Method and system for predicting changes in a patient's physiological parameters

Abstract

This disclosure relates to a method and system for predicting changes in target physiological parameters of a target patient that are expected to result from a drug administered to the target patient. The method and system may include deriving one or more parameter estimation functions based on historical data, each parameter estimation function modeling how distinct parameters of a prediction function vary according to one or more initial physiological parameters. The method and system may further include receiving user input indicating values ​​for one or more initial physiological parameters for a target patient, calculating values ​​for each parameter of the prediction function by applying the derived parameter estimation function to the indicated one or more values, and predicting changes in the target physiological parameters using the prediction function and the calculated parameter values. The predicted changes may then be displayed on a user interface.

Description

[Technical Field]

[0001] This disclosure relates to a method and system for predicting changes in one or more physiological parameters of a patient. More specifically, this disclosure relates to a method and system for predicting changes in one or more target physiological parameters of a patient that are expected to result from a course of medication administered to the patient. [Background technology]

[0002] Patients are generally administered medications to induce changes in one or more of their target physiological parameters. This is especially true for patients with chronic diseases requiring long-term treatment. As a result, when a patient is administered a medication or course of medication, the patient and their caregivers (e.g., healthcare providers or HCPs) generally expect to see some measurable change in one or more of the patient's target physiological parameters. Changes in a patient's physiological parameters can be considered the outcome or result of the administered medication or course of medication. However, the outcomes resulting from a medication or course of medication can vary from patient to patient. [Overview of the project]

[0003] In one aspect, the present disclosure is directed to a method for predicting changes in a target physiological parameter of a target patient that are expected to result from a drug administered to the target patient. The method includes, in one or more computing devices, accessing historical data indicative of changes in the target physiological parameters of a plurality of patients observed over an observation period resulting from administration of the drug; in one or more computing devices, deriving, based on the historical data, one or more parameter estimation functions, each parameter estimation function modeling how distinct parameters of a prediction function for predicting changes in the target physiological parameter over time vary according to one or more starting physiological parameters observed in a plurality of patients; in one or more computing devices, receiving user input indicative of values for each of the one or more starting physiological parameters for the target patient; in one or more computing devices, calculating values for each parameter of the prediction function by applying the one or more derived parameter estimation functions to the one or more values indicated by the received user input; in one or more computing devices, predicting changes in the target physiological parameter of the target patient at a plurality of future time points using the prediction function and the calculated parameter values; and displaying the predicted changes in the target physiological parameter of the target patient on a user interface of at least one of the computing devices to assist at least one of the target patient and a medical professional in determining whether the drug should be administered to the target patient.

[0004] In some embodiments, the drug is an anti-obesity drug and the target physiological parameter is the patient's weight.

[0005] In some embodiments, the drug is an anti-diabetic agent and the target physiological parameter is the patient's A1C level.

[0006] In some embodiments, the one or more starting physiological parameters include at least one of age, A1C level, height, weight, resting heart rate, and biological sex.

[0007] In some embodiments, the prediction function is a probability distribution function.

[0008] In some embodiments, predicting a change in a target physiological parameter of a target patient includes predicting, for each of a plurality of future time points, the expected change in the target physiological parameter and the prediction interval for the expected change.

[0009] In some embodiments, at least one of the parameter estimation functions models how a parameter indicating a measure of statistical dispersion varies according to one or more starting physiological parameters.

[0010] In some embodiments, the plurality of patients in the historical data have been administered different dosage levels of a drug such that the historical data shows how the change in the target physiological parameter varies with the dosage level. In such embodiments, at least one of the derived parameter estimation functions may model how one of the parameters of the prediction function varies according to the dosage level, the user input received at one or more computing devices may further indicate the target dosage level for the drug to be administered to the target patient, the value for at least one parameter of the prediction function may be calculated by one or more computing devices based at least in part on the target dosage level, and the predicted change in the target physiological parameter of the target patient displayed on the user interface to assist at least one of the target patient and the medical professional in determining whether the target dosage level of the drug should be administered to the target patient may be calculated based at least in part on the target dosage level.

[0011] In some embodiments, the method may further include receiving, on one or more computing devices, further user input indicating actual changes to a target physiological parameter observed in a target patient in response to a course of drugs previously administered to the target patient, and the calculated parameter values ​​of the predictive function are calculated based at least in part on the further user input.

[0012] In some embodiments, before user input is received, the parameter estimation function is derived by one or more computing devices and stored in memory accessible by one or more computing devices.

[0013] In some embodiments, the access and derivation steps are implemented in a first computing device, while the receiving, calculating, predicting, and displaying steps are implemented in a second computing device.

[0014] In some embodiments, one or more parameter estimation functions are derived from historical data using linear regression.

[0015] In another aspect, the disclosure relates to a system for predicting changes in target physiological parameters of a target patient that are expected to result from a drug administered to the target patient, the system comprising a user interface, one or more memory systems for storing computer executable instructions, and one or more processors, which are communicatively coupled to the one or more memory systems and the user interface, and which access historical data showing changes observed over an observation period in target physiological parameters of multiple patients resulting from drug administration, and one or more parameter estimation functions, each parameter estimation function modulating how distinct parameters of a prediction function for predicting changes in target physiological parameters over time vary according to one or more initial physiological parameters observed in multiple patients. The system comprises one or more processors configured to execute instructions for: deriving one or more parameter estimation functions based on historical data; receiving user input indicating values ​​for each of one or more initial physiological parameters for a target patient; calculating values ​​for each parameter of a prediction function by applying one or more derived parameter estimation functions to one or more values ​​indicated by the received user input; using the prediction function and the calculated parameter values ​​to predict changes in the target physiological parameters of the target patient at multiple future points in time; and displaying the predicted changes in the target physiological parameters of the target patient on a user interface to help at least one of the target patient and a healthcare professional determine whether a drug should be administered to the target patient.

[0016] In yet another aspect, the disclosure relates to a non-temporary computer-readable medium storing computer-executable instructions for predicting changes in target physiological parameters of target patients that are expected to result from a drug administered to the target patient, wherein, when executed by one or more processors, the computer-executable instructions cause the processors to access historical data showing changes observed by one or more processors over an observation period of target physiological parameters of multiple patients resulting from drug administration, and one or more processors, based on the historical data, one or more parameter estimation functions, each parameter estimation function modeling how distinct parameters of a prediction function for predicting changes in target physiological parameters over time vary according to one or more initial physiological parameters observed in multiple patients The system is configured to derive a parameter estimation function, receive user inputs indicating values ​​for each of one or more initial physiological parameters for a target patient in one or more processors, calculate values ​​for each parameter of a prediction function by applying one or more derived parameter estimation functions to one or more values ​​indicated by the received user inputs, have one or more processors use the prediction function and the calculated parameter values ​​to predict changes in the target patient's target physiological parameters at multiple future points in time, and display the predicted changes in the target patient's target physiological parameters on a user interface communicatively coupled to at least one of the one or more processors to help at least one of the target patient and healthcare professionals determine whether a drug should be administered to the target patient. [Brief explanation of the drawing]

[0017] Additional embodiments of the present disclosure, as well as their features and advantages, will become more apparent by referring to the description herein in conjunction with the accompanying drawings. Components in the drawings are not necessarily to scale. Furthermore, in the drawings, similar reference figures indicate corresponding parts throughout different drawings.

[0018] [Figure 1] This is a conceptual block diagram illustrating an exemplary computing device for implementing the method of this disclosure. [Figure 2] This is a conceptual block diagram illustrating an alternative exemplary computing system for implementing the method of this disclosure. [Figure 3A] This flowchart illustrates exemplary logic for predicting changes in a patient's target physiological parameters that are expected to result from a course of medication administered to or to be administered to the patient. [Figure 3B] This flowchart illustrates exemplary logic for predicting changes in a patient's target physiological parameters that are expected to result from a course of medication administered to or to be administered to the patient. [Figure 4] This is an example screen of a web interface on a mobile device for receiving user input. [Figure 5A] This is an example screen of a web interface for presenting the predicted changes in a patient's weight that are expected to result from a course of medication. [Figure 5B] This is an example screen of a web interface for presenting the predicted changes in a patient's weight that are expected to result from a course of medication. [Figure 6] This is another exemplary screen of a web interface for presenting the predicted changes in a patient's A1C level that are expected to result from a course of medication. [Figure 7] Another exemplary screen of a web interface, configured for display on laptops, desktops, and / or tablets, and for presenting the predicted changes in a patient's weight expected to result from the medication. [Figure 8] This is another exemplary screen of a web interface, configured for display on a laptop or desktop, and for presenting the predicted changes in a patient's A1C levels that are expected to result from the medication. [Modes for carrying out the invention]

[0019] To facilitate understanding of the principles of this disclosure, the embodiments illustrated in the drawings will be described below using specific terminology. Nevertheless, it will be understood that this does not intend to limit the scope of the invention.

[0020] This disclosure relates to a method and system for predicting changes in a patient's target physiological parameters that are expected to result from a course of medication administered to or to be administered to the patient. When a patient takes medication, the patient generally expects to see some measurable change in one or more of their physiological parameters. This is especially true for patients taking medication for relatively long periods to treat chronic diseases. For example, a patient taking blood pressure medication to lower blood pressure often expects to see a reduction in blood pressure. A patient taking anti-obesity medication often expects to see a reduction in weight. Similarly, a patient taking medication to treat diabetes often expects to see an improvement in their hemoglobin A1C (HbA1C or A1C) score over time. Thus, changes in one or more target physiological parameters that may be observed in a particular patient as a result of taking medication can be considered a drug outcome.

[0021] However, outcomes resulting from medication can vary significantly from patient to patient. Some patients may experience dramatic changes or improvements in physiological responses, while others may experience only minor changes. The time required for patients to observe changes in target physiological parameters can also vary; some patients may experience dramatic changes relatively quickly, while others may only see the same changes after taking the medication for a longer period. To decide whether to start or continue taking medication, individual patients and their caregivers (e.g., healthcare providers or HCPs) often desire to predict, as accurately as possible, the outcomes that this individual patient might expect to see from this medication. In other words, individual patients and their caregivers may desire to personalize outcome predictions to take into account the specific individual patient's initial physiological parameters in order to improve the accuracy of such predictions.

[0022] Such outcomes can be an important factor in helping patients and their caregivers decide whether or not to take medication. When such outcomes are predicted for multiple types of medication, these predicted outcomes can help patients and their caregivers compare and contrast the possible effects of different medications, ultimately helping them decide which medication to take. Such predicted outcomes can also be useful in helping patients and their caregivers form realistic expectations about the extent to which a medication may or may not improve the patient's condition. Furthermore, accurate predicted outcomes can help patients and their caregivers determine if there are any particular unexpected medical complications or causes that need to be considered. For example, if a patient's actual outcome after taking medication for a period of time does not match the patient's initially predicted outcome, this may prompt patients and their caregivers to further investigate the underlying cause of the patient's condition or illness and adjust the patient's treatment plan as necessary.

[0023] The inventors have recognized that when certain initial physiological parameters of a patient are taken into consideration, the patient can often predict with greater accuracy the outcomes they may expect from a course of medication. As used herein, the patient's initial physiological parameters include physiological parameters measured before the patient begins taking a course of the target medication, and predict the effects of the course of the target medication. This is because the outcomes a patient may expect from a course of medication in the future are often predictably varied by the patient's initial physiological state. The inventors have further recognized that the way in which outcomes are predictably varied by the patient's initial physiological state can be determined, or at least estimated, from an analysis of historical data showing outcomes observed in multiple patients who have received the medication. Preferably, the multiple patients in the historical data exhibit a wide variety of physiological states and present different initial physiological parameters before being administered the medication. By applying standard regression techniques to this historical data, it is possible to estimate the way in which outcomes may be expected to vary by the patient's different initial physiological state.

[0024] For example, a particular medication may be prescribed to help a patient lose weight. The amount of weight a particular patient is expected to lose as a result of a course of the targeted medication may vary based on the patient's initial physiological parameters, such as height, weight, resting heart rate, and A1C level measured before starting the course of the targeted medication. Furthermore, how the patient's expected weight loss varies depending on the aforementioned initial physiological parameters can be estimated by analyzing historical data showing outcomes in a population of users with different initial height, weight, resting heart rate, and A1C levels. By taking these current physiological parameters into account and analyzing historical data, the amount of weight a particular patient is expected to lose as a result of taking this medication can be estimated with greater accuracy.

[0025] As used herein, the terms “logic,” “control logic,” “application,” or “computer executable instruction” may include software and / or firmware running on one or more programmable processors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), hardwired logic, or combinations thereof. Accordingly, various logics may be implemented in any suitable manner according to the embodiments, and still be subject to the embodiments disclosed herein.

[0026] Figure 1 is a conceptual block diagram illustrating an exemplary computing device 101 for implementing the method of this disclosure. The computing device 101 exemplifies a mobile device such as a smartphone. Alternatively, any suitable computing device may be used, including, but not limited to, a laptop, desktop, tablet, or server computer.

[0027] The computing device 101 includes at least one processor 102 that executes software and / or firmware stored in the memory 104 of the device 101. The software / firmware code includes instructions that, when executed by the processor 102, cause the computing device 101 to perform the functions described herein. At least one processor 102 may be configured to execute control logic 106, which may be stored in the memory 104 and may operate to predict changes in one or more target physiological parameters of a patient, as discussed in detail herein. The memory 104 may be any suitable computer-readable medium accessible by the processor 102. The memory 104 may be a single storage device or multiple storage devices, may be located inside or outside the processor 102, and may include both volatile and non-volatile media. The exemplary memory 104 includes random-access memory (RAM), read-only memory (ROM), erasable programmable ROM (EEPROM), flash memory, magnetic storage devices, optical disk storage, or any other suitable medium configured to store data and accessible by the processor 102.

[0028] The computing device 101 includes a user interface 108 which communicates with a processor 102 and is operable to provide user input data to the system and to receive and display data, information, and prompts generated by the system. The user interface 108 includes at least one input device for receiving user input and providing user input to the system. In the illustrated embodiment, the user interface 108 is a graphical user interface (GUI) which includes a touchscreen display operable to display data and receive user input. The touchscreen display enables the user to interact with presented information, menus, buttons, and other data to receive information from the system and provide user input to the system. Alternatively, a keyboard, keypad, microphone, mouse pointer, or other suitable user input device may be provided.

[0029] Figure 2 is a conceptual block diagram illustrating an alternative exemplary computing system 200 for implementing the method of the present disclosure. The system 200 may comprise a computing device 201. Similar to computing device 101, device 201 may include memory 204 (similar to memory 104), a processor 202 (similar to processor 102), and a user interface 208 (similar to user interface 108). In addition, computing device 201 may comprise a communication interface 212 configured to establish a communication link 230 with a corresponding communication interface 214 of another computing device 250. Computing device 250 may, but does not have to be, be located remotely from computing device 101, for example, in another room, another building, another city, another country, or another continent. For example, computing device 250 may be, for example, one or more cloud servers configured to provide or offer a remote server or virtual computing device. The communication link 230 may be a direct communication link established directly between computing device 101 and computing device 250 via a wired or wireless connection (e.g., Bluetooth, NFC, or WiFi connection). Alternatively or in addition, all or part of the communication link 230 may include one or more intermediate devices and / or traverse a data network (not shown) such as a Local Area Network (LAN), a cellular network, or the Internet. Similar to computing devices 201 and 101, computing device 250 may comprise a processor 252 and memory 254 (or, in the case of a cloud computing service, a virtual processor and virtual memory). In the embodiment depicted in Figure 2, control logic 206 (similar to control logic 106) that operates to predict changes in one or more target physiological parameters of a subject may be stored in memory 254 and configured to be executed by processor 252.In the embodiment depicted in Figure 2, user input may be received via the user interface 208 of computing device 201 and transmitted to computing device 250 via the communication link 230. Computing device 250 may then implement the logic considered herein to predict one or more target physiological parameters of a patient. Computing device 250 may store or further process the output (e.g., target physiological parameters), return the output to computing device 201, or transfer the output to another device (not shown). If the output is returned to computing device 201, device 201 may display the output, or information derived from the output, to the user via the user interface 208.

[0030] In one exemplary embodiment, the communication link 230 between computing device 201 and computing device 250 may be a web link established on the World Wide Web. Computing device 201 may be configured to present a web page or web interface to its user via a browser or similar interface. User input from the user received via user interface 208 may pass through the web page or web interface and be transmitted to computing device 250 via the web link. In this embodiment, computing device 250 may be a remote server (or one or more cloud servers implementing a virtual server) configured to receive user input, predict one or more physiological parameters of a patient, and provide and return the predicted parameters to computing device 201 via the web link 250. Computing device 201 may then present the predicted parameters, or information derived from the predicted parameters, to its user via the web page or web interface.

[0031] Figures 3A and 3B present flowcharts illustrating exemplary logic 300 for predicting changes in a patient's target physiological parameters that are expected to result from a course of medication administered to or to be administered to the patient. As discussed herein, logic 300 may be executed on processors 102, 202, and / or 252. Logic 300 may be executed on a single processor among these processors operating independently, or in coordination among two or more processors. When logic 300 is implemented in coordination among two or more processors, the processors may cooperate so that different processors among the two or more processors perform different steps, or so that multiple processors perform one or more steps.

[0032] In some embodiments, the target physiological parameter may be the patient's body weight. Alternatively or in addition, the target physiological parameter may be the patient's hemoglobin A1C (HbA1C, or A1C) level. Alternatively or in addition, the target physiological parameter may be any measurable physiological parameter of the patient, such as blood pressure, heart rate, muscle-to-fat ratio, Body Mass Index (BMI), levels of one or more biomarkers in the patient's blood, tissue, or other body medium, or the severity level of one or more symptoms observed in the patient (but not limited to these). The drug may be any pharmaceutical product that is expected to produce a change in one or more target physiological parameters. For example, the drug may be one of a class of drugs that is expected to produce a change in the patient's A1C level or the patient's body weight when administered to the patient once or repeatedly over a period of time. Such drugs may include SGLT2, GLP-1 agonists, GIP / GLP-1 agonists, GIP / GLP-1 / glucagon agonists, sulfonylureas, insulin, insulin analogs, and / or other drugs approved by regulatory authorities for use in the treatment of diabetes and / or chronic weight management. Examples of such drugs include, but are not limited to, tilzepatide, semaglutide, letatoltide, metformin, insulin glargine, and insulin lysopro. In one embodiment, such a drug is tilzepatide.

[0033] Logical 300 may begin in step 302, in which one or more computing devices access historical data showing changes observed over an observation period in multiple patients' target physiological parameters resulting from drug administration. The observation period should preferably be long enough to allow observation of changes in target physiological parameters resulting from the drug, such as several weeks, several months, or several years. The historical data may include data derived from clinical trials in which the effects of a drug on target physiological parameters have been studied and observed in a large population of patients. The historical data may also include other types of data that quantitatively measure or show the effects of a drug on target physiological parameters in a population of subjects who have previously been administered the drug. Such alternative types of data may include, or may be derived from, real-world data and / or observational studies, electronic health records from one or more health systems, insurance claim data from one or more healthcare payers, etc.

[0034] As described herein, the patients studied in the historical data should preferably exhibit a wide variety of initial physiological states, so that the historical data may be used to estimate how outcomes vary depending on different initial physiological states. The initial physiological states included in the historical data may include any physiological parameters that can be measured in the patient. Illustrative examples of suitable initial physiological parameters include the patient's height, weight, resting heart rate or pulse rate, and initial A1C level. Further examples of suitable initial physiological parameters include the patient's biological sex, age, race or ethnicity, blood pressure, body mass index (BMI), the presence or level of one or more genetic markers in the patient, and / or the presence or level of one or more biomarkers present in the patient's blood, tissue, or some other bodily medium. Suitable physiological parameters may also include elements of the patient's medical history, such as indicators of one or more diseases or illnesses present in the patient, indicators of one or more medications the patient is currently taking or has taken in the past, indicators of one or more treatments or therapies the patient is currently receiving or has received, and / or indicators of the patient's family history (e.g., diseases or illnesses present in the patient's immediate relatives or extended family).

[0035] In some embodiments, the historical data may also include outcomes recorded from patients administered at different dose levels of the drug, so that the historical data can be used to estimate how outcomes vary with different dose levels. For example, the historical data may include outcomes recorded from patients administered at one of a given set of dose levels (e.g., three, four, or five given dose levels). In some embodiments, the historical data may include outcomes recorded from patients administered at significantly varying dose levels.

[0036] In step 304, one or more computing devices derive one or more parameter estimation functions based on historical data. Each parameter estimation function may be associated with a distinct parameter of a predictive function and may be used to model a distinct parameter of the predictive function.

[0037] A predictive function can be a computer-implemented model or engine, a set of computer-implemented logic, or a sequence of computer-implemented steps for predicting the expected changes over time in a target patient's target physiological parameters that are expected to occur from the administration of a target drug (or from the administration of a course of a target drug). For example, a predictive function may take time t as input and output the expected changes in the target patient's target physiological parameters. A predictive function may include one or more configurable parameters that can vary. Such configurable parameters may be adjusted to take into account the target patient's initial physiological parameters. Each of the aforementioned parameter estimation functions may be used to model how each parameter of the predictive function varies according to one or more initial physiological parameters of the target patient. Parameter estimation functions may be derived by applying one of several standard regression techniques to historical data. For example, simple linear regression or multiple linear regression may be used to predict the assumed linear relationship between independent variables (e.g., the parameters of the predictive function being modeled) and one or more dependent variables (e.g., one or more initial physiological states, dose levels, and / or predicted time).

[0038] An example of a prediction function is provided in Equation 1 below:

[0039]

number

[0040] This exemplary prediction function is an exponential decay function that accepts the input time t and outputs the expected change in the target physiological parameter from the baseline. Specifically, this prediction function is composed of two exponential decay functions, namely, 1 - exp{tk i} and 1 - exp{dk i}. In this exemplary prediction function, there are three configurable parameters, namely, the first parameter θ ij associated with the first parameter estimation function, the second parameter k i associated with the second parameter estimation function, and the third parameter ε ijThere are several parameters. Each parameter is explained in more detail below.

[0041] θ ij θ can be estimated based on the first parameter estimation function. In the example given in Equation 1, θ ij This varies based on the subject j and the dose level i. Therefore, θ ij The parameter estimation function can accept, as input values, any or all of the set of initial physiological parameters of the target patient, as well as the target dose level to be administered to the target patient. The first parameter estimation function is then θ ij These inputs can be processed according to mathematical formulas or equations to output a numerical estimate for θ. This first parameter estimation function can be derived based on the historical data mentioned above. For example, the first parameter estimation function is θ by (i) the initial physiological parameters observed in multiple patients studied in the historical data and (ii) the dose levels administered to multiple patients. ij To estimate how θ fluctuates, it can be determined by applying any standard regression technique (e.g., linear regression) to the aforementioned historical data. ij This varies based on the subject j and the dose level i, but in other embodiments, θ ij It should be understood that it can be estimated based on different parameters. For example, θ ij This can be additionally estimated based on the duration d of the observation period in the historical data. θ ij This can be estimated based on fewer input parameters, more input parameters, or different input parameters. In some embodiments, θ ij This can be set to a single invariant value that applies to all subjects and all dose levels, for example, θ ij It does not change at all based on any input parameter.

[0042] Similarly, k i k can be estimated based on a second parameter estimation function. In the example given in Equation 1, k i k varies based on the dose level i. Therefore,i The parameter estimation function for k can accept the target dose level administered to the target patient as input. Then the parameter estimation function is k i To output a numerical estimate for k, the dose level can be processed according to a mathematical formula or equation. Here again, k i The parameter estimation function for k is: i To estimate how it varies with dose levels, this can be determined by applying any standard regression technique to the aforementioned historical data. In this exemplary embodiment, k i In other embodiments, k varies based solely on the dose level i, but in other embodiments, k i is, θ ij It should be understood that k can be estimated based on different parameters, including any or all of the parameters used to estimate k. i If it is desired to estimate k i The estimation function for k is, again, determined by the aforementioned additional / different parameters in the historical data. i How it has changed can be determined by applying the standard regression techniques described above.

[0043] ε ij is, ε ij Since θ is a random variable and not an estimated value, ij and k i This is different. ε ij This can be modeled by a probability distribution function such as a normal Gaussian distribution (but is not limited to this). In this embodiment, ε ij It is assumed that ε has a predicted value of zero. ij The probability distribution for θ may include measures of variability such as the standard deviation σ. In the example given in Equation 1, the standard deviation σ varies based on the subject j and the dose level i. Therefore, θ ij and k iSimilarly, the standard deviation σ can be estimated using a third parameter estimation function that accepts any or all of the target patient's initial physiological parameters, as well as the target dose level administered to the target patient, as input values. This third parameter estimation function can again be derived by applying any standard regression technique to the historical data described above to estimate how σ varies with the initial physiological parameters and dose level. In this example, σ varies based on subject j and dose level i, but it should be understood that in other embodiments, σ ​​can be estimated based on fewer input parameters, more input parameters, or different input parameters. For example, σ can also be estimated based on the predictive time t, either additionally or alternatively. In some embodiments, σ ​​may be set to a single invariant value that applies regardless of the patient's current physiological parameters, dose level, and / or predictive time t; for example, σ does not vary at all based on any input parameters. This consideration is ε ij This focuses on the standard deviation σ of the probability distribution function for modeling, but also on the statistical variance (e.g., σ). 2 ), the interquartile range, and / or a random variable ε, including but not limited to these ranges. ij It should be understood that any measure of statistical variability can be used.

[0044] In step 306, one or more computing devices receive user input indicating values ​​for each of one or more initial physiological parameters for the target patient. For example, if the set of initial physiological parameters includes the target patient's initial weight, A1C level, resting heart rate, and height, one or more processors may receive user input indicating values ​​for each of these initial physiological parameters (e.g., initial weight = 279 lbs, initial A1C level = 8.2, initial heart rate = 88 beats per minute, initial height = 5 feet 7 inches). Optionally, in step 306, one or more computing devices may also receive user input indicating a target dose level expected to be administered to the target patient. Additionally or alternatively, the processor may also receive user input indicating the duration of a forecast period over which the user wishes to predict changes in the patient's target physiological parameters.

[0045] In step 308, one or more computing devices calculate values ​​for each parameter of the prediction function by applying one or more derived parameter estimation functions to one or more values ​​indicated by the received user input. For example, parameter θ ij This can be estimated by applying the first parameter estimation function derived in step 304 to some or all of the initial physiological parameters and / or dose levels received in step 306. Similarly, parameter k i The parameter σ can be estimated by applying the second parameter estimation function derived in step 304 to the dose level received in step 306. The parameter σ can be estimated by applying the third parameter estimation function derived in step 304 to some or all of the initial physiological parameters and / or dose levels received in step 306.

[0046] parameter θ ij , k iWhen , and σ have all been estimated using their respective parameter estimation functions, the prediction function provided in Equation 1 is then given by the predicted value Y of the change in the patient's target physiological parameter at multiple time points t during the prediction period. ij This can be used (in step 310) to estimate. Multiple future time points may extend to a future foresight period, which is long enough that changes in the target physiological parameters are expected to result from the administration of the drug or a course of the drug. For example, the foresight period may extend to one month, three months, six months, nine months, and / or twelve months in the future. Multiple future time points may include two, three, or any number of time points within the foresight period.

[0047] In the example prediction function given in Equation 1, Y ij ε is a random variable ij Since it is determined based on a partial basis, it is a random variable. Therefore, Equation 1 represents not only the expected change in the target physiological parameter, but also Y. ij It can be used to determine the probabilities associated with values ​​higher than expected and values ​​lower than expected. For example, given a specific time t within the forecast period, equation 1 is: Y ij The 90% prediction interval for Y, for example, the period during which a patient's target physiological parameter is expected to fall within a 90% probability. ij It can be used to determine the range of values ​​for Y at time t. Such a prediction interval is Y at time t. ij It can be used to derive and present a range of expected values ​​for a given value. Naturally, different prediction intervals (e.g., 85% intervals or 95% intervals) can be used.

[0048] In step 312, the processor may display the predicted change in the patient's target physiological parameter, which is determined by equation 1, such as Y ij This may include displaying the expected value for Y. Alternatively or in addition, the processor may include Y ijThe processor may display probabilities associated with values ​​higher than expected and / or lower than expected. In some embodiments, the processor may also display predicted intervals for the expected change. The processor may display the predicted change on a user display communicatively coupled to the processor. Such a display may be integrated into the same device embodying the processor, or be physically coupled (e.g., a processor 102 on computing device 101 may display the results on a user interface 108), or such a display may be remote from the processor implementation logic 300 (e.g., a processor 252 on computing device 250 may display the predicted change Y ij This information may be communicated to the computing device 201 via the communication link 230, thereby enabling the computing device 201 to display the predicted changes on the user interface 208. The displayed predicted changes can be used to help at least one of the target patient and a healthcare professional (e.g., a healthcare professional directly or indirectly responsible for caring for or advising the target patient) determine whether the drug should be administered to the target patient.

[0049] Figures 3A and 3B, as well as the preceding discussion, present one exemplary embodiment of logic 300, but it should be understood that any of the steps in logic 300 can be adjusted, extended, rearranged, deleted, and / or modified in various ways. For example, θ ij and k i As presented in the previous discussion, θ is a mathematical (numerical) parameter that can be estimated using each parameter estimation function, ij and k i Also ε ij These can be similar random variables, each of which can be modeled by an associated probability distribution function. Also, ε ij Similarly, θ ij and / or k iThe probability distribution function of may have parameters that can be estimated using estimation functions derived from applying standard regression techniques to the aforementioned historical data. For example, θ ij and / or k i The type of distribution to model the distribution (e.g., normal (Gaussian), binomial, Poisson, etc.), the expected value of the distribution, and / or statistical measures of the distribution's variability (e.g., variance, standard deviation, interquartile range, or range) can all be estimated based on the patient's current physiological parameters, dose levels, and / or some or all of the predictive period by applying regression techniques to historical data.

[0050] In some embodiments, a user may use logic 300 to predict changes in a patient's target physiological parameters after the patient has already taken a drug over a previous time period. In such embodiments, the patient may have already observed certain actual changes in the target physiological parameters as a result of taking the drug over a previous observation period. In such embodiments, logic 300 may be extended and / or modified to take into account actual changes observed by the patient in order to improve the accuracy of the predicted changes in the target physiological parameters that are expected to result from continued administration of the drug. For example, a patient may have taken a drug for weight loss over a previous 3-month observation period. During these 3 months, the patient may have observed certain actual changes in his / her weight (e.g., a weight loss of 10 lbs) as a result of taking the drug. Logic 300 may be extended and / or modified to take into account the actual observed changes in the patient's weight over the previous 3-month observation period while predicting further expected changes in the patient's weight if the patient continues to take the drug. In this way, logic 300 may improve the accuracy of its predictions by customizing its logic for the specific patient under consideration.

[0051] Logic 300 may be extended and / or modified in this way by adding an optional step (not shown) between step 302 and step 304, in which case the processor receives several observed changes in the patient's target physiological parameters during the previous observation period. d' indicates the duration of the previous observation period, L indicates the number of observed changes received, and Z indicates the number of observed changes received. t’ Let represent the actual observed change in the target physiological parameter at time t', where t' is the time point within the previous observation period. Then, the parameter specific to patient j is calculated.

[0052]

number

[0053]

number

[0054] Next, parameters

[0055]

number

[0056]

number

[0057] Equation 3 is the parameter

[0058]

number

[0059]

number

[0060] Logic 300 represents a technical improvement over prior known systems in several respects. Firstly, because Logic 300 uses a prediction function (e.g., an exponential decay function provided in Equation 1 or Equation 3) that takes time t as input and outputs the expected change in the target physiological parameter at the input time t, the prediction function can be used to estimate the expected change in the target physiological parameter at multiple time points t. Thus, unlike prior known systems, Logic 300 is not constrained to providing estimated changes at only one or two predetermined future time points (e.g., estimating only changes over the next three or six months), but can estimate predicted changes at any time t specified by the user.

[0061] Secondly, logic 300 is constructed such that certain computational or memory-intensive steps can potentially be performed in advance by a first computing device with access to larger computational and / or memory resources. These first resource-intensive steps yield an output (e.g., a parameter estimation function derived by step 304), which can be stored in memory and / or communicated to a computing device with access to relatively fewer computational and / or memory resources. When a user provides initial physiological parameters for a specific target patient and requests a personalized prediction for the target patient, the stored output can then be used to calculate the expected changes in the target physiological parameters of the target patient with relatively little time and / or consumption of computational or memory resources.

[0062] Specifically, the historical data accessed in steps 302 and 304 (Figure 3A) is large in size and may require larger computational and / or memory resources to handle. Similarly, the derivation of the parameter estimation function in step 304 can be a resource-intensive task because it requires performing one or more regressions on a large dataset. If steps 302 and 304 are repeated each time the user wishes to calculate the expected changes in the target physiological parameters of a target patient, it would be difficult to implement logic 300 on a computing device with access to only limited computational and / or memory resources. Furthermore, it may be difficult to implement logic 300 quickly, meaning that logic 300 may require a long runtime.

[0063] However, logic 300 is constructed such that steps 302 and 304 can be implemented in advance (for example, before the user requests an estimate for a specific target patient) on a first computing device (e.g., processor 252 in computing device 250) with relatively large computing and memory resources. Once implemented, the output of step 304 (e.g., the derived parameter estimation function) can be stored in memory and / or communicated to other devices such as computing device 201, which has relatively fewer computing and / or memory resources. This stored output incorporating the derived parameter estimation function can be embedded in a relatively compact code module or data structure that does not require much memory (or much bandwidth for transfer) to store, and can then be utilized at runtime, for example, when the user actually requests a prediction for a specific target patient. This improves responsiveness from the user's perspective because less time is required to generate a prediction once requested. Furthermore, since steps 306 to 312 require relatively little computing resources, they can be implemented on a computing device (such as computing device 201) that has relatively less computing and / or memory resources.

[0064] Thirdly, logic 300 improves the security of the historical data accessed in step 302. The historical data accessed in step 302 is typically not only large in size but can also be commercially or strategically valuable. For example, clinical trials can require years of effort, as well as significant financial and resource expenditures, making the resulting data extremely valuable. Historical data may also be subject to certain privacy restrictions to protect sensitive information associated with multiple patients being studied, and consequently, leakage of historical data to unauthorized users could lead to liability for the owner and / or administrator of the historical data. For these and other reasons, the owner and / or administrator of the historical data often desires to restrict access to the historical data to mitigate the possibility of the data being leaked or misused by others. To improve the security of historical data, steps 302 and 304 (the only step requiring direct access to historical data) can be pre-implemented on a relatively secure computing device (e.g., a device operating under enhanced cybersecurity measures). Since the parameter estimation function cannot be easily reverse-engineered to derive the source historical data, the output of step 304 (e.g., the derived parameter estimation function) may then be exposed or transferred to other parties or less secure computing devices to implement steps 306-312. To further improve the security of the historical data, the derived parameter estimation function may be embedded in a code module or data structure in an opaque manner that is not easily discoverable through reverse engineering. For example, the derived parameter estimation function may be incorporated into compiled machine code, and only the compiled machine code may be made available to other parties or less secure computing devices, so that other parties or devices have no way of recovering the source code that sets the derived parameter estimation function in a human-readable format.In this way, less secure computing devices or other parties may benefit from historical data without the owner and / or manager of the historical data having to disclose the source data.

[0065] Figure 4 presents an exemplary screen 400 of a web browser or web-enabled application interface (hereinafter referred to as the "web interface") on a mobile device for receiving user input from a user. In this example, the web interface is configured to predict either a change in the patient's weight or A1C level that is expected to result from a course of GIP / GLP-1 agonist administered to the patient. The user may be the patient, or someone entering data on behalf of the patient, such as a caregiver, family member, or healthcare provider (HCP). Screenshot 400 includes separate fields that can be entered by the user. In this example, these fields include field 402 for receiving user input indicating the patient's starting weight, field 404 for receiving user input indicating the patient's starting A1C level, field 406 for receiving user input indicating the patient's starting resting heart rate, and field 408 for receiving user input indicating the patient's starting height. Further fields, different fields, or different units of measurement may be used in the web interface 400. In this example, the user provides input indicating that the patient has a current weight of 279 lbs, a current A1C level of 8.2, a current resting heart rate of 88, and a current height of 5 feet 7 inches. Once the user has finished providing user input, the user may proceed to the next step in logic 300 by touching or activating button 410 ("View Projected Path"). For example, pressing button 410 may cause the user input to be sent by computing device 201 to computing device 250 for processing via web link 230.

[0066] Pressing button 410 may transition the web interface to an exemplary screen 500 shown in Figure 5A. Screenshot 500 includes a toggle button 502 that allows the user to select a target physiological parameter, for example, whether to visualize the patient's A1C level or the predicted change in the patient's weight. In the illustrated example, the user has selected to visualize the predicted change in the patient's weight. Screen 500 further includes a drop-down menu 504 that allows the user to select a dose level between, for example, 5, 15, or 15 mg. In the illustrated example, the drop-down menu 504 is set to a dose level of 15 mg. The user's selection in the drop-down menu 504 indicates the dose level i in the formula discussed earlier. Screen 500 also includes a selection bar 520 that allows the user to select an appropriate forecast period, for example, to forecast one month, three months, six months, nine months, or twelve months into the future. In the illustrated example, the user has selected a forecast period of six months.

[0067] Screen 500 further includes a results panel 506. Panel 506 includes a graph 514. In graph 514, the horizontal axis represents time, and the vertical axis represents the magnitude or level of the target physiological parameter. Graph 514 includes three curves 516a, b, and c, showing predictions of the level of the target physiological parameter at multiple time points over a forecast period. Curve 516a shows the mean predicted level, curve 516b shows the maximum predicted level, and curve 516c shows the minimum predicted level over time. Each of curves 516a, b, and c can be derived from Equation 1 (or Equation 3), as discussed above. In this embodiment, the maximum and minimum predicted levels are derived from the 90% prediction interval from Equation 1 (or Equation 3), as discussed earlier.

[0068] Graph 514 further includes a selected time indicator 518. The time indicator 518 may include any visual features that indicate the selected current time to the user. In some embodiments, the time indicator 518 may appear as an icon, symbol, or text indicating the selected current time. In the specific embodiment depicted in Figure 5A, the selected time indicator 518 appears as a vertical bar superimposed on curves 516a, b, and c. The selected time indicator 518 may be selected and moved left or right by a user interacting with the web interface to select different points in time within the forecast period. In the example shown in Figure 5A, the selected time indicator 518 is currently positioned above the date June 24, 2023. The results panel 506 further displays the forecast level 508 for the target physiological parameter at the selected time, as well as the maximum forecast level 510 and the minimum forecast level 512. In this case, since the user has selected toggle button 502 and chosen "Weight", forecast level 508 will show the patient's forecast weight (in this case: 207 lbs), maximum forecast weight (209 lbs), and minimum forecast weight (205 lbs) at the selected future time on June 24, 2023.

[0069] Figure 5B shows screen 500 after the user changes the forecast period to 9 months in selection bar 520 and moves the selected time indicator 518 to the right to select a different date, August 11, 2023. Because the selected time has changed, the forecast level 508 now shows 202 lbs, while the maximum and minimum forecast levels now show 206 lbs and 198 lbs, respectively.

[0070] Selecting the side effect profile link 524 takes the user to a screen that informs them about potential side effects of the medication they are taking. Selecting the details link 522 takes the user to a screen that provides additional details on how the web interface calculates expected changes to the user's weight and / or A1C levels.

[0071] Figure 6 shows screen 500 after the user changes the toggle button 502 to select an A1C level instead of weight. As a result, the projected level 508 shows a projected A1C level of 5.7% on the selected date, June 24, 2023, with the maximum and minimum projected levels showing 5.9% and 5.5%, respectively.

[0072] Figure 7 shows an exemplary screen 700 for a web interface configured to be displayed on a monitor connected to a laptop or desktop (as opposed to on a mobile device such as a smartphone). Similar to screen 500, screen 700 includes a toggle button 702 (similar to toggle button 502), a drop-down menu 704 (similar to drop-down menu 504) for selecting a dose level, and a selection bar 720 (similar to selection bar 520) for selecting a prediction period. Screen 700 further includes a results panel 706 (similar to results panel 506) which includes a graph 714 (similar to graph 514). Graph 714 further includes a first curve 714a showing the mean predicted level for the target physiological parameter during the prediction period, a second curve 714b showing the maximum predicted level, and a third curve 714c showing the minimum predicted level. A selected time indicator 718 (similar to selected time indicator 518) is superimposed on the three curves. The results panel 706 further includes numerical readouts 708 (similar to readout 508) of the expected level of the target physiological parameter at the selected time, as well as numerical readouts 710 and 712 (similar to readouts 510 and 512) for the maximum and minimum expected levels, respectively. Finally, screen 700 further includes adverse event profile links 724 and 722, similar to links 524 and 522 in Figure 5.

[0073] Figure 7 shows an exemplary screen when the user selects "Weight" using the toggle button 702, and as a result, the results panel 706 displays the predicted level of the patient's weight. Figure 8 shows another exemplary screen from the same web interface when the user selects "A1C" using the toggle button 702, and as a result, the results panel 706 displays the predicted level of the patient's A1C level.

[0074] Various aspects of this disclosure may be used individually, in combination, or in various arrangements not specifically considered in the embodiments described above, and therefore, in their application, are not limited to the details and arrangements of components described or illustrated in the drawings. For example, an aspect described in one embodiment may be combined in any way with an aspect described in another embodiment. Thus, although this teaching has been described in conjunction with various embodiments and examples, it is not intended that this teaching be limited to such embodiments or examples. Rather, the teachings of the present invention encompass various alternatives, modifications, and equivalents, as will be understood by those skilled in the art. Accordingly, the foregoing description and drawings are for illustrative purposes only.

Claims

1. A method for predicting changes in target physiological parameters of a target patient that are expected to result from a drug administered to the target patient, wherein the method is Accessing historical data on one or more computing devices that shows changes observed over an observation period in the target physiological parameters of multiple patients resulting from the administration of the drug, In the one or more computing devices, one or more parameter estimation functions are derived based on the historical data, each parameter estimation function models how a separate parameter of a prediction function for predicting changes in the target physiological parameter over time varies according to one or more initial physiological parameters observed in the multiple patients. The one or more computing devices receive user input indicating values ​​for each of the one or more initial physiological parameters for the target patient, The one or more computing devices calculate values ​​for each parameter of the prediction function by applying the one or more derived parameter estimation functions to the one or more values ​​indicated by the received user input, Using one or more computing devices, predict changes in the target physiological parameters of the target patient at multiple future points in time, using the prediction function and the calculated parameter values. A method comprising displaying the predicted changes in the target physiological parameters of the target patient on the user interface of at least one of the target patient and a healthcare professional in order to help them determine whether the drug should be administered to the target patient.

2. The method according to claim 1, wherein the drug is an anti-obesity drug and the target physiological parameter is body weight.

3. The method according to claim 1, wherein the drug is an antidiabetic drug and the target physiological parameter is at the A1C level.

4. The method according to any one of claims 1 to 3, wherein the one or more initial physiological parameters include at least one of age, A1C level, height, weight, resting heart rate, and biological sex.

5. The method according to any one of claims 1 to 4, wherein the prediction function is a probability distribution function.

6. The method according to claim 5, wherein predicting changes in the target physiological parameters of the target patient includes predicting, for each of the plurality of future time points, the expected changes in the target physiological parameters and the predicted interval for the expected changes.

7. The method according to claim 5 or 6, wherein at least one of the parameter estimation functions models how a parameter representing a measure of statistical variance varies according to one or more initial physiological parameters.

8. The multiple patients in the historical data have been administered different dose levels of the drug such that the historical data shows how the changes in the target physiological parameters vary with dose levels. At least one of the derived parameter estimation functions models how one of the parameters of the prediction function changes according to the dose level. The user input received by the one or more computing devices further indicates a target dose level for the drug to be administered to the target patient. The value for at least one parameter of the prediction function is calculated by one or more computing devices based at least partially on the target dose level. The method according to any one of claims 1 to 7, wherein, in order to assist at least one of the target patient and the healthcare professional in determining whether the target dose level of the drug should be administered to the target patient, the predicted change in the target physiological parameter of the target patient, displayed on the user interface, is calculated at least in part on the target dose level.

9. The method according to any one of claims 1 to 8, further comprising receiving further user inputs indicating actual changes in the target physiological parameters observed in the target patient in response to a course of the drug previously administered to the target patient, wherein the calculated parameter values ​​of the prediction function are calculated based at least in part on the further user inputs.

10. The method according to any one of claims 1 to 9, wherein, before the user input is received, the parameter estimation function is derived by one or more computing devices and stored in memory accessible by one or more computing devices.

11. The method according to any one of claims 1 to 10, wherein the access step and the derivation step are implemented in a first computing device, and the receiving step, the calculation step, the prediction step, and the display step are implemented in a second computing device.

12. The method according to any one of claims 1 to 12, wherein one or more parameter estimation functions are derived from the historical data using linear regression.

13. A system for predicting changes in target physiological parameters of a target patient that are expected to result from a drug administered to the target patient, wherein the system User interface and One or more memory systems for storing computer executable instructions, One or more processors, which are communicatively coupled to the one or more memory systems and the user interface, and Access historical data showing changes observed over an observation period in the target physiological parameters of multiple patients resulting from the administration of the drug, One or more parameter estimation functions, each parameter estimation function modeling how a separate parameter of a predictive function for predicting changes in the target physiological parameter over time varies according to one or more initial physiological parameters observed in the multiple patients, are derived based on the historical data. The system receives user input indicating values ​​for each of the one or more initial physiological parameters for the target patient. By applying the one or more derived parameter estimation functions to the one or more values ​​indicated by the received user input, the values ​​for each parameter of the prediction function are calculated. Using the prediction function and the calculated parameter values, the changes in the target physiological parameters of the target patient at multiple future points in time are predicted. A system comprising one or more processors operable to execute instructions for displaying the predicted changes in the target physiological parameters of the target patient on the user interface, in order to help at least one of the target patient and a healthcare professional determine whether the drug should be administered to the target patient.

14. The system according to claim 13, wherein the drug is an anti-obesity drug and the target physiological parameter is body weight.

15. The system according to claim 13, wherein the drug is an antidiabetic drug and the target physiological parameter is at the A1C level.

16. The system according to any one of claims 13 to 15, wherein the one or more initial physiological parameters include at least one of age, A1C level, height, weight, resting heart rate, and biological sex.

17. The system according to any one of claims 13 to 16, wherein the prediction function is a probability distribution function.

18. The system according to claim 17, wherein predicting changes in the target physiological parameters of the target patient includes predicting, for each of the plurality of future time points, the expected changes in the target physiological parameters and the predicted interval for the expected changes.

19. The system according to claim 17 or 18, wherein at least one of the parameter estimation functions models how a parameter representing a measure of statistical variance varies according to one or more initial physiological parameters.

20. The multiple patients in the historical data have been administered different dose levels of the drug such that the historical data shows how the changes in the target physiological parameters vary with dose levels. At least one of the derived parameter estimation functions models how one of the parameters of the prediction function changes according to the dose level. The received user input further indicates the target dose level for the drug to be administered to the target patient. The value for at least one parameter of the prediction function is calculated by one or more processors based at least partially on the target dose level. The system according to any one of claims 13 to 19, wherein, in order to assist at least one of the target patient and the healthcare professional in determining whether the target dose level of the drug should be administered to the target patient, the predicted change in the target physiological parameter of the target patient, displayed on the user interface, is calculated at least in part on the target dose level.

21. The system according to any one of claims 13 to 20, wherein one or more processors are further operable to execute instructions to receive further user input indicating actual changes to the target physiological parameters observed in the target patient in response to a course of the drug previously administered to the target patient, and the calculated parameter values ​​of the prediction function are calculated based at least in part on the further user input.

22. The system according to any one of claims 13 to 21, wherein one or more processors are operable to derive the parameter estimation function and store it in one or more memory systems before the user input is received.

23. The system according to any one of claims 13 to 22, wherein the one or more processors include a first processor capable of implementing the access step and the derivation step, and a second processor capable of implementing the receive step, the compute step, the predict step, and the display step.

24. The system according to any one of claims 13 to 23, wherein one or more parameter estimation functions are derived from the historical data using linear regression.

25. A non-temporary computer-readable medium that stores computer-executable instructions that, when executed by one or more processors, are operable to cause one or more processors to implement the method according to any one of claims 1 to 12.

Images