood of hypoglycemia being larger than a certain percentage (e.g., 5%).

[0091] The Effective Dose-LBGR Response Uncertainty Modeler 320 can calculate the variance of the residuals produced by the Effective Dose-LBGR Response Modeler 310. It can also apply a constant learning rule to update its estimate of variance (sibgr).

[0092] In some implementations, alternative definitions of Low Blood Glucose Risk may be employed which could include a Low Blood Glucose Index (LBGI) or other percentiles.

[0093] In some implementations, the Effective Dose-LBGR Response Modeler 310 may employ non-linear regression models, machine-learning methods, including neural-networks, and / or additional predictors in addition to Effective Doses. In addition, a population model may be used in regularization that is multi-variate and includes all regression model parameters. Also, historical CGM data can be used to produce a Digital Twin or Clone of the patient and simulate this model as an alternate method to produce a Low Blood Glucose Risk value corresponding to an Effective Dose.

[0094] Alternative embodiments of the Effective Dose-LBGR Response Uncertainty Modeler 320 may employ a variance model that changes with the size of Effective Dose and / or a variance model that changes with other features.Target Dose AdjusterDexcom Docket No. 0991-PCT01

[0095] As shown FIG. 7, the Target Dose Adjuster 160 can receive as inputs the Dose-EFBG Model from the Dose-EFBG Modeler 140 and the Dose-LBGR Model from the Dose-LBGR Modeler 150 and output a Target Dose resulting from a tradeoff between efficiency and safety. The Target dose can be provided to the Dose Finalizer 170.

[0096] The submodules of the Target Dose Adjuster 160 can perform several primary steps. The various steps can be performed in parallel, sequentially, or in any combination thereof.

[0097] In the first step, an efficient Target Dose Calculator 330 can take the Dose-EFBG Model as input and produce an Efficient Dose from an inversion of the model. The Dose-EFBG Model can contain the elements necessary to (i) predict an Estimated Fasting Blood Glucose value and the error around it from a basal insulin dose and / or (ii) assess the quality of the predictions. In some implementations of the Dose-EFBG Modeler 140, the prediction can be performed with a linear model with a coefficient Cefbg and intercept lefbg , the error around the prediction can be described as having a percentage (e.g., 68%) chance to fall within sefb of the prediction, and the quality of the predictions can be assessed with the r-square value (r^fbg). If the quality of the predictions is deemed too poor (r^fbg to low), no Efficient Dose is produced. Otherwise, the Efficient Dose can be the dose that satisfies the desired Estimated Fasting Blood Glucose with a certain level of confidence (accounting for the prediction error). In some implementations, the Efficient Dose can be the dose based on sefbg. The patient’s actual response can be equally likely to lie above or below the target fasting blood glucose.

[0098] In the second step, a Safe Target Dose Calculator 340 can take the Dose-LBGR Model as input and produce a Safe Dose from an inversion of the model. The Dose-LBGR Model can contain the elements necessary to (i) predict a Low Blood Glucose Risk value and the error around it from a basal insulin dose and / or (ii) assess the quality of the predictions. In some implementations of the Dose-LBGR Modeler 150, the prediction can be performed with a linear model with a coefficient Cibgr and intercept Iibgr, the error around the prediction can be described as having a percentage (e.g., 68%) chance to fall within sibgr of the prediction, and the quality of the predictions can be assessed with the r-square value (i^ibgr). If the quality of the predictions is deemed too poor (r2^ to low), no Safe Dose is produced, otherwise, the Safe Dose can be the dose that satisfies the desired Low Blood Glucose Risk with a certain level of confidence (accounting for the prediction error). In some implementations, the Safe Dose can be the dose, based on sibgr. that yields a percentage (e.g., 95%) chance that the actual LBGR response corresponds to hypoglycemia exposure below a defined threshold.Dexcom Docket No. 0991-PCT01

[0099] In the third step a Dose Reconciliator 350 can take the Efficient Dose and the Safe Dose as inputs and produce a Target Dose. The Target Dose can be the result of a tradeoff between the Efficient Dose and the Safe Dose. In some implementations the Target Dose can be equal to the Safe Dose.

[0100] In some implementations the Target Dose Adjuster 160 may use any of the various embodiments of the Dose-EFBG Model and the Dose-LBGR Model described above. Also, the level of confidence that is used can be set to any suitable value.

[0101] In some embodiments, the Dose Reconciliator 350 may employ a minimum of the Efficient Dose and the Safe Dose, a maximum of the Efficient Dose and the Safe Dose or a weighted average of the Efficient Dose and the Safe Dose, with weights computed from model accuracy and / or prediction error.Dose Finalizer

[0102] As shown in FIG. 8, inputs to the Dose Finalizer 170 can include the Target Dose, outputs from the CGM Module 110 ({I<Thresh(L),d}, {I<Thresh(LL),d}, and RefBG), the Hypo Event History {Hi, H2, . . ., Hm], the most recent Adjusted Dose (RDM), and / or the Dose at Low BG Guardrail DLBGGM. The module can apply a set of safeguards and outputs a new Adjusted Dose.

[0103] The Dose Finalizer 170 can apply a set of safeguards to a proposed Target Dose to mitigate hypoglycemia risk and prevent large dose changes with respect to the current Adjusted Dose RDM. In this regard the Dose Finalizer 170 can define a Target Deadband surrounding a target glucose value, which is a design parameter of the algorithm. The Target Deadband can be a range of glucose values within which the Adjusted Dose does not change. The individual submodules of the Dose Finalizer 170 can operate as follows.

[0104] The Consistency Verifier 360 can apply a set of rules to ensure the proposed Target Dose responds to the clinical state of the patient. For instance, it may set the Adjusted Dose to the current dose RDM if RefBG is in the Target Deadband. It may also set the Adjusted Dose to the current dose RDM if RefBG is below the Target Deadband and the Target Dose is greater than the current dose RDM. The Consistency Verifier 360 may also Set the Adjusted Dose to a value greater than RDM if RefBG is above the Target Deadband.

[0105] The Hypo-Based Guard 370 can ensure that the resulting Adjusted Dose does not produce hypoglycemia, e.g., by a percentage (e.g., 95%) of the time, by limiting the Adjusted Dose =Dexcom Docket No. 0991-PCT01 min(Adjusted Dose, DLBGGt i). This module can further prevent any dose increases in the presence of recent indications from {I<Thresh(L)70,d} or {I<Thresh<LL),d} or the Hypo Event History {Hi, H2, ..., Hn,}.

[0106] The Maximum Change Guard 380 can support a smooth change of doses by setting a Maximum Change. If the proposed Target Dose may change the current Adjusted Dose RDn by more than the Maximum Change, the maximum change can be applied instead by setting Adjusted Dose = RDM + Maximum Change.

[0107] In some implementations, the Overbasalization Check 390 can create a warning if the Target Dose is beyond clinically acceptable limits of basal insulin therapy for the patient. This is a prediction that overbasalization may occur in the future. It also can create a warning if the Adjusted Dose produced by the Maximum Change Guard is beyond clinically acceptable limits of basal insulin therapy for the patient. This is an indication that overbasalization is currently present.

[0108] In some implementations, the Dose Finalizer 170 may employ any of a variety of different rules for establishing maximum allowable changes. In some implementations the Hypo-Based Guard 370 can be based on risk or other hypoglycemia related metrics. In some implementations, overbasalization can introduce a constraint on the dose, as opposed to simply being a warning.Termination Checker

[0109] As shown in FIG. 9, the Termination Checker 180 can include a Termination Checker Submodule 410 that can receive as inputs the Estimated Fasting Blood Glucose History, Hypo Event History, Low Blood Glucose Risk History, Time Below L Indicator History, Time Below LL Indicator History, Reference Blood Glucose (RefBG), Daily Effective Dose History and / or the Recommendation History and can output the Termination Report. As shown in FIG. 1:• Estimated Fasting Blood Glucose History refers to the output of CGM Module• Reported Hypo Event History refers to the output of Reported Hypo Module• Low Blood Glucose Risk History refers to the output of CGM Module• CGM Below L Indicator History refers to the output of CGM Module• CGM Below LL Indicator History refers to the output of CGM Module• Reference Blood Glucose RefBG refers to the output of CGM ModuleDexcom Docket No. 0991-PCT01• Daily Effective Dose History refers to the output of Insulin Module

[0110] The Recommendation History that can be used as an input, where each element of Recommendation History Rd (for d = 1, 2, . . t) can be the recommendation element for the d-th day, is described by both an Adjusted Dose (output of Dose Finalizer) and a Termination Report (output of Termination Checker). Likewise, the termination Report that can be used as an input, can contain the termination information such as if a termination took place and if yes, the timestamp and reason of the termination.

[0111] The Termination Checker 180 can perform several primary steps. In the first step, the Termination Checker Submodule 410 can receive the various inputs and output a Termination Report. In some implementations, the Termination Checker Submodule 410 can check what elements in a given list of termination criteria are verified. The Termination Report can returns “yes” if any criteria is verified and if “yes”, it can specify which ones. In some implementations, several different criteria can terminate the titration process. For example: a. A number of past elements (e.g., 7) of the Estimated Fasting Blood Glucose History are within the desired target range and the past elements of Time Below L Indicator History are below a desired target and the past elements of Time Below LL Indicator History are below a desired target and there are no elements in the Hypo Event History with a timestamp within the past (e.g., 7) days. b. A number of past Adjusted Doses (e.g., 7) obtained from the last elements of Recommendation History are equal to zero. c. The Termination Deadline has been reached: which can be a predetermined termination deadline (e.g., 30 days).

[0112] In the second step, the Termination Reporter 420 can receive a Termination Report and Recommendation History as inputs from the Termination Checker Submodule 410. The Termination Reporter 420 can return the last Termination Report from Recommendation History if this one indicates that a termination took place and return the Termination Report from the Termination Checker 180 otherwise.

[0113] In some implementations, the Termination Checker Submodule 410 may use different inputs. For instance, termination decisions can be based on alternative CGM or insulin dose statistics that can be computed by the CGM or Insulin Modules, respectively. The TerminationDexcom Docket No. 0991-PCT01Checker Submodule 410 also may use alternative glycemic or insulin metrics that more broadly reflect diabetes management criteria (e.g. time in range, estimated HbAlc).

[0114] Likewise, different embodiments may use different termination criteria on which to make termination decisions. For instance, termination decisions may be based on alternative statistics related to the patient’s experience of hypoglycemia that can be computed by the Reported Hypo Module 130. In some implementations, termination decisions may be based on patient satisfaction or caregiver satisfaction. In some implementations, the termination criteria may use different assessment windows or any metric reaching the desired target range for any amount of time.

[0115] In some implementations, the Termination Report may be based on the system’s inability to reach a target RefBG without exposure to hypoglycemia within a desired number of titration days. In this case the report may indicate that variability in the CGM trace, regardless of whether it is due to the patient’s physiology or behavior, limits the ability to reach a target RefBG without also introducing prandial insulin, e.g., multiple daily injection therapy.Reference Blood Glucose State Parameter

[0116] Some implementations of the basal insulin titration system described herein may employ a variety of different blood glucose state parameters that may be quantified and output by the CGM module 110 in a number of different ways. For example, in some embodiments, the use of the Estimated Fasting Blood Glucose (EFBG) state parameter may be eliminated and an alternative reference blood glucose (RefBG) state parameter can be used instead. As used herein, the term "reference blood glucose" (“RefBG”) generally refers to any BG metric that can serve as an index of the quality of hyperglycemia management via basal insulin therapy that can be computed without the knowledge or assumption that the patient is in a steady fasting state at any particular time.) In part, this embodiment avoids the use of EFBG because it has been found that, in practice, due in part to patient behavior (e.g. the patient may not actually be in a fasting state when BG is stable), it is highly variable from day to day and often does not represent the lowest blood glucose value of the day.

[0117] Thus, in some embodiments, the reference blood glucose (RefBG) may be used by the Dose Finalizer 170 to produce the Adjusted Dose. That is, the Dose Finalizer 170 may produce an Adjusted Dose that produces a desired value of the reference blood glucose (RefBG) . The reference blood glucose (RefBG), which in some cases may be correlated with EFBG, is advantageous because it serves as a stable metric that generally does not have as much variabilityDexcom Docket No. 0991-PCT01 as the EFBG. Moreover, unlike the EFBG, the reference blood glucose (RefBG) does not depend on the time of day. Accordingly, the reference blood glucose (RefBG) may be used as a representation of a users’ typical blood glucose level and can serve as a substitute for EFBG from a titration perspective.

[0118] The reference blood glucose (RefBG) may be quantified, for example, as the Xth percentile of daily (e.g., time series) CGM values, where X can include 10, 11, 12 or other percentiles. Reference blood glucose RefBG computed in this way (i) is reflective of the quality of hyperglycemia management (and is, by definition, not the lowest blood glucose value of the day), (ii) is such that the its value can be realized at any time of the day (and is therefore independent of when meals are taken), and (iii) has been observed to have lower day-to-day variability than average blood glucose computed at any fixed time or range of times of the day. The percentile that is selected should be a value that (i) reduces the average EBG to an appropriate level and (ii) meets glycemic goals with the smallest possible insulin dose. In some implementations, the Adjusted Dose determined by the Dose Finalizer 170 can be selected so that the reference blood glucose (RefBG) is maintained within some range of values, e.g., 87-120 mg / dL, while also reducing the risk of hypoglycemia.

[0119] In addition to the use of the reference blood glucose (RefBG), some embodiments of the basal insulin titration system may also employ a second percentile value of daily CGM values, referred to as the daily Low Blood Glucose Percentile (LBGP), which may be used instead of the Low Blood Glucose Risk (LBGR) described above. The percentile that is selected can be a value that exposes the user to a significant risk of hypoglycemia e.g., an EBG value of less than 70mg / dl. For example, the LBGP may be quantified as 1stpercentile of daily CGM values, while the RefBG may be quantified as the 12thpercentile of daily CGM values. Similar to the LBGR, the LBGP can be used to reduce risks from low blood glucose (e.g., hypoglycemia). However, the use of an LBGP may be advantageous because unlike the LBGR, the LBGP is a direct measurement that does not need to be converted into a risk metric. That is, the use of the LBGP allows the titration algorithm to operate directly on low EBG values without the need to convert any such values to a risk metric.

[0120] FIG. 10 shows a functional block diagram of a basal insulin titration system similar to that shown in FIG. 1, except that it employs the reference blood glucose (RefBG) and the daily Low Blood Glucose Percentile (LBGP) as described above instead of the EFBG and the LBGR. In this embodiment only a single dose-blood glucose state response modeler, which in this case is a Dose- LBGP modeler 145, is employed. That is, the Dose-LBGP modeler 145 replaces the Dose-EFBG Modeler 140 and Dose-LBGR Modeler 150 used in the basal insulin titration system of FIG. 1.Dexcom Docket No. 0991-PCT01Termination Checker

[0121] Some illustrative examples of the Termination Checker 180 discussed above employ a limited set of criteria to determine when the titration process should be terminated. For instance, one rule may specify that the titration process should terminate if the seven past Adjusted doses are zero. Another rule may specify that the titration process should terminate after 30 days if the titration process has not yet finished. Yet another rule may specify that the termination process should terminate if the EFBG remains in range within a predetermined number of days without exposure to hypoglycemia. In the examples presented below additional functions and features of some embodiments of the Termination Checker 180 are discussed.

[0122] In contrast to the previously described examples of the Termination Checker 180 in which termination is effectively defined as the deterministic convergence of the EBG (or the RefBG) to a healthy range of values, the examples of the Termination Checker 180 discussed below implicitly acknowledge the statistical nature of the titration problem. The Termination Checker 180 can accomplish this by taking an evidence-based approach to the problem, declaring termination only when it has gathered sufficient evidence.

[0123] In some embodiments of the basal insulin titration system described herein the Termination Checker 180 can use as inputs the reference blood glucose RefBG, the Low Blood Glucose Percentile (LBGP), the indicators of low and very low blood glucose 7<Thresh(L) and / <Thresh(LL), the effective dose ED and / or the patient’s body weight. Based on these inputs the Termination Checker 180 can determine whether termination should occur positively (positive termination) or negatively (negative termination). Positive termination may indicate that the titration system has found a satisfactory Adjusted Dose. Negative termination may indicate that no satisfactory Adjusted Dose has been found and is unlikely to exist.

[0124] Negative termination may arise for any of a number of reasons. For example, the titration system may have determined that regardless of the dose that is recommended, the EBG (or the RefBG) remains unacceptably high or the transition from being above an upper EBG limit to being in range also comes with exposure to low EBG values because of natural variability. For instance, the termination system may determine that there are relatively high doses for which the reference blood glucose RefBG is satisfactory but the Low Blood Glucose Percentile (LBGP) is unacceptably low at those doses. All of these reasons for terminating negatively may indicate that treatment of the user with basal insulin alone is not sufficient and that the user may require the addition ofDexcom Docket No. 0991-PCT01 prandial insulin. That is, negative termination can provide an important indicator that prandial insulin may now need to be added.

[0125] In some embodiments the Termination Checker 180 may determine whether there is sufficient evidence to support positive or negative termination and if not, the Termination Checker 180 can cause the Dose Finalizer 170 to produce a number of candidate doses that are selected for the purpose of collecting measurements that would support a termination decision in a subsequent iteration of the titration process. That is, as a termination decision approaches and if the evidence supporting the decision is not sufficient, the Termination Checker 180 can determine that alternative doses surrounding the Adjusted Dose should be produced in subsequent iterations to provide further evidence supporting the termination decision. In this way the Termination Checker 180 can effectively direct and change the operation of the Dose Finalizer 170 to guide the search for alternative potential Adjusted Doses by controlling or configuring the Dose Finalizer 170 to take into account not only the Target Doses from the Target Dose Adjuster 160 but also the candidate doses from the Termination Checker 180. The evidence that is produced in support of the termination decision may be included in the Termination Report generated by the Termination Checker 180.

[0126] When the Termination Checker 180 determines with sufficient evidence that termination should occur positively, it may recognize that there may be more than one satisfactory Adjusted Dose that could be produced that gives rise to an EBG or a RefBG that remains within a desired range. If this is the case the titration system can generally select the smallest satisfactory Adjusted Dose.Continuous Tracking of Events and Internal Statistics

[0127] In some embodiments, time can be accounted for as a discrete quantity (i.e., a calendar day). For example, the state parameters produced by the CGM Module 110 such as the Estimated Fast Blood glucose (EFBG) and the like and the Effective Dose (ED) produced by the Insulin Module 120 are shown in one example as being tracked on a daily basis with daily timestamps. In these embodiments the subsequent analysis of the data by the titration system can be performed on integer values of time corresponding to calendar days, with the transition between days occurring at midnight. However, tracking time in this way may limit the effectiveness of the titration system.

[0128] For example, it is assumed that the patient is complying with the daily dose recommendations produced by these embodiments of the titration system and taking the daily doseDexcom Docket No. 0991-PCT01 recommendation at the same time every day. However, in practice the patient may not take the recommended dose at the same time every day, but may take it earlier or later. If, for instance, the patient takes the recommended dose (e.g., several hours) late, it is unclear if it should be treated as being taken on the day it was recommended or on the next day. However, neither option is completely satisfactory. As a result, accounting for time as a discrete quantity defined by calendar days is too coarse for tracking the patient’s compliance with dosage recommendations.

[0129] To address this problem, in some implementations, time may be accounted for by treating time as a continuous variable and tracking the occurrence of all discrete events (e.g., dose recommendations, dose injections) and all internal statistics (e.g., low-BG percentiles and effective insulin doses) with a continuous time stamp. This addresses the problem discussed above concerning the timing of recommended doses and the time at which the patient complies with the recommendation. That is, by treating time as a continuous variable, the titration system is capable of adapting to patient behavior by tracking insulin injections regardless of the time of day in which they are performed, even if doses are split across multiple injections. It also allows the titration algorithm to be invoked at any time of day and not necessarily at a single predetermined time each day. This approach can simplify handling of practical difficulties such as patients traveling across time zones and / or manually changing the time zone on their device. Both of these actions can alter the definition of calendric midnight-to-midnight day; however, such changes are transparent to an approach that treats time as a continuous variable without calendric demarcations.Dose Adapter

[0130] The ability to track events and internal statistics using a continuous timestamp instead of an integer timestamp based on discrete days allows the basal insulin termination system to provide improved features and additional features and functionality. One example of such functionality can be provided by a module that is referred to herein as a Dose Adapter, which, in addition to titrating, or instructing, an ideal daily basal insulin dose that is optimal under the assumption that the daily dose is taken at the same time (or times) every day, is able to provide next dose advice that adapts the ideal daily dose to the recent history of basal injections. In some embodiments, the Dose Adapter may be used by the patient to provide an adapted adjusted dose to the patient at any time the patient requests the Dose Adapter to provide one.

[0131] As shown in the functional block diagram of FIG. 11, the Dose Adapter may be incorporated in a Dose Adapter System 600 that can incorporate the functionality performed by theDexcom Docket No. 0991-PCT01 basal titration system 610 (e.g., basal titration system shown in FIGs. 1 or 10) and the Dose Adapter 620 as described herein. In some implementations, the Dose Adapter may be a separate and independent system or application that receives data such as the ideal daily basal insulin dose from the basal titration system 610.

[0132] For example, the Dose Adapter can be able to address the problem noted above in which a patient does not take a daily Adjusted Dose at the same time each day as the titration system specifies, but takes it earlier or later than recommended, or splits it over multiple injections or fails to take it altogether. The Dose Adapter can accomplish this by fine-tuning the daily Adjusted Dose from the Dose Finalizer 170 using the titration algorithm and the timestamp specifying when the patient actually took the previous dose. Accordingly, if the patient takes the daily Adjusted Dose at the recommended time each and every day, the Adapted Adjusted Dose provided by the Dose Adapter can be the same as the daily Adjusted Dose provided by the Dose Finalizer 170. If, alternatively, the patient does not take the daily Adjusted Dose as specified, the Dose Adapter can determine the dose that should be taken at the time the patient invokes it, taking into account when the previous dose was taken and the daily Adjusted Dose. That is, if the patient takes a daily Adjusted Dose 8 hours early, the Dose Adapter is able to modify the daily Adjusted Dose to take into account that it has been 32 hours since the last dose was taken. Conversely, if it has only been, e.g., 17 hours, since the last daily Adjusted Dose was taken, the Dose Adapter can reduce the Adjusted Dose to take into account that it has been less than 24 hours since the last dose was taken.

[0133] In some implementations, the Dose Adapter may be controlled or configured to perform a secondary titration using the titration algorithm employed by the basal insulin titration system. The Dose Adapter may define a variable based on the patient’s basal insulin on board (BIOB). The BIOB may be treated as being zero if the patient is taking the daily Adjusted Dose at the recommended time. The BIOB may be treated as being negative if the patient is taking the Adjusted Dose later than the recommended time or misses a dose altogether or if the dose that is actually taken is smaller than recommended. Alternatively, if the patient is taking the daily Adjusted Dose earlier than the recommended time or takes a dose that is larger than recommended, the BIOB may be treated as positive. The variable that is related to BIOB may then be used in the secondary titration to determine an adapted Adjusted dose that the patient should take at the time the patient invokes the Dose Adapter.

[0134] In some implementations, the Dose Adapter can provide dose recommendations that minimize the difference between the anticipated future effective dose values and the ideal doseDexcom Docket No. 0991-PCT01 recommendation of the dose finalizer, which can be the ideal daily dose computed under the assumption that the patient injects basal insulin at the same times every day.

[0135] In some embodiments, the Dose Adapter may be invoked by the user on an on-demand basis so the user can determine the dose that should be taken at that very time based on when the previous dose was actually taken. The output of the Dose Adapter (e.g., the next recommended dose, which may include an amount and timing for the next recommended dose of insulin or other medicament) may be provided to the user and / or to a medicament delivery device to control an amount of medicament delivered to the user of the medicament delivery device.

[0136] FIG. 12 shows one example of an illustrative operating environment in which the basal titration adjustment system and method described herein may operate. Of course, those of ordinary skill in the art will recognize that the basal insulin titration adjustment system and method may be used in other environments as well and in some cases may even operate as an independent device or, alternatively, the method may be implemented on any suitable computing device that incorporates additional functionality. As shown, the environment can include an insulin delivery device 510, a glucose monitor 520, a processor (e.g., a processor module, which may include one or more processing units and / or processing unit cores) 530, a subject 540, an activity monitor 550, and a smartphone 560.

[0137] One or more of the insulin delivery device 510, the glucose monitor 520, the processor 530, the activity monitor 550, and the smartphone 560 may be in communication through a network. The network may be a variety of network types including the public switched telephone network (PSTN), a cellular telephone network, and a packet switched network (e.g., the Internet). Although only insulin delivery device 510, one glucose monitor 520, one processor 530, one subject 540, one activity monitor 550, and one smartphone 560 are shown in FIG. 1 , there is no limit to the number of insulin delivery devices 510, glucose monitors 520, processors 530, subjects 540, activity monitors 550, and smartphones 560 that may be supported.

[0138] The insulin delivery device 510 may be an example of a medicament delivery device, and may include any device that dispenses insulin, such as syringes, smart pens, pumps (e.g., external, mechanical, patch, or implanted), and inhalers, for example. The insulin delivery device 510 may also include devices that dispense other drags that help control glucose levels like glucagon (dualhormone artificial pancreas), GLP-1, etc. Depending on the particular type of device employed, the insulin delivery device may or may not be a connected device that is in communication through a network.Dexcom Docket No. 0991-PCT01

[0139] The glucose monitor 520 may be any type of CGM or SMBG (self-monitoring of blood glucose) device, depending on the implementation. The glucose monitor 520 may 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 glucose readings, the glucose monitor 520 may record user interactions such as when and how a user (e.g., a subject, a patient, a caregiver, a medical professional, etc.) views their glucose traces and how they respond to alerts and alarms. The user interaction can provide insights into the timing and motivation of treatment decisions including why and when they are considering the effects of meals or dosing insulin.

[0140] The processor 530 collects data from the insulin delivery device 510 and the glucose monitor 520 and the subject 540 and, in some embodiments, may run the basal insulin titration methods described herein. The processor 530 (as well as the insulin delivery device 510, the glucose monitor 520, the activity monitor 550, and / or the smartphone 560) may be implemented using a variety of computing devices such as smartphones, desktop computers, laptop computers, and tablets, for example. In this regard the processor 530 may be a processor that is available in one of the other devices shown in the operating environment of FIG. 12, such as the smartphone 560, the activity monitor 550 or, depending on the type employed (e.g., a smart pen), the insulin device 510. Other types of computing devices in which the processor 530 may be incorporated also may be supported. A suitable computing device is illustrated in FIG. 13 as the computing device 1300.

[0141] To provide a robust system, the calculations performed by the processor 530 may be dynamic and distributed depending on which devices and processors are connected. For example, cloud computing may be used when there is connectivity, a smartphone processor may be used when there is no connectivity and then a transmitter or a smartwatch may be used when the smartphone is not connected. Complex calculations, like model optimization, may only ran when more powerful processors are available. When powerful processors are not available, algorithms may use the most recent parameters or simpler approximations.

[0142] The subject 540 can provide inputs to the system including information about meals, activity, and diabetes treatments (e.g., basal insulin injections) using, for instance, any computing device that is in communication with the system, such as the smartphone 560 or other computing device of the subject 540. These inputs can be user-initiated or prompted by the system. These inputs can describe current, previous, and / or upcoming events.

[0143] The activity monitor 550 may be any device that monitors the user's physiologic and mental state. One example is a fitness tracker that monitors activity, exercise, and sleep withDexcom Docket No. 0991-PCT01 accelerometers, gyroscopes, heart rate, and oxygen sensors. This can also include smartphones as they can detect location and user activity / interactions, or a smart home device (e.g., Amazon Alexa). In some implementations, the activity monitor 550 may include devices that detect meals.

[0144] The smartphone 560 can be used as an activity and context monitor, data entry device, data collection (talking to devices with Bluetooth, NFC, Wi-Fi, etc.) and run applications (e.g., apps) that estimate nutritional information through manual entry or automated entry (e.g., photos).

[0145] In some embodiments the methods described herein for titrating basal insulin may be implemented as a dedicated application that may reside in whole or in part on any suitable device or combination of devices such as any of the devices shown FIG. 12 (e.g., the smartphone 560, activity monitor 550, glucose monitor 520). Alternatively, these methods may be performed in whole or in part in the cloud (e.g., as a web-based application), or in part in the cloud and in part on one or more devices.

[0146] FIG. 13 shows an exemplary computing device 1300 in which example embodiments and aspects may be implemented. The computing device 1300 is only one example of a suitable computing device and is not intended to suggest any limitation as to the scope of use or functionality. Numerous other general purpose or special puipose computing devices environments or configurations may be used. Examples of well-known computing devices, environments, and / or configurations that may be 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 that include any of the above systems or devices, and the like.

[0147] Computer-executable instructions, such as program modules, being executed by a computer may be used. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Distributed computing environments may be used where tasks are performed by remote processing devices that are linked through a communications network or other data transmission medium. In a distributed computing environment, program modules and other data may be located in both local and remote computer storage media including memory storage devices.

[0148] With reference to FIG. 13, an exemplary system for implementing aspects described herein includes a computing device, such as computing device 1300. In its most basic configuration, computing device 1300 typically includes at least one processing unit 1302 and memory 1304.Dexcom Docket No. 0991-PCT01Depending on the exact configuration and type of computing device, memory 1304 may be volatile (such as random access memory (RAM)), non-volatile (such as read-only memory (ROM), flash memory, etc.), or some combination of the two. This most basic configuration is illustrated in FIG. 13 by dashed line 1306.

[0149] Computing device 1300 may have additional features / functionality. For example, computing device 1300 may include additional storage (removable and / or non-removable) including, but not limited to, magnetic or optical disks or tape. Such additional storage is illustrated in FIG. 13 by removable storage 1308 and non-removable storage 1310. One or more of the removable storage 1308, non-removable storage 1310, and / or memory 1304 may comprise computer-readable storage media storing instructions executable by the at least one processing unit 1302 to perform one or more of the operations described herein. Computing device 1300 may include one or more input devices 1314 and output devices 1316, which may be built-in to the computing device and / or peripherally connected to the computing device. Input devices 1314 may include a keyboard, mouse, pen, trackball, touch screen, microphone, biometric security device (e.g., fingerprint scanner, eye scanner, face scanner, etc.), and / or other suitable component(s) configured to receive, detect, and / or process input (e.g., user input) for use by the computing device 1300. Computing device 1300 may further include one or more output devices 1316, such as a display, audio speaker, haptic feedback device, and / or other suitable component(s) for providing output from the computing device 1300 (e.g., to a user). The input and / or output devices 1314, 1316 may be included in a user interface system of the computing device 1300.

[0150] Computing device 1300 may include one or more communication connection(s) 1312, which enable communication over a communication medium to another computing entity. The communication medium conveys information such as computer-executable instructions, audio or video input or output, or other data in a modulated data signal. A modulated data signal is a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media can use an electrical, optical, RF, or other carrier.Additional Examples

[0151] In a first example, a method of improving glycemic outcomes for a person with diabetes comprises: measuring or determining a first value of a reference parameter associated with the person, wherein the reference parameter is based at least in part on a respective plurality of continuous or intermittent glucose monitoring samples; deriving a target dose for a medicamentDexcom Docket No. 0991-PCT01 based at least in part on a history of continuous or intermittent glucose monitoring; providing a first therapeutic recommendation based on the first value of the reference parameter, the first therapeutic recommendation comprising an adjusted dose of the medicament; measuring or determining a second value of the reference parameter associated with the person; and providing a second therapeutic recommendation based on the first therapeutic recommendation and the second value of the reference parameter, the second therapeutic recommendation comprising an update to the adjusted dose of the medicament.

[0152] A second example includes the first example, and further includes the method, wherein the reference parameter is a reference glucose value.

[0153] A third example includes the first and / or second examples, and further includes the method, wherein the reference glucose value is a value of blood glucose or interstitial glucose.

[0154] A fourth example includes one or more of the first through third examples, and further includes the method, wherein the reference glucose value is interstitial glucose, and further comprising deriving the reference interstitial glucose value from a time series of continuous glucose monitoring (CGM) values.

[0155] A fifth example includes one or more of the first through fourth examples, and further includes the method, wherein the reference glucose value is associated with a 12th percentile of the time series of CGM values.

[0156] A sixth example includes one or more of the first through fifth examples, and further includes the method, wherein the second therapeutic recommendation is further based on at least a portion of the time series of CGM values.

[0157] A seventh example includes one or more of the first through sixth examples, and further includes the method, wherein the first and second therapeutic recommendations include insulin amounts to be delivered.

[0158] An eighth example includes one or more of the first through seventh examples, and further includes the method, wherein the insulin amounts to be delivered are transmitted to a medicament delivery device.

[0159] A ninth example includes one or more of the first through eighth examples, and further includes the method, wherein the medicament delivery device is a pump or smart pen.Dexcom Docket No. 0991-PCT01

[0160] In a tenth example, a method of determining a medicament dose comprises: receiving Continuous Glucose Monitoring (CGM) data for a subject over a period of time: defining a reference blood glucose (RefBG) metric based at least in part on the CGM data, wherein the RefBG metric is a statistic derived from a specified time window of the CGM data; modeling one or more additional metrics of a therapeutic performance of the subject using the RefBG metric to produce one or more models of therapeutic performance; establishing (i) a medicament dose and (ii) a RefBG target value using the one or more models of therapeutic performance; evaluating whether one or more values of the RefBG metric are above the RefBG target value, and, when the one or more values of the RefBG metric are above the RefBG target, avoiding use of the corresponding medicament dose in future updates of the medicament dose; and adjusting the medicament dose to provide a subsequent medicament dose based, at least in part, on the evaluation of the one or more values of the RefBG metric.

[0161] An eleventh example includes the tenth example, and further includes the method, wherein the RefBG metric comprises the Xth percentile of an individual’s or population’s CGM data from the specified time window, and wherein X is selected as a percentile that minimizes inter-day variance or that minimizes a cost function that penalizes both inter-day variance and proximity to true fasting blood glucose samples with weights of the cost function chosen to reflect user preferences.

[0162] A twelfth example includes the tenth and / or eleventh examples, and further includes the method, wherein the medicament comprises insulin and wherein the subsequent medicament dose comprises an adjusted insulin dose.

[0163] A thirteenth example includes one or more of the tenth through twelfth examples, and further includes the method, wherein an indication of the adjusted insulin dose is transmitted to a medicament delivery device.

[0164] A fourteenth example includes one or more of the tenth through thirteenth examples, and further includes the method, wherein the medicament delivery device comprises a pump or smart pen.

[0165] A fifteenth example includes one or more of the tenth through fourteenth examples, and further includes the method, wherein the one or more additional metrics includes Ale, Time in Range, or True Fasting Blood Glucose.Dexcom Docket No. 0991-PCT01

[0166] A sixteenth example includes one or more of the tenth through fifteenth examples, and further includes the method, wherein the initial medicament dose is determined by a health care practitioner or is determined from individual biomarkers.

[0167] In a seventeenth example, a system for determining a medicament dose comprises: a processor; and computer-readable storage media storing instructions executable by the processor to: receive Continuous Glucose Monitoring (CGM) data for a subject over a period of time; define a reference blood glucose (RefBG) metric based at least in part on the CGM data, wherein the RefBG metric is a statistic derived from a specified time window of the CGM data; model one or more additional metrics of a therapeutic performance of the subject using the RefBG metric to produce one or more models of therapeutic performance; establish a RefBG target value using the one or more models of therapeutic performance; for a plurality of historical RefBG values, evaluate whether each respective historical RefBG value is above the RefBG target value, and, when the respective historical RefBG value is above the RefBG target, mark a dose value or values that produced the respective historical RefBG value as “to-be-avoided” in future updates of the medicament dose; and adjust a current medicament dose to provide fixed-size or proportional dose adjustment away from the dose value or values marked as “to-be-avoided.”

[0168] An eighteenth example includes the seventeenth example, and further includes the system, wherein an indication of the adjusted current medicament dose is transmitted to a medicament delivery device.

[0169] A nineteenth example includes the seventeenth and / or eighteenth examples, and further includes the system, wherein the medicament delivery device comprises a pump or smart pen.

[0170] A twentieth example includes one or more of the seventeenth through nineteenth examples, and further includes the system, wherein the one or more additional metrics includes Ale, Time in Range, or True Fasting Blood Glucose.

Claims

Dexcom Docket No. 0991-PCT01Claims1. A method of improving glycemic outcomes for a person with diabetes, comprising: a. Measuring or determining a first value of a reference parameter associated with the person, wherein the reference parameter is based at least in part on a respective plurality of continuous or intermittent glucose monitoring samples; b. Deriving a target dose for a medicament based at least in part on a history of continuous or intermittent glucose monitoring; c. Providing a first therapeutic recommendation based on the first value of the reference parameter, the first therapeutic recommendation comprising an adjusted dose of the medicament; d. Measuring or determining a second value of the reference parameter associated with the person; and e. Providing a second therapeutic recommendation based on the first therapeutic recommendation and the second value of the reference parameter, the second therapeutic recommendation comprising an update to the adjusted dose of the medicament.

2. The method of claim 1, wherein the reference parameter is a reference glucose value.

3. The method of claim 2, wherein the reference glucose value is a value of blood glucose or interstitial glucose.

4. The method of claim 3, wherein the reference glucose value is interstitial glucose, and further comprising deriving the reference interstitial glucose value from a time series of continuous glucose monitoring (CGM) values.

5. The method of claim 4, wherein the reference glucose value is associated with a 12thpercentile of the time series of CGM values.

6. The method of claim 4, wherein the second therapeutic recommendation is further based on at least a portion of the time series of CGM values.

7. The method of any one of claims 1-6, wherein the first and second therapeutic recommendations include insulin amounts to be delivered.

8. The method of claim 7, wherein the insulin amounts to be delivered are transmitted to a medicament delivery device.

9. The method of claim 8, wherein the medicament delivery device is a pump or smart pen.

10. A method of detenuining a medicament dose, comprising:Dexcom Docket No. 0991-PCT01 receiving Continuous Glucose Monitoring (CGM) data for a subject over a period of time; defining a reference blood glucose (RefBG) metric based at least in part on the CGM data, wherein the RefBG metric is a statistic derived from a specified time window of the CGM data; modeling one or more additional metrics of a therapeutic performance of the subject using the RefBG metric to produce one or more models of therapeutic performance; establishing (i) a medicament dose and (ii) a RefBG target value using the one or more models of therapeutic performance; evaluating whether one or more values of the RefBG metric are above the RefBG target value, and, when the one or more values of the RefBG metric are above the RefBG target, avoiding use of the corresponding medicament dose in future updates of the medicament dose; and adjusting the medicament dose to provide a subsequent medicament dose based, at least in part, on the evaluation of the one or more values of the RefBG metric.

11. The method of claim 10, wherein the RefBG metric comprises the Xth percentile of an individual’s or population’s CGM data from the specified time window, and wherein X is selected as a percentile that minimizes inter-day variance or that minimizes a cost function that penalizes both inter-day variance and proximity to true fasting blood glucose samples with weights of the cost function chosen to reflect user preferences.

12. The method of claim 10, wherein the medicament comprises insulin and wherein the subsequent medicament dose comprises an adjusted insulin dose.

13. The method of claim 12, wherein an indication of the adjusted insulin dose is transmitted to a medicament delivery device.

14. The method of claim 13, wherein the medicament delivery device comprises a pump or smart pen.

15. The method of any one of claims 10-14, wherein the one or more additional metrics includes Ale, Time in Range, or True Fasting Blood Glucose.

16. The method of any one of claims 10-14, wherein the initial medicament dose is determined by a health care practitioner or is determined from individual biomarkers.

17. A system for determining a medicament dose, the system comprising: a processor; and computer-readable storage media storing instructions executable by the processor to:Dexcom Docket No. 0991-PCT01 receive Continuous Glucose Monitoring (CGM) data for a subject over a period of time; define a reference blood glucose (RefBG) metric based at least in part on the CGM data, wherein the RefBG metric is a statistic derived from a specified time window of the CGM data; model one or more additional metrics of a therapeutic performance of the subject using the RefBG metric to produce one or more models of therapeutic performance; establish a RefBG target value using the one or more models of therapeutic performance; for a plurality of historical RefBG values, evaluate whether each respective historical RefBG value is above the RefBG target value, and, when the respective historical RefBG value is above the RefBG target, mark a dose value or values that produced the respective historical RefBG value as “to-be-avoided” in future updates of the medicament dose; and adjust a current medicament dose to provide fixed-size or proportional dose adjustment away from the dose value or values marked as “to-be-avoided.”18. The system of claim 17, wherein an indication of the adjusted current medicament dose is transmitted to a medicament delivery device.

19. The system of claim 18, wherein the medicament delivery device comprises a pump or smart pen.

20. The system of any one of claims 17-20, wherein the one or more additional metrics includes Ale, Time in Range, or True Fasting Blood Glucose.

Images