Dosage Calculator

Abstract

1. A method for generating a dosage calculator for determining a dosage of a drug to administer to a patient, the method comprising: receiving population data, the population data comprising actual population data and / or simulated population data calculated using a pharmacokinetic-pharmacodynamic PKPD model of the drug; and training a machine learning dosage calculator using the population data.

Description

[Technical Field]

[0001] The present disclosure relates to a method for generating a dosage calculator for determining a dosage of a drug to administer to a patient, a dosage calculator, and a method for determining a dosage. [Background technology]

[0002] Pharmacokinetic / pharmacodynamic models, such as compartmental models, are useful methods for understanding the effects of drugs on the human body. Compartmental models can be used to model the concentration of a drug in the central compartment, the region of the body where the drug exerts its effect. Thus, compartmental models can provide a relationship between drug dosage and drug effect.

[0003] Compartmental models typically use differential equations that require iterative, time-based calculations, which are resource- and time-intensive.

[0004] The present disclosure relates to methods and apparatus for determining patient drug dosage in an efficient manner that can be deployed on personal computing devices such as those found in smartphones. Summary of the Invention

[0005] According to a first aspect of the present disclosure, there is provided a method of generating a dosage calculator for determining a dosage of a drug to administer to a patient, the method comprising: receiving population data, including actual population data and / or simulated population data calculated using a pharmacokinetic-pharmacodynamic PKPD model of the drug; and training a machine learning dosage calculator using the population data.

[0006] The machine learning dosage calculator may be configured to calculate dosages to administer to a patient by processing patient data and the patient's target PKPD metrics.

[0007] The population data may include drug data for multiple patients, where each patient's drug data may include patient data, dosage data, and PKPD metrics.

[0008] The patient data may include one or more patient data parameters that can affect the effect of the drug on the patient. The patient data may include factors that affect the absorption of the drug into and / or the clearance of the drug from the body. The patient data may include one or more input parameters used in the PKPD model of the drug.

[0009] The dosage data may include the dosage of a drug administered to a patient.

[0010] PKPD metrics may include one or more parameters that indicate the effect of a drug on a patient's body.

[0011] PKPD metrics may include one or more of the following: the time profile of drug concentration at the effective site in the patient's body; the maximum drug concentration at the effective site in the patient's body; trough drug concentrations at the effective site in the patient's body; the average drug concentration at the effective site in the patient's body; Area under the curve of the time profile, the ratio of maximum drug concentration to trough drug concentration; the ratio of the maximum drug concentration to the area under the curve of the time profile, or Physiological measurements.

[0012] The population data may include simulated population data, including simulated patient data, simulated dosages, and / or simulated PKPD metrics.

[0013] The simulated population data may include simulated population data of at least 10,000 simulated patients.

[0014] The method may include calculating simulated population data using a PKPD model of the drug.

[0015] Calculating simulated population data using the PKPD model is defining simulated patient data for a patient population including a plurality of simulated patients and a corresponding plurality of simulated dosages; calculating PKPD metrics for each of the plurality of simulated patients by processing the simulated dosage and simulated patient data using the PKPD model to determine simulated PKPD metrics; and determining simulated population data including simulated patient data, simulated dosages, and simulated plasma level metrics.

[0016] The simulated patient data may include one or more of renal function metrics, liver function metrics, patient age, patient ethnicity, patient gender, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, pharmacogenomic profile, and patient medication list.

[0017] Defining simulated patient data involves: receiving a population distribution for each parameter type of simulated patient data; and generating each simulated patient by probabilistic selection of each parameter type according to a respective population distribution.

[0018] Defining simulated patient data involves: defining a number of discrete values ​​for each parameter type; and generating each simulated patient data as a different combination of one discrete value from each parameter type.

[0019] Generating each simulated patient may include generating simulated patient data representing all combinations of one discrete value from each parameter type.

[0020] The PKPD model may include a plasma level prediction model, and the PKPD metric may include a plasma level metric.

[0021] A PKPD model may include a time-based differential equation model to model the time dependence of the concentration of a drug in an effective region of the body as a function of patient data.

[0022] The PKPD model may include a two-compartment model to model the time dependence of central compartment drug concentrations as a function of patient data.

[0023] Training the machine learning dosage calculator may include training the machine learning dosage calculator using simulated population data and actual population data.

[0024] The actual population data may include actual patient data including one or more of renal function metrics, liver function metrics, patient age, patient ethnicity, patient sex, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, pharmacogenomic profile, and patient medication list. The actual population data may include actual dosage data and measured PKPD metrics for each patient.

[0025] Using population data to train a machine learning dosage calculator training a machine learning dosage calculator using simulated population data to generate a partially trained machine learning dosage calculator; and training a partially trained machine learning dosage calculator using real population data.

[0026] The method may further include validating the machine learning model using additional population data.

[0027] The additional population data may be different from the population data. The additional population data may include additional actual population data and / or additional simulated population data.

[0028] The method may further include locking the machine learning dosage calculator to prevent further adjustments to the machine learning dosage calculator.

[0029] The method may further include calibrating the machine learning algorithm for the patient based on measured PKPD metrics obtained from physiological testing of the patient.

[0030] The physiological tests may include blood tests, urine tests, cerebrospinal fluid analysis, or biopsies. The measured PKPD parameters may include drug concentrations from the physiological tests.

[0031] The machine learning dosage calculator may be configured to directly calculate dosages to administer to patients by processing patient data and target PKPD metrics.

[0032] Machine learning dosage calculator processing the patient data with a dosage calculator to determine an ideal dosage regimen; Based on the ideal dosing regimen, the dosage can be calculated directly by selecting the dosage to administer to the patient from a selection of available dosing regimens.

[0033] The patient data may include one or more of renal function metrics, liver function metrics, patient age, patient ethnicity, patient gender, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, pharmacogenomic profile, and patient medication list.

[0034] The medication may include an anticoagulant, preferably a direct oral anticoagulant, preferably dabigatran.

[0035] According to a second aspect of the present disclosure, there is provided a computer-implemented method for determining a dosage of a drug to administer to a patient, the method comprising: receiving patient data relating to a patient; processing the patient data with a dosage calculator to determine a dosage to administer to the patient, the dosage calculator comprising a machine learning algorithm trained using population data, including actual population data and / or simulated population data obtained from a pharmacokinetic-pharmacodynamic PKPD model; and indicating the dosage.

[0036] The method comprises: receiving updated patient data; processing the updated patient data with a dosage calculator to determine an updated dosage; and indicating the updated dosage.

[0037] The updated patient data may include measured PKPD metrics obtained from the patient's physiological testing.

[0038] PKPD metrics may include drug concentrations.

[0039] The method may further include calibrating the dosage calculator by adjusting the dosage calculator and / or the PKPD model using the drug concentration.

[0040] The PKPD model may include a time-based differential equation model for modeling the time dependence of the concentration of the drug in effective regions of the body as a function of patient data. The PKPD model may include a two-compartment model for modeling the time dependence of the central compartment drug concentration as a function of patient data.

[0041] Processing the patient data with a dosage calculator to determine a dosage of dabigatran to administer to the patient includes: receiving a target PKPD metric; Calculating the dosage to administer to the patient by processing the patient data and the target PKPD metrics in a dosage calculator.

[0042] The method may include determining the target PKPD metric as an individualized target PKPD metric based on the patient data.

[0043] The method comprises: processing the patient data with a dosage calculator to determine PKPD metrics; PKPD metrics and and indicating one or more of the patient risks based on the PKPD metrics.

[0044] PKPD metrics may include drug concentrations at effective sites in the patient's body.

[0045] The patient data may include one or more dosing times at which the patient received a dose of the drug. The PKPD metrics may include time-dependent drug concentrations at effective sites in the patient's body based on the one or more dosing times.

[0046] PKPD metrics may include one or more of the following: the time profile of drug concentration at the effective site in the patient's body; the maximum drug concentration at the effective site in the patient's body; trough drug concentrations at the effective site in the patient's body; the average drug concentration at the effective site in the patient's body; Area under the curve of the time profile, the ratio of maximum drug concentration to trough drug concentration; the ratio of the maximum drug concentration to the area under the curve of the time profile, or Physiological measurements.

[0047] According to a third aspect of the present disclosure, there is provided a computer-implemented method for determining a dosage of dabigatran to administer to a patient, the method comprising: receiving patient data regarding the patient, the patient data including renal function metrics; processing the patient data with a dosage calculator to determine a dosage of dabigatran to administer to the patient, the dosage calculator being derived from a plasma level prediction model; and indicating the dosage.

[0048] The patient data may further include one or more of the patient's age, the patient's ethnicity, the patient's gender, the patient's weight, the patient's hemoglobin level, the patient's left ventricular function, and the patient's medication list.

[0049] The patient's medication list may include an indication of whether the patient is taking one or more medications, including proton pump inhibitors, calcium channel blockers, nonsteroidal anti-inflammatory drugs, H2 receptor antagonists, verapamil, amiodarone, clopidogrel, aspirin, and diltiazem.

[0050] The patient data may further include one or more of reported side effects, alcohol intake, smoking history, patient's blood coagulation metrics, treatment intent, patient's genetic determinants, patient's co-conditions, patient's activity level, patient's medication compliance, patient's liver function, patient's history of thrombosis, patient's history of bleeding, patient's history of cancer, family history of thrombosis, family history of stroke, family history of bleeding, patient's cardiovascular history, patient's metabolic history, patient's blood pressure history, patient's platelet count, patient's heart rate, and patient's hematocrit.

[0051] Therapeutic objectives may include prevention of thrombosis, embolism, and / or stroke (optionally including patients with non-valvular atrial fibrillation and one or more risk factors (previous stroke or transient ischemic attack, heart failure, diabetes, or hypertension)), treatment of active thrombosis and / or treatment of active pulmonary embolism, prevention of venous thromboembolism in people who have undergone surgery (optionally including hip or knee replacement therapy).

[0052] The method comprises: receiving updated patient data; processing the updated patient data with a dosage calculator to determine an updated dosage; and indicating the updated dosage.

[0053] The updated patient data may include the patient's blood clotting metrics and / or drug concentrations from blood test results.

[0054] The method may further include calibrating the dosage calculator by adjusting the dosage calculator and / or plasma level prediction model using the patient's blood coagulation metrics and / or drug concentrations.

[0055] The plasma level prediction model may include a time-based differential equation model to model the time dependence of the plasma concentration of dabigatran as a function of patient data.

[0056] Processing the patient data with a dosage calculator to determine a dosage of dabigatran to administer to the patient includes: receiving a target plasma level metric; Calculating the dosage to administer to the patient by processing the patient data and the target plasma level metric in a dosage calculator.

[0057] Processing the patient data with a dosage calculator to determine a dosage of dabigatran to administer to the patient includes: Setting an initial dose estimate; processing the dose estimates with a dosage calculator to estimate plasma level metrics; comparing the plasma level metric to a target plasma level metric; and determining a dosage to administer to the patient by refining the dose estimate based on the comparison.

[0058] Refining dose estimates is iteratively adjusting the dose estimate and recalculating the plasma level metric until the difference between the estimated plasma level metric and the target plasma level metric is less than a difference threshold; establishing one or more additional values ​​for the dose estimation and processing the one or more additional values ​​with a dosage calculator to estimate one or more additional plasma level metrics; and determining a dosage to administer to the patient by interpolating a dosage value corresponding to the target plasma level metric based on a relationship between the initial value of the dose estimate, the one or more further values ​​of the dose estimate, the plasma level metric, and the one or more further plasma level metrics.

[0059] Target plasma level metrics may include one or more of the following: target trough plasma levels, including ideal therapeutic levels; a target maximum plasma level that is less than a maximum level threshold; target mean plasma levels over the dosing interval at steady state, including the ideal therapeutic level; a target area under the curve of the plasma level time profile that includes the ideal therapeutic level; the target ratio of maximum plasma level to trough plasma level, including the ideal therapeutic ratio and therapeutic level; or The target ratio of maximum plasma level to area under the curve of the plasma level time profile, including the ideal therapeutic ratio and therapeutic level.

[0060] The patient data may include one or more goal-dependent patient parameters including reported side effects, patient's thrombosis history, patient's bleeding history, patient's cancer history, patient's stroke history, patient's liver function metrics, patient's cardiac function metrics, patient's brain status, patient's smoking history, patient's alcohol history, patient's blood pressure, patient's activity level, patient's medication compliance, patient's mobility status, patient's menstrual status, patient's inflammation status, patient's infection status, patient's concomitant medications, patient's comorbidities, blood coagulation metrics, patient's genetic profile, family stroke history, family bleeding history, family hypertension history, patient's cardiovascular history, patient's metabolic history, patient's blood pressure history, patient's blood pressure, patient's heart rate, patient's platelet count, patient's hematocrit, and patient's hydration status, and the method Determining the target plasma level metric as an individualized target plasma level metric based on one or more target-dependent patient parameters.

[0061] The method comprises: determining a bleeding risk score for the patient based on one or more of the goal-dependent patient parameters; determining a thrombosis risk score for the patient based on one or more of the goal-dependent patient parameters; and determining an individualized target plasma level metric based on the bleeding risk score and / or the thrombosis risk score.

[0062] The method comprises: The patient's bleeding risk score was calculated as follows: time since one or more bleeding risk events, the bleeding risk events including any of trauma, surgical procedure, needle stick, or patient fall, as indicated by the patient's bleeding history; the severity of the blood clotting risk event; the patient's age, the patient's cancer history, the patient's blood pressure, Patient liver function metrics, time since hemorrhagic stroke as indicated by the patient's stroke history; the extent of amyloid in the patient's brain as indicated by the patient's brain status; blood coagulation metrics, indications for aspirin, clopidogrel, NSAIDs, steroids, or antidepressants as indicated by the patient's concomitant medications; and / or the amount of alcohol consumption indicated by the patient's alcohol history. The patient's thrombosis risk score was calculated as follows: The time from one or more blood clotting risk events, wherein the blood clotting risk events are: time, as indicated by the patient's history of thrombosis, including any of the following: thrombosis, trauma, surgical procedure, needle puncture, or patient fall; episodes of atrial fibrillation as indicated by the patient's cardiac performance metrics; chemotherapy episodes as indicated by the patient's cancer history; the severity of the blood clotting risk event; the patient's cancer history, the patient's blood pressure, patient mobility, the patient's blood pressure, blood coagulation metrics, the patient's genetic profile, time since ischemic stroke as indicated by the patient's stroke history; the patient's menstrual status, Indications for oral contraceptives as indicated by the patient's concomitant medications, and and determining the hydration status of the patient based on one or more of: Determining the individualized target plasma level metric is determined based on the bleeding risk score and / or the thrombosis risk score.

[0063] Individualized target plasma level metrics include: trough plasma levels, including individualized adjustment to ideal therapeutic levels based on one or more goal-dependent patient parameters; and / or It may include the ratio of maximum plasma level to area under the curve of the plasma level time profile that is below the bleeding risk threshold.

[0064] The ideal treatment level is Trough plasma level range of 75–150 ng / ml; a trough plasma level range of 110–115 ng / ml, or It may include a trough plasma level of 112 ng / ml.

[0065] The method comprises: processing the patient data with a dosage calculator to determine plasma level metrics; Plasma level metrics, the patient's risk of thrombosis based on plasma level metrics, or and indicating one or more of the patient's bleeding risk based on the plasma level metric.

[0066] The plasma level metrics may include a plasma level time profile. Indicating one or more of the plasma level metric, the patient's thrombosis risk, or the patient's bleeding risk may include indicating to the patient or a medical professional. The indicating may be done via a user interface.

[0067] The patient data may include one or more dosing times at which the patient received a dose of dabigatran. The plasma level metric may include a time-dependent plasma level metric based on the one or more dosing times.

[0068] Plasma level metrics are: Plasma level time profile, Maximum plasma level, trough plasma levels, mean plasma levels over the dosing interval at steady state; Area under the curve of the plasma level time profile, the ratio of maximum plasma level to trough plasma level, or The parameters may include one or more of: the ratio of maximum plasma level to the area under the curve of the plasma level time profile.

[0069] The dosage calculator may include a machine learning algorithm trained using a plasma level prediction model.

[0070] The dosage calculator may include a machine learning algorithm trained using simulated population data obtained from the plasma level prediction model.

[0071] Dosage calculator Simulated population data derived from plasma level prediction models and It may include machine learning algorithms trained using real population data.

[0072] The machine learning algorithm may be locked to prevent further adjustments to the machine learning algorithm.

[0073] The machine learning algorithm may include a tunable machine learning algorithm. receiving updated patient data including the patient's blood coagulation metrics from blood test results; and adjusting the machine learning model based on the patient's blood clotting metrics.

[0074] The dosage calculator may include a plasma level prediction model.

[0075] The dosage calculator may include one or more look-up tables defined according to simulated population data obtained from the plasma level prediction model.

[0076] Processing the patient data with a dosage calculator to determine a dosage of dabigatran to administer to the patient includes: processing the patient data with a dosage calculator to determine an ideal dosage regimen; and selecting a dosage to administer to the patient from a selection of available dosage regimens based on the ideal dosage regimen.

[0077] Available dosing regimen options may include dosages containing 75 mg, 110 mg, or 150 mg of dabigatran.

[0078] The choice of available medication regimens is any multiple of 1 mg, any multiple of 10 mg, or Dosages containing any multiple of 25 mg may be included.

[0079] The choice of available dosing regimens may include the selection of microparticulate or liquid formulations to titrate the ideal dosage for the ideal dosing regimen.

[0080] Processing the patient data with a dosage calculator to determine a dosage of dabigatran to administer to the patient may include processing the patient data with a dosage calculator to determine one or more of a dosage amount, a dosage time, a dosage frequency, and / or a dosage form.

[0081] The dosage form may contain a sustained release formulation of dabigatran with a specific release time.

[0082] A specific release time may be at least 6 hours.

[0083] Indicating the dosage may include indicating the dosage to a medical professional and / or to a patient.

[0084] Indicating the dosage may include indicating the dosage to a medical professional and / or the patient via a user interface. The user interface may include a digital app. The method may include performing one or more of the steps within the digital app.

[0085] Disclosed herein are any methods for use in stroke prevention, thrombosis treatment, thromboprophylaxis, or dosage management for invasive treatment of a patient.

[0086] The patient may be a cancer patient.

[0087] According to a fourth aspect of the present disclosure, there is provided a method for generating a dosage calculator for determining a dosage of dabigatran to administer to a patient, the method comprising: receiving simulated population data calculated using the plasma level prediction model; and training a machine learning dosage calculator using the simulated population data.

[0088] The simulated population data may include simulated population data of at least 100,000 simulated patients.

[0089] The simulated population data may include simulated patient data, simulated medication doses, and / or simulated plasma level metrics.

[0090] The method may include calculating simulated population data using a plasma level prediction model.

[0091] Using plasma level prediction models to calculate simulated population data defining simulated patient data for a patient population including a plurality of simulated patients and a corresponding plurality of simulated dosages; calculating a plasma level metric for each of the plurality of simulated patients by processing the simulated dosage and simulated patient data using the plasma level prediction model to determine a simulated plasma level metric; and determining simulated population data including simulated patient data, simulated dosages, and simulated plasma level metrics.

[0092] The simulated patient data may include one or more of a renal function metric, a patient's age, a patient's ethnicity, a patient's gender, a patient's weight, a patient's hemoglobin level, a patient's left ventricular function, and a patient's medication list.

[0093] Defining the simulated patient data may include receiving a population distribution for each parameter type of the simulated patient data and generating each simulated patient by probabilistic selection of each parameter type according to the respective population distribution. Defining the simulated patient data may include defining a plurality of discrete values ​​for each parameter type and generating each simulated patient data as a different combination of one discrete value from each parameter type. Generating each simulated patient may include generating simulated patient data for all possible combinations of one discrete value from each parameter type.

[0094] The plasma level prediction model may include a time-based differential equation model to model the time dependence of the plasma concentration of dabigatran as a function of patient data.

[0095] The plasma level prediction model may include a compartmental model to model the time dependence of the plasma concentration of dabigatran as a function of patient data.

[0096] Training the machine learning dosage calculator may include training the machine learning dosage calculator using simulated population data and actual patient data.

[0097] The method may include validating the machine learning model using further simulated population data, where the further simulated population data may differ from the simulated population data.

[0098] The method may include locking the machine learning dosage calculator to prevent further adjustments to the machine learning dosage calculator.

[0099] The method may further include calibrating the machine learning algorithm for the patient based on measured drug plasma levels or coagulation measurements obtained from the patient's blood tests.

[0100] The machine learning dosage calculator may be configured to directly calculate the dosage to administer to the patient by processing the patient data and the target plasma level metric.

[0101] The patient data may include one or more of a renal function metric, a patient's age, a patient's ethnicity, a patient's gender, a patient's weight, a patient's hemoglobin level, a patient's left ventricular function, and a patient's medication list.

[0102] According to a fifth aspect of the present disclosure, there is provided a dosage calculator for determining a dosage of dabigatran to administer to a patient, the dosage calculator comprising: receiving patient data regarding the patient, the patient data including renal function metrics; processing the patient data with a dosage calculator to determine a dosage of dabigatran to administer to the patient, the dosage calculator being derived from a plasma level prediction model; and indicating the dosage.

[0103] According to a sixth aspect of the present disclosure, there is provided a computer-readable medium containing instructions that, when executed by one or more processors, cause the one or more processors to perform any of the methods disclosed herein.

[0104] According to a seventh aspect of the present disclosure, there is provided a method for administering a dosage of a drug to a patient, the method comprising: administering a dosage of a drug to the patient, wherein the dosage of the drug comprises: receiving patient data relating to a patient; and processing the patient data with a dosage calculator to determine a dosage of the drug to administer to the patient, the dosage calculator including a machine learning algorithm trained using population data including simulated population data obtained from a pharmacokinetic-pharmacodynamic PKPD model.

[0105] A computer program may be provided that, when executed on a computer, causes the computer to configure any apparatus, including the circuits, controllers, converters, or devices disclosed herein, or to perform any method disclosed herein. The computer program may be software implemented, and the computer may be considered to be implemented in any suitable hardware, such as, but not limited to, a digital signal processor, a microcontroller, and implementation in a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or an electronically erasable programmable read-only memory (EEPROM). The software may be an assembly program.

[0106] The computer program may be provided on a computer-readable medium (which may be a physical computer-readable medium, such as a disk or memory device) or may be embodied as a transitory signal. Such a transitory signal may be a network download, including an Internet download. One or more non-transitory computer-readable storage media may be provided that store computer-executable instructions that, when executed by a computing system, cause the computing system to perform any of the methods disclosed herein.

[0107] One or more embodiments will now be described, by way of example only, with reference to the accompanying drawings. [Brief explanation of the drawings]

[0108] [Figure 1] 1 shows the relationship between dabigatran plasma levels and the probability of clinical outcomes. [Figure 2] 1 illustrates a method for determining a dosage of a drug to administer to a patient, according to one embodiment of the present disclosure. [Figure 3] 1 illustrates an exemplary two-compartment model that may provide a PKPD model according to one or more embodiments. [Figure 4A] 1 shows a comparison of dabigatran plasma levels on day 10 for the Liesenfeld method and a plasma level prediction model according to an embodiment of the present disclosure. [Figure 4B] 1 shows a comparison of the relationship between renal function and plasma levels as determined by the Leisenfeld model and a plasma level prediction model according to an embodiment of the present disclosure for a fixed dose of 150 mg. [Figure 5] 1 illustrates a model for determining a dosage of a drug to administer to a patient, according to one embodiment of the present disclosure. [Figure 6] 1 illustrates a method for generating an ML dosage calculator according to one embodiment of the present disclosure. [Figure 7] A subsample of simulated population data is shown. [Figure 8] 1 shows validation results of a dense neural network-based ML dosage calculator, according to one embodiment of the present disclosure. [Figure 9] 10 shows validation results of a dense neural network-based ML dosage calculator according to another embodiment of the present disclosure. [Figure 10] 1 illustrates a system suitable for performing any of the computer-implemented methods disclosed herein. [Figure 11] 1 shows the estimated plasma level time profile for a simulated patient taking a 25 mg dabigatran dosage twice daily on day 10, calculated according to one embodiment of the present disclosure. [Figure 12] 1 illustrates an exemplary patient monitoring process according to one embodiment of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0109] overview The present disclosure relates to a method for generating a drug dosage calculator that includes a machine learning (ML) model (also referred to as an ML algorithm) trained using population data. The population data may include drug data for a plurality of patients, and may include actual population data from one or more real patient studies and / or simulated data output from a pharmacokinetic-pharmacodynamic PKPD model of the drug.

[0110] FIG. 6 illustrates a method for generating an ML dosage calculator according to one embodiment of the present disclosure.

[0111] The first step 638 involves receiving population data, including actual population data and / or simulated population data calculated using a PKPD model.

[0112] The second step 640 involves using the population data to train the ML dosage calculator.

[0113] The population data may include drug data for multiple patients. For each patient, the drug data may include patient data, dosage data, and PKPD metrics.

[0114] The patient data may include one or more patient data parameters that can affect the effect of the drug on the patient. For example, the patient data may include factors that affect the absorption and / or clearance of the drug into and from the body. The patient data may include one or more input parameters used in the PKPD model of the drug. The patient data may include, among others, one or more of renal function metrics, liver function metrics, patient age, patient ethnicity, patient sex, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, a pharmacogenomic profile (e.g., a CYP enzyme profile and / or a drug transporter profile), and the patient's medication list.

[0115] Dosage data can include the dosage of a drug administered to a patient, which can include a nominal dosage and a duration of administration, e.g., a dosage administered daily for 10 days.

[0116] PKPD metrics may include any parameter that indicates the effect of a drug on a patient's body. For example, PKPD metrics may include drug concentrations in specific compartments of the body, such as plasma concentrations, tissue concentrations, urine concentrations, etc. PKPD metrics may be related to physiological measurements (e.g., blood pressure or electrocardiogram). The physiological measurements may be metrics for monitoring a disease or condition treated by the drug.

[0117] Real population data The actual population data may include multiple drug data collected from one or more patient studies or from a drug monitoring program. Each patient record of drug data may include actual patient data (e.g., age, weight, renal function metrics (such as creatinine clearance), BMI, concomitant medications, etc.), actual dosage data including the drug dosage administered to the patient, and actual (measured) PKPD metrics (such as blood drug concentrations or other physiological measurements).

[0118] Simulated population data The simulated population data may be obtained from a PKPD model for a drug. The simulated population data may include simulated drug data for a plurality of simulated patients. The simulated drug data for each simulated patient may include simulated patient data (e.g., age, weight, renal function metrics (e.g., creatinine clearance), BMI, concomitant medications, etc.), simulated dosage data including simulated drug dosages administered to the patient, and simulated PKPD metrics (e.g., blood drug concentrations) calculated by processing the simulated patient data and simulated dosage data with the PKPD model.

[0119] The PKPD model may include any time-dependent model capable of outputting the dependence of a PKPD metric based on received inputs including drug dosage and (actual or simulated) patient data. For example, the PKPD model may include a compartmental model, such as a two-compartment model, to model the time dependence of central compartment drug concentration as a function of dosage and patient data. The PKPD metric may include the concentration of the drug in the central compartment. The central compartment may be associated with one or more regions of the body where the drug acts, such as blood, liver, tissue, or brain. As discussed further below, the PKPD metric may include one or more other metrics, such as a physiological measurement (blood pressure, electrocardiogram, etc.), maximum drug concentration, trough drug concentration, mean drug concentration, time profile of drug concentration, area under the curve of the time profile, or one or more combinations or ratios thereof.

[0120] The patient data may include any input variables used in the PKPD model. For example, the patient data may include one or more variables that can affect drug absorption and / or clearance of the drug from the body. For example, in embodiments, the patient data may include, among others, one or more of: renal function metrics, liver function metrics, patient age, patient ethnicity, patient sex, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, a pharmacogenomic profile (e.g., a CYP enzyme profile and / or a drug transporter profile), and a patient medication list.

[0121] Data output from the PKPD model may include simulated population data. The simulated population data may include data output from the PKPD model after processing of simulated patient data, which may represent known population variability in a population. For example, a Monte Carlo-type approach may be used to generate the simulated patient data using known distributions for each parameter type of the patient data (see below). In some examples, the simulated population data may include PKPD metrics output by the PKPD model. In some examples, the simulated population data may include dosages to administer to each patient in a population set to achieve target PKPD metrics. An ML model may be trained using the simulated population data to provide an ML dosage calculator.

[0122] 3 shows an exemplary two-compartment model that may provide a PKPD model, according to one or more embodiments. The central compartment 301 may represent a target site in the body where a drug provides a therapeutic effect. The central compartment 301 has a central compartment drug concentration (Cc) and a central compartment volume (Vc). The drug uptake into the central compartment 301 from an administered drug may be defined by:

number

[0123] Drug clearance from the central compartment 301 (e.g., via the kidney) may be defined as:

number

[0124] The peripheral compartment 303 may represent a site in the body where a drug may be present but does not necessarily provide a therapeutic effect. The peripheral compartment 303 has a peripheral compartment drug concentration (Cp) and a peripheral compartment volume (Vp). The rate of change of the peripheral compartment drug concentration Cp due to transfer of drug between the two compartments may be defined as:

number

[0125] Combining all of this, the rate of change of drug concentration in the central compartment (Cc) can be written as:

number

[0126] The method may include calculating simulated population data using a PKPD model. For example, the method may include (i) defining simulated patient data for a simulated patient population including a plurality of simulated patients, and (ii) calculating PKPD metrics for each of the plurality of simulated patients using the PKPD model to define the simulated population data. The simulated population data may include simulated patient data, simulated dosages, and calculated PKPD metrics.

[0127] The simulated patient population may include at least 10,000 simulated patients, at least 100,000 simulated patients, at least 1,000,000 simulated patients, or at least 10,000,000 simulated patients. Defining the simulated patient data may include (i) receiving a population distribution for each parameter type of the patient data (e.g., age, weight, other medications, liver function, renal function, pharmacogenomic profile, etc.) and (ii) generating a plurality of simulated patients through probability-based selection of each parameter type according to the respective population distribution. For example, Monte Carlo-type analysis may be used to define the simulated patient data. In other examples, the simulated patient data may include a more systematic coverage of the parameter space defined by the patient data. In some examples, the simulated patient data may also include simulated dosages. The simulated patient data (including simulated dosages) may be processed using a PKPD model to obtain simulated population data including simulated PKPD metrics.

[0128] In some examples, defining the simulated patient data may include (i) defining a plurality of discrete values ​​for each parameter type and (ii) generating each simulated patient data as a different combination of one discrete value from each parameter type. In some examples, the different combinations may generate simulated patient data representing all combinations of one discrete variable from each parameter type. The parameter values ​​for a particular parameter type may include discrete steps between each parameter value that are fine enough to represent a continuous variable, according to the biological range of the parameter important for determining drug levels.

[0129] Figure 7 shows a subsample of simulated population data for the PKPD model of dabigatran. The simulated population data includes simulated dose 742, simulated patient data 744 (age, sex, CrCl, hemoglobin level, weight, ethnicity, heart failure, medication (proton pump inhibitor), amiodarone, verapamil), and simulated plasma level metrics 746, including trough plasma level on day 10 and maximum plasma levels on days 0 and 10.

[0130] Training the ML Dosage Calculator After receiving the population data, the ML dosage calculator can be trained using the population data. The ML dosage calculator can include any known ML architecture, such as an artificial neural network or a generative model. In some embodiments, the population data can include both actual population data and simulated population data from a plasma level prediction model as ML training data. The ML training data can include each population data set, with a higher weighting assigned to the actual population data than to the simulated population data.

[0131] In some examples, training of the ML dosage calculator may occur in two steps. In a first, pre-training step, the ML dosage calculator may be trained using simulated population data. A second, refinement step may include further using the refined training data to train the ML dosage calculator. In some examples, the refined training data may include real population data, such as clinical data from real patients, including patient data, associated drug dosages, and resulting PKPD metrics. In some examples, the refined training data may relate to a different drug with a similar PKPD pathway as the drug used in the first pre-training step. In this way, the ML dosage calculator may be pre-trained in a general manner and then refined using real-world patient data and / or for related drugs with similar underlying structures.

[0132] ML Dosage Calculator Operation The present disclosure also provides a method for determining a dosage of a drug to administer to a patient, utilizing an ML dosing calculator trained using population data. The method may determine the dosage without requiring physiological measurements. The ML dosage calculator may process patient data for an individual patient to indicate a dosage to administer to the patient. The ML dosage calculator may receive target PKPD metrics and calculate a dosage to administer to the patient by processing the patient data and the target PKPD metrics. The PKPD metrics may include an ideal therapeutic level (ITL) or an ideal drug concentration.

[0133] FIG. 2 illustrates a method 210 for determining a dosage of a drug to administer to a patient, according to one embodiment of the present disclosure.

[0134] The first step 212 involves receiving patient data for an individual patient. The patient data may be of the same type as that used to train the ML dosage calculator.

[0135] The second step 214 involves processing the patient data with an ML dosage calculator to determine the dosage of the drug to administer to the patient.

[0136] The third step 216 involves indicating the dosage.

[0137] In a first set of examples, the ML dosage calculator may perform the second step 214 by replicating the PKPD model and determining PKPD metrics for specific dose and patient data. The dosage calculator may perform the calculations iteratively and revise the dosage estimates to obtain a dosage to administer to the patient such that the calculated PKPD metrics match a target PKPD metric, such as an ideal therapeutic level (ITL).

[0138] FIG. 5 illustrates a method for determining a dosage to administer to a patient (step 214) according to one embodiment of the present disclosure.

[0139] The first step 526 involves setting an initial value for the dosage estimate, for example, based on received patient data.

[0140] The second step 528 involves processing the dosage estimates with a dosage calculator to determine PKPD metrics, which may include trough drug concentration (Ctrough), maximum drug concentration (Cmax), the ratio of maximum to trough drug concentration, the average drug concentration over the dosing interval at steady state (Caverage) (i.e., after the drug has accumulated and stabilized over several days), or any other drug concentration metric described herein.

[0141] The third step 530 involves comparing the PKPD metrics to target PKPD metrics, which may include the same metrics as the PKPD metrics (Ctrough, Cmax, Coverage, etc.).

[0142] A fourth decision step 532 involves determining whether the difference between the values ​​of the PKPD metric and the target PKPD metric is within a difference threshold. If not, the method proceeds to a fifth step 534 to refine the dosage estimate before returning to the second step 528.

[0143] If the difference between the two values ​​is within the difference threshold, the method proceeds to a sixth step 536 and outputs or indicates the dosage estimate as the dosage to administer to the patient.

[0144] In some embodiments, a loop around the second through fifth steps 528-534 may be performed iteratively until the value falls within the difference threshold. In some embodiments, the loop may be performed at least twice, and a dosage value corresponding to the target PKPD metric may be interpolated. In some embodiments, the loop may correspond to an optimization routine.

[0145] Figure 8 shows the validation results of the dense neural network-based ML dosage calculator. For the specific example of dabigatran, described further below, the ML dosage calculator was trained on simulated population data of 384,000 simulated patients. The ML dosage calculator was then tested on simulated population data of 1,008,000 simulated patients. The simulated population data was generated using plausible individual demographic, physiological, or interacting drug combinations for the simulated patient data (e.g., renal clearance 200-25 ml / min, body weight 40-200 kg, age 18-110 years, presence or absence of co-administered medications, etc.). The test simulation data included CrCl levels not generated in the training dataset to provide additional rigor. The simulated population data was obtained based on Equations 1-7 using a PKPD model, including the plasma level prediction model described below. The validation results are divided into two plots: (i) a first plot 848 showing the ML dosage calculator's prediction of whether a dosage is optimal for dosages identified as suboptimal by the plasma level prediction model, and (ii) a second plot 850 showing the ML dosage calculator's prediction of whether a dosage is optimal for dosages identified as optimal by the plasma level prediction model. The x-axis scale of both plots 848, 850 is the ML model's prediction confidence of whether a plasma level is within the ideal therapeutic level (ITL), with 0 corresponding to a day 10 trough plasma level outside the range of 75-150 ng / ml and 1.0 corresponding to a day 10 trough plasma level within the range of 75-150 ng / ml. A threshold of 0.5 is set to separate the classification into optimal and non-optimal doses. Figure 8 shows that the ML dosage calculator properly identifies optimal and non-optimal dosages for approximately 99% of the simulated population data.It will be appreciated that, although not certain, further refinements and improvements can be made, such as improving training data (more data, more realistic data, actual data), and performance may be superior to subjective assessments of dosage by healthcare professionals (HCPs) (e.g., clinicians) who may not have complete knowledge of patient data.

[0146] In a second set of examples, the ML dosing calculator can receive target PKPD metrics and patient data and directly determine the dosage to administer to the patient. Figure 9 shows validation results for a dense neural network-based ML dosing calculator. The ML dosing calculator was trained on simulated population data of 384,000 simulated patients for the specific example of dabigatran. The ML dosing calculator was then tested on simulated population data of 57,500 simulated patients. The simulated population data was obtained based on Equations 1-7 using a PKPD model including the plasma level prediction model described below. The test population data was limited to data for which the plasma level prediction model indicated trough plasma levels in the range of 110-115 ng / ml. The x-axis corresponds to the dosage used in the plasma level prediction model to provide trough plasma levels in the range of 110-115 ng / ml. The y-axis corresponds to the dosage predicted by the ML dosage calculator to provide trough plasma levels in the range of 110–115 ng / ml. The vertical spread at each point on the plot corresponds to the accuracy of the ML dosage calculator relative to the plasma level prediction model. Given the narrow range of target trough plasma levels, the ML dosage calculator outputs only a limited spread of recommended dosages. This directly demonstrates the strong performance of the ML dosage calculator.

[0147] A key advantage of the ML dosage calculator is that it can operate significantly faster than two-compartment differential equation models. Efficient processing means that the ML dosage calculator requires much less processing power than calculators that attempt to directly implement the differential equations of a two-compartment model. Such low processing power requirements enable large-scale deployment of the ML dosage calculator to HCPs and / or patients, for example, via cloud platforms and / or personal computing devices. Other advantages of the ML dosage calculator include: (i) a wealth of rapidly available data, including optimized dosage and PKPD metrics such as trough levels, Cmax, ratios, or time profiles (similar to the profile in Figure 4A). Such time profiles can be used to indicate time-dependent risks to patients (see the "Output" section below). And (ii) precise, individualized, and accurate dosing. Furthermore, the model enables continuous learning, including incorporating new adjustment parameters as data accumulates. Such adjustments can be performed at regulated stages or steps to ensure ongoing regulatory compliance and safety.

[0148] In some examples, the ML dosage calculator may be locked after training and validation, preventing the ML dosage calculator from "learning" from any future data. Such an approach can improve safety and facilitate regulatory approval by ensuring that the ML dosage calculator does not change to an unsafe regimen based on errors in further training data. Alternatively, in some instances where safety requirements may be more relaxed, such as when a drug is used in a terminal illness, a highly monitored environment, or a clinical situation where the HCP has no alternatives, the ML dosage calculator may be free to change and use live data as further training data to provide a more accurate dosage calculator.

[0149] The benefits of training an ML dose calculator using simulated population data from a validated PKPD model relate to the safety, reliability, and regulatory approval of the ML model. As illustrated by the above examples, a large number of patients may be required to successfully train and maximize the benefits of an ML dosage calculator. Obtaining patient data for a comparable number of real patients for a particular drug can be time-consuming and expensive. As a result, ML models may struggle to gain regulatory approval, especially if part of the training phase is proposed to be performed as part of deployment due to the lack of predictability of such an approach. Embodiments of the present disclosure relate to training and validating ML models using large simulated patient datasets based on validated and approved analytical models developed from real (and achievable) patient data. The resulting ML dosage calculator is safe, predictable, and can be locked to prevent further training (potential exceptions being defined, regulatory-approved, update points).

[0150] A still further advantage of training an ML dosage calculator with simulated data is that it can provide outputs for covariate combinations that are not well represented in the original patient dataset. In this way, the ML dosage calculator can reduce bias and poor predictions by artificially enriching for specific cases of interest and improving the weighting and consideration of less common characteristic combinations.

[0151] Furthermore, large-scale deployment of safe, predictable, and efficient ML dosing algorithms to patients and HCPs can enable the collection of real patient data, including measured PKPD metrics such as drug plasma levels (see the "Continuous Therapy Management" section below). Such deployment is not feasible with two-compartment differential equation-based PKPD models due to processing constraints. The collected real patient data can then be used to alter the ML dosage calculator and / or PKPD model, further improving the accuracy and precision of the dosage calculator.

[0152] The following description often refers to the development, use, and advantages of the ML dosage calculator in relation to the specific example drug, dabigatran. However, the development, use, and advantages of the described ML model are applicable to any drug whose absorption and clearance can be modeled by pharmacokinetic and / or pharmacodynamic modeling, such as two-compartment differential equation-based models. For example, drugs including vancomycin, phenobarbital, butalbital, and etomidate can be modeled using two-compartment PKPD models. Other drugs can be modeled using PKPD models similar to those in Equations 1-4 above. As with Equations 5-7 described below, specific refinements for other drugs may be provided by corresponding actual patient studies and PKPD analyses for the corresponding drugs, as well as contributing individual patient data (which may differ from the patient data listed for dabigatran). In the case of dabigatran, the central compartment comprises plasma. For other drugs, the central compartment may represent sites other than plasma, such as the liver, muscle tissue, or brain. As a result, a dosage calculator can be generated from actual and / or simulated population data output from a PKPD model of a drug that predicts the time dependence of the appropriate PKPD metric. In the case of dabigatran, the PKPD metric is a plasma level metric, although the PKPD metric can include other metrics such as drug concentration in central compartments (e.g., liver, muscle tissue, etc.).

[0153] Example - Dabigatran Pharmacokinetics (Pk) - Plasma Level Variability The body's blood clotting system has shifted to a specific equilibrium, which represents a trade-off. Blood clotting is a protective factor that helps repair internal blood vessel damage or trauma. However, too much blood clotting can block blood vessels when this is not desired. This clot can then break down and embolize into distant blood vessels, potentially with devastating consequences. The multiple molecular components involved in blood clotting are constantly being replaced.

[0154] Thrombi can form anywhere in the vascular system, particularly in the veins of the legs, the arteries of the chest and neck, and within the heart. Thrombi that embolize from the heart or cervical arteries or form directly within the cerebral vasculature can cause a stroke.

[0155] When the ventricles are enlarged or not contracting normally, the likelihood of a blood clot forming in the heart increases. Atrial fibrillation (AF), characterized by uncontrolled electrical activation of the heart's atria and contraction of the myocardium, is an arrhythmia that accounts for approximately 30% of all hospitalizations and occurs in nearly 10% of people over the age of 80, with prevalence increasing with the aging population. The clinical consequences of uncontrolled AF include a five-fold increase in strokes and blood clots, the need for hospitalization, and consequent increased healthcare costs, currently estimated at $8 billion annually in the United States alone.

[0156] Management of AF involves methods to restore normal sinus rhythm, control heart rate, and, if possible, prevent recurrence. If these methods fail, anticoagulants are used to inhibit blood clot formation. In the past, warfarin was used successfully, significantly reducing strokes by approximately 60%. However, its narrow therapeutic range, metabolic variations between individuals, and multiple interactions with several concomitant medications and foods can lead to serious bleeding risks. Warfarin is difficult to prescribe at the appropriate dose and requires frequent measurement of blood clotting time (international normalized ratio, INR) to allow for periodic dose adjustments.

[0157] Another condition for which anticoagulants are used is the treatment of thromboembolism, in which a blood clot forms in a vein (VTE), which, if untreated, can lead to disability (pain, scaling ulcers, leg edema), deep vein thrombosis (DVT), and, if the clot breaks off, pulmonary embolism and death. There are approximately 600,000 cases of VTE in the United States each year.

[0158] New drugs have been developed that allow for simplified dose management compared to warfarin and have been shown to reduce the likelihood of major bleeding events. These direct-acting oral anticoagulants (DOACs) act through different, more targeted mechanisms. While warfarin and other similar anticoagulants are indirect inhibitors of vitamin K via both the intrinsic and extrinsic pathways, DOACs act directly on a common pathway downstream of the blood coagulation cascade. For example, dabigatran, one DOAC, acts directly by inhibiting the thrombin molecule in both its free and bound forms. Thrombin plays a central role in clot formation. These new mechanisms of action are particularly important in that they require less monitoring, require less frequent follow-up, have a more immediate onset and offset of drug effects, link plasma drug levels to action, and have fewer drug and food interactions.

[0159] Dabigatran etexilate is a NOAC approved for the prevention of stroke in nonvalvular atrial fibrillation and the treatment and prevention of venous thromboembolism and has been studied in thousands of patients. In the Phase III RELY trial, dabigatran was shown to be superior to warfarin in patients with nonvalvular AF, with similar or lower rates of both ischemic stroke and major bleeding compared with adjusted-dose warfarin (INR 2.0-3.0), and with a slightly reduced risk of intracranial hemorrhage. Dabigatran also reduces the tendency to form blood clots and is therefore used to reduce the incidence of stroke from other systemic embolic causes in patients.

[0160] DOACs such as dabigatran also have drawbacks, including a lack of efficacy and safety data in patients with severe chronic kidney or liver disease or severe valvular disease, a lack of readily available blood level and follow-up monitoring, and high patient costs in some medical fields. Furthermore, while reversal agents are now available for some DOACs, they are expensive and do not target all forms of bleeding.

[0161] Three doses of dabigatran (75, 110, and 150 mg) taken twice daily are available in Europe, but only two (75 and 150 mg) are available in the United States. Prescriptions may combine tablet doses to achieve recommended posology for individuals who may not meet the usual criteria.

[0162] Restrictions in dosage units and prescribing guidelines can result in very limited dosing flexibility for healthcare professionals (HCPs). As a result, patients may be prescribed an inappropriate starting dose, and some patients, for example, those with impaired renal function, may be excluded from dabigatran treatment. FDA real-world evidence studies provide evidence that underdosing with current DOACs can lead to excessive thrombotic events, and overdosing, especially in the presence of impaired renal function, can lead to excessive bleeding.

[0163] The problem of inappropriate starting doses (or appropriate starting doses) can be further exacerbated by inadequate monitoring. Unlike warfarin, routine monitoring of blood coagulation in patients taking DOACs is not currently recommended except for certain patients, particularly those with subclinical thrombosis, renal failure, the elderly, or those taking certain concomitant medications. In these latter groups, monitoring is necessary but often not performed. The recommendation is that such patients should be reviewed at least annually. This recommendation is rarely adhered to and often not reviewed at all. Reasons include delegation to general practitioners and general physicians, who may be busy and / or lack sufficient information and understanding. Furthermore, even when hematology specialists provide clear instructions to primary care, these instructions are often not adequately followed.

[0164] An additional challenge resulting from inadequate monitoring is managing risk during surgical invasive procedures such as hip and knee replacement or lumbar puncture. Perioperative patient management involves careful assessment of the relative risk of bleeding or the likelihood of a thromboembolic event. Current guidelines are one-size-fits-all, resulting in some patients being stopped from anticoagulation too early, thus placing them at higher risk of thrombosis, and others being stopped from anticoagulation too late, placing them at higher risk of bleeding. In addition to the risk to the patient, there are secondary harms due to prolonged hospital stays and bed occupancy while waiting for the anticoagulation effect to wear off. Furthermore, delays in investigations can occur. For example, an unplanned lumbar puncture, which may be needed in the short term to diagnose a neurological condition, may be delayed for an unnecessarily long period due to concerns about ongoing anticoagulation.

[0165] A particular area of ​​further need relates to the risk of thrombosis in cancer. This is particularly difficult to manage in cancer because not only is the risk of thrombosis increased, but so is the risk of bleeding. As chemotherapy becomes more home-based, this is becoming an increasing problem in cancer because of the increased insertion of chemotherapy lines, further increasing the risk of thrombosis. In particular, the need for precise control and knowledge of the actual risk of bleeding would be very helpful.

[0166] Dabigatran is formed by hydrolysis of the dabigatran etexilate ester in the gastrointestinal tract; only approximately 5% of the drug is absorbed, with the remainder (approximately 90%) excreted in the feces. This may be the result of saturation of the Pgp transporter system through which the drug is absorbed. Peak absorption occurs approximately 1 hour later; food does not affect total absorption, but may delay the peak by up to 2 hours. Interestingly, removing the pellet from the capsule increases availability by 75%. After absorption, the drug is slowly excreted with a half-life of approximately 12–17 hours; therefore, the drug is often administered twice daily to maintain reasonably constant levels. Total and peak systemic exposure are proportional to the dose, ranging from 50–400 mg twice daily.

[0167] The primary route of dabigatran excretion is via the kidneys (80% of total clearance), with the remainder being metabolized to equally active acylglucuronides, which account for an average of approximately 10% of the parent drug in plasma. In some cases, this may contribute to the drug's activity. 80% of absorbed drug is excreted in the urine. There is wide variability in the plasma concentrations achieved at any given dose, depending on absorption, renal function, and other patient factors. However, the patient package insert does not provide guidance on dosage modifications, except for renal impairment, as follows: "No Pradaxa dose adjustment is recommended for patients with mild or moderate renal impairment [see Clinical Pharmacology (12.3)]. Reduce the Pradaxa dose in patients with severe renal impairment (CrCl 15-30 mL / min) [see Dosage and Administration (2.1) and Clinical Pharmacology (12.3)]. Dosing is not recommended for patients with a CrCl <15 mL / min or on dialysis."

[0168] However, 43% of patients with atrial fibrillation and impaired renal function were potentially overdosed and had a high hazard ratio for major bleeding.

[0169] Dabigatran pharmacokinetics and interactions with other drugs known to affect Pgp result in drug- and time-dependent dabigatran absorption. Ketoconazole increases total exposure by approximately 1.4-fold, while verapamil increases exposure by 2.4-fold compared to dabigatran, depending on the timing of drug administration. Quinidine increases total absorption by only 53% and maximum concentration (Cmax) by 56%. These increases may increase the potential for bleeding, depending on the individual.

[0170] Because of the significant impact of renal function on dabigatran PK and increased bleeding, US guidelines for dabigatran use are complex and exclude patients who may benefit from dabigatran if it were prescribed and monitored more sophisticatedly. The guidelines are as follows: · For patients with creatinine clearance (CrCl) greater than 30 mL / min, the dose is 110–150 mg twice daily (approximately 12 hours apart); · For patients with CrCL 15-30 mL / min, the dose is 75 mg orally twice daily, and concomitant use of P-gp inhibitors or antivirals should be avoided; Dabigatran is contraindicated in patients with a CrCl <15 mL / min or in patients undergoing dialysis.

[0171] However, there is still a large inter-subject variability in the coagulation measurements obtained and outcomes between under-dosing (increased morbidity and mortality) and over-dosing (increased incidence of bleeding and gastrointestinal disturbances).

[0172] Furthermore, the conversion process from the prodrug dabigatran etexilate and the low absorption rate from the gastrointestinal tract (only 3-7% of dabigatran is actually absorbed from the gastrointestinal tract) can result in large intersubject variability, contributing to fluctuations in circulating plasma levels (up to 50%).

[0173] Several studies have shown a relationship between plasma drug levels and clinical outcomes, suggesting that a trough level of 75–150 ng / ml provides the best balance between reducing stroke and minimizing bleeding. However, measuring drug levels in general practice is expensive and rarely performed. Trough levels are the plasma levels before the next scheduled dose (typically the morning dose). There appears to be a relationship between drug levels and specific measures of dabigatran's activity on blood clotting (i.e., dilute thrombin time (dTT) and ecarin thrombin time (ECT)), which could form a method for monitoring patient outcomes.

[0174] In summary, the large interpatient variability in drug plasma levels occurs due to the influence of renal function, the effects of other drugs, low absorption rates, and other factors (discussed further below). Figure 1 (adapted from Reilly [1]) illustrates the relationship between dabigatran plasma levels and the probability of clinical outcomes. The figure shows a first line and shaded area 102 representing the event probability of ischemic stroke and a second line and shaded area 104 representing the event probability of major hemorrhage. The figure also shows a first range 106 of trough plasma levels measured for a patient population administered a first dosage of 110 mg dabigatran (administered to the first patient population) and a second range 108 of trough plasma levels measured for a patient population administered a second dosage of 150 mg dabigatran (administered to the second patient population). The two doses used in the study demonstrate a very wide range of trough plasma levels. For the 110 mg dose, trough plasma levels ranged from approximately 25 to 160 ng / ml, while for the 150 mg dose, they ranged from approximately 40 to 220 ng / ml, indicating large inter-subject differences in drug absorption and clearance from the body (due to the factors listed above and described herein).

[0175] The method of Figure 2 can be advantageously implemented using an ML dosage calculator to determine the dosage of dabigatran to administer to a patient. The ML dosage calculator can be trained using actual population data and / or simulated population data from a PKPD model that includes a plasma level prediction model.

[0176] In the case of dabigatran, the patient data includes a renal function metric (which may be measured creatinine clearance). In some examples, the renal function metric may include other renal function metrics, such as inulin cystatin C, beta trace protein, 51Cr-EDTA (radioactive chromium complexed with ethylenediaminetetraacetic acid), or 99Tc-EDTA (radioactive technetium complexed with ethylenediaminetetraacetic acid). As described above and below, renal function is the most important contributor to the variability of dabigatran drug plasma levels between patients. By including a renal function metric, the present method advantageously accounts for renal function, which is the largest contributor to the variability of dabigatran drug plasma levels.

[0177] The term dabigatran drug plasma level may also be referred to herein as dabigatran plasma level, plasma level, drug plasma level, or simply plasma level. The term "level" may also be referred to as concentration. As described below, embodiments of the present method can process patient data (including multiple parameters that contribute to plasma level variability, such as demographic data, medication data, and patient condition data) to advantageously provide more accurate medication dosages to patients.

[0178] For any drug, the method of FIG. 2 can calculate a patient's starting dose. The method can also continue to monitor patient data and provide the patient with a revised dose as the patient data changes. In the specific example of dabigatran, the method can address the provision of an inaccurate starting dose and the lack of continuous monitoring of the DOACs described above. As explained below, the method may also suggest a dosage that differs from the available standard dosage. In some embodiments, the method may recommend a new dosage provided by a new dosage form, including a liquid or microgranular formulation, or other known solubility-enhancing formulation. As discussed above, microgranules can improve absorption, which can reduce inter-subject variability in drug plasma levels. This can advantageously make dabigatran available to patients for whom the drug is currently contraindicated and provide precise dosing for patients for whom achieving adequate plasma levels is important (e.g., cancer patients).

[0179] As also described below, for any drug, the method of FIG. 2 can also encompass personalization by providing individualized target PKPD metrics on an individual basis based on patient data.

[0180] PKPD model of dabigatran - Plasma level prediction model In connection with the second step 214, the PKPD model for dabigatran may include a plasma level prediction model. The plasma level prediction model includes a theoretical model capable of predicting dabigatran drug plasma levels based on dosage and patient data. In some examples, the plasma level prediction model may be a time-dependent differential equation-based (or rate equation-based) model. The plasma level prediction model may include a compartmental model, such as a two-compartment model, or any other PKPD analytical model for predicting drug plasma levels as a function of patient data.

[0181] Estimates of specific parameters for the two-compartment model (Equations 1–4) above can be obtained from patient studies. For example, Liesenfeld [2] generated a two-compartment model for dabigatran. Their model was based on a population pharmacokinetic analysis of all data from a large, prospective, randomized, open-label study (RE-LY) comparing two doses of dabigatran (110 and 150 mg) with warfarin over a two-year period. Plasma levels were collected before and approximately 2 hours after the start of treatment, 4 weeks after the initiation of treatment, and additionally at 3, 6, and 12 hours from a subset of patients. Various covariates (patient data) considered relevant to the drug's PK were collected, including age, weight, ethnicity, renal function, creatinine clearance (CrCl), left ventricular function, heart failure, hemoglobin, and PgP inhibitors (including verapamil, amiodarone, clopidogrel, diltiazem, proton pump inhibitors, and H2 receptor antagonists). The analysis showed the individual effects on total drug exposure as measured by plasma area under the curve (AUC), as listed in Table 1 below. [Table 1] Where Vd = volume of distribution of the drug; AUC = area under the curve of plasma drug level versus time (to infinity); Cmax = peak plasma drug level; CrCl = creatinine clearance; PPI = proton pump inhibitor.

[0182] Modification factors for total body clearance (Cl), central compartment volume (Vc), and drug relative bioavailability (F), based on individual patient characteristics, were as follows:

number

number

number

[0183] According to one or more embodiments of the present disclosure, a two-compartment plasma level prediction model may be defined by Equations 1-4, with expressions for Cl, Vc, and F adjusted according to Equations 5-7, and parameter estimates taken from Liesenfeld.

[0184] Figures 4A and 4B show a comparison and validation of the disclosed model with a model based on Liesenfeld's work. Figure 4A shows a comparison of dabigatran plasma levels on day 10 for the Liesenfeld method 420 and the plasma level prediction model 418 described above according to an embodiment of the present disclosure. Input parameters included Ka = 0.754, Vc = 673, Vp = 345, Q = 35.5, gender = male, Wt = 80.3 kg, ETHN = Caucasian, no concomitant medications, and the specified dosage and measured CrCl. It can be seen that the two curves 418, 420 of the separate models are in good agreement. F was revised to a value of 0.75 to provide the best overlap. Figure 4B shows a comparison of the relationship between renal function (CrCl) and plasma levels determined by the Liesenfeld model 424 and the plasma level prediction model 422 described above for a fixed dose of 150 mg. Again, the curves 422, 424 from the two models are in good agreement and show that relatively small changes in CrCl at the lower end of renal function (i.e., <75 ml / min) result in relatively large fluctuations in circulating drug plasma levels.

[0185] As can be seen from the model equations above, as well as Figures 4A and 4B, renal function is currently the primary parameter identified by Liesenfeld as affecting plasma levels and used by regulatory agencies to determine starting doses, but there are several other covariates (patient input data) that can be considered when calculating the trough plasma levels that can be achieved for a particular patient. This is especially true when considering single or multiple extreme cases, such as a subject who is female, South Asian, well over 80 years old, taking a PgP inhibitor, and has poor renal function (where levels may be three-fold higher than in patients with "normal average" values). Such patients may require a starting dose lower than the currently available 75 mg.

[0186] As discussed below, providing an ML dosage calculator based on a PKPD model can enable: (i) providing accurate and precise dosage, (ii) providing individualized dosage, (iii) utilizing a wide range of patient data, including data not previously considered in pharmacometric modeling, (iv) providing dosage monitoring and adjustment, (v) defining new dosing regimens, and (vi) providing thrombosis and bleeding risk as a function of time since last dose. Note that such benefits are applicable to all drugs, not just the specific example of dabigatran.

[0187] Embodiments of the present disclosure may include ML dosage calculators trained using a two-compartment model based on Equations 1-4 and modified by other expressions for F, Vc, and / or Cl derived from other patient studies of dabigatran, including studies conducted using the devices and methods disclosed herein. Other embodiments may include ML dosage calculators trained using a simpler single-compartment model that does not include the peripheral compartment 303. However, a two-compartment model may provide more accurate results than a single-compartment model. Still further embodiments include ML dosage calculators trained using single- or two-compartment (e.g., Equations 1-4)-based models for drugs other than dabigatran.

[0188] Dosage Calculator The plasma level prediction model can predict plasma levels throughout the day, including peak or maximum plasma levels (Cmax) and trough levels (Ctrough), based on individual patient data. Returning to the method of Figure 2, the present method advantageously implements an ML dosage calculator trained using actual population data and / or simulated population data from the plasma level prediction model, allowing for calculation of the starting dose (or dose adjustment) using pharmacokinetic principles.

[0189] In some examples, the dosage calculator may receive a target plasma level metric and calculate a dosage to administer to a patient by processing the patient data and the target plasma level metric. The target plasma level metric may include an ideal therapeutic level (ITL) for trough plasma levels (e.g., 75-150 ng / ml or a more targeted level such as 112.5 ng / ml). The ITL is intended to provide an optimal balance between reduced risk of thrombosis and increased risk of bleeding. The target plasma level metric may be related to other plasma metrics, such as maximum concentration, and may be individualized based on patient data, as described in the "Individualized Target PKPD Metrics" section below.

[0190] Patient Data As described above, the patient data may include any variables used as inputs to the PKPD model used to train the ML dosage calculator. The patient data may include one or more variables that can affect drug absorption and / or clearance of the drug from the body, and may include, among others, one or more of renal function metrics, liver function metrics, patient age, patient ethnicity, patient sex, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, a pharmacogenomic profile (e.g., a CYP enzyme profile and / or a drug transporter profile), and the patient's medication list.

[0191] For dabigatran, current regulatory dosing guidelines consider only a limited number of patient parameters. In the EU, starting doses are based on body weight, age, treatment modality (prophylactic or reactive), and concomitant medications. For children, dosing guidelines can be particularly complex. In the United States, dosing guidelines are based on individual renal function, as the majority of the drug is eliminated through the kidneys, but do not take into account patient demographics or the large intersubject variability in drug absorption. However, dosing flexibility is limited because only two doses are available.

[0192] The plasma level prediction model described above indicates that renal function metrics, specifically creatinine clearance (CrCl), are the largest contributors to variability in drug plasma levels among patient populations. Age, weight, hemoglobin, gender, ethnicity, heart failure, and the presence of other medications (specifically, proton pump inhibitors, amiodarone, and verapamil) can also all modulate circulating drug levels in plasma.

[0193] Returning to the first step 212 of FIG. 2 , in the simplest embodiment, the patient data includes renal function metrics. In such embodiments, mean population values ​​or surrogate substituted values ​​may be used for other parameters involved in the dose calculator and plasma level prediction model described above. In some embodiments, the patient data may include one or more of the other patient parameters defined in Equations 5-7, namely, age, weight, hemoglobin, gender, ethnicity, presence of heart failure, and other medications. In some examples, the patient data may include the most readily accessible data, such as age, weight, gender, ethnicity, history of heart failure, and concomitant medications. In some examples, parameters requiring specific medical testing, such as hemoglobin level or CrCl, may be estimated using surrogate substitution, as described below.

[0194] Patient weight can present a particular problem for current guidelines, as the international guidelines for dabigatran only apply up to a maximum of 140 kilograms. There is also limited evidence for patients with a body mass index (BMI) greater than 40. As more and more patients exceed this weight limit, and without knowledge of what to do for them regarding DOACs, many patients end up on warfarin, resulting in increased risk. By including weight as a patient data parameter, dosage calculators can favorably account for overweight patients.

[0195] Furthermore, changes in the endothelium due to disease or simply aging may affect the absorption of dabigatran. By including age as a patient data parameter, the dosage calculator can advantageously account for age dependency on plasma drug levels.

[0196] The plasma level prediction models and calculators described above may require values ​​for age, sex, CrCl, hemoglobin, weight, ethnicity, heart failure, and concomitant medications. While the use of specific interfering medications is likely known, along with weight, age, sex, and the possible presence or absence of heart failure, in some situations, CrCl and / or hemoglobin levels may not be known without prior testing.

[0197] In some examples, CrCl or hemoglobin values ​​can be provided by population-based surrogate estimates. For example, an estimated creatinine clearance rate (eCCR) can be provided by the Cockcroft-Gault formula. Creatinine clearance can be estimated using serum creatinine levels. The Cockcroft-Gault (CG) formula uses a patient's weight (kg) and gender to predict CrCl (mg / dL). If the patient is female, the resulting CrCl is multiplied by 0.85 to correct for lower CrCl in women. The CG formula relies on age as the primary predictor of CrCl. The CG formula can be written as follows:

number

[0198] Serum creatinine can be measured, derived from age-related changes, and potentially further revised depending on other diseases or conditions. In the absence of a direct measurement of CrCl, eCCR can be substituted for CrCl in the dosage calculations described herein.

[0199] In some examples, similar substitutions can be made for hemoglobin levels. Hemoglobin values ​​in men aged 20-60 are typically 14.5-15 gm per 100 ml, with younger men tending to have higher values. After the age of 50, there is a gradual and significant decrease, averaging 12.4 gm per 100 ml for men aged 80-90. For women after the age of 20, the average hemoglobin value remains near 13 gm per 100 ml (Hawkins [3]). When hemoglobin is not directly measured, these relationships can be used to define a surrogate estimate of hemoglobin in the dosage calculations described herein.

[0200] Other patient data parameters The present disclosure may encompass other PKPD models and resulting trained ML dosage calculators that are currently available or may become available in the future after appropriate patient studies.

[0201] Thus, according to embodiments of the present disclosure, dabigatran patient data may include other patient parameters that may be drivers of endothelial function, including one or more of smoking history, alcohol intake, patient's history of thrombosis, patient's history of bleeding, patient's history of cancer, patient's cardiovascular history, patient's metabolic history, patient's platelet count, patient's genetic determinants, patient's hematocrit, patient's liver function, patient's blood pressure, patient's comorbidities, patient's activity level, patient's dosage compliance, family history of thrombosis, family history of stroke, family history of bleeding (e.g., anemia), and family history of hypertension.

[0202] Patient data may also include medications other than those specifically mentioned above, including calcium channel blockers, nonsteroidal anti-inflammatory drugs, H2 receptor antagonists, clopidogrel, aspirin, and diltiazem. For any drug, patient data may also include cytochrome P450 data. Cytochrome P450 is a set of enzymes found throughout the body that metabolize drugs, primarily in the liver. In some examples, administered drugs can decrease or increase the activity of these enzymes, resulting in higher or lower plasma levels of other co-administered drugs. In the case of dabigatran, other drugs may affect dabigatran blood levels by altering the way dabigatran is metabolized. Therefore, models and calculators utilizing cytochrome P450 data can account for drug interactions.

[0203] The patient's bleeding history may include bleeding events indicating the severity and duration of the events. The patient's bleeding history may be related to endothelial status or bleeding tendency, such as the tendency for superficial bruising on the arms and legs, and other measures of skin elasticity. Other methods for assessing vascular data and / or endothelial status include image analysis of retinal vessels from fundus photography. The flexibility of the model facilitates the incorporation of alternative measures of endothelial status, which may be incorporated into future plasma level prediction models, for example, by evaluation and training using machine learning capabilities.

[0204] The patient's thrombotic history can include thrombotic events indicating the severity and time of the event. A family history of thrombosis can include first-degree relatives and can include those under the age of 50.

[0205] The patient's history of risk events may include a record of intermittent atrial fibrillation events, including their duration, occurrence, and frequency. Such a record may be provided from cardiac data from the smartwatch.

[0206] In relation to cancer history, cancer poses a particularly high risk of thrombosis due to highly activated endothelium, which subsequently releases various clotting factors. Even with DOAC treatment, thrombosis can commonly occur in the cancer setting, with a high recurrence rate of approximately 17-18% and a 20% risk of major bleeding. This is particularly true for certain cancers, such as those of the lung, pancreas, and colon, further requiring patient monitoring and precise dosing. Overall, it is estimated that only one-third of patients with thrombosis are adequately managed, even in the best medical centers, and in other settings where management is by general practitioners or pulmonologists, optimal management occurs in less than 10% of cases. As described below, embodiments of the present disclosure can include personalized target plasma level metrics for cancer patients to account for their unique PK-PD. Also, as described below, embodiments of the present disclosure can include enhanced monitoring to calibrate and / or personalize plasma level prediction models and dosage calculators.

[0207] A patient's genetic determinants may replace, refine, or improve (for example) a model or computer classification of ethnicity.

[0208] A patient's activity level may represent an indication of personal risk exposure, for example, that the patient is engaging in risky activities or other lifestyle factors such as recreational drugs, physical activity, etc.

[0209] Patient comorbidities may include an indication of whether the patient has AF and the severity of the AF. If the severity of AF is particularly severe (associated with episodic dizziness, syncope, and tachycardia), a higher dose may be justified.

[0210] If the patient has a family history of stroke or high blood pressure, higher doses may be better tolerated, even if accompanied by some bleeding.

[0211] Patient's dosage compliance can indicate the tendency of the patient to take the drug according to the prescribed regimen.The plasma level prediction model and dosage calculator can be refined (for example, through future patient testing) to take into account patient's compliance.The method can include monitoring patient's compliance and adjusting dosage accordingly.

[0212] The patient's metabolic history may suggest diabetes, which, as a comorbidity, may alter clotting and circulation.

[0213] One or more of the above patient parameters may be incorporated into the PKPD model and / or ML dose calculator via appropriate patient testing. Patient data including one or more (new) patient parameters may be recorded along with patient dosage (amount and time) and directly measured PKPD metrics to alter the PKPD model and / or ML dose calculator accordingly to account for the additional dependency of the one or more new patient parameters.

[0214] The patient data may also include additional patient data parameters. For example, the patient data parameters may include parameters that can affect individualized target PKPD metrics, as described in the relevant sections below. The additional patient data parameters may include side effect reporting (see the "Side Effects Monitoring" section below), the time the dose was taken, and other examples described herein.

[0215] Receiving patient data The patient data may be received by one or more of manual data entry (either by the patient, HCP, or third party on a computing device such as a personal computing device), data from medical records stored in a database or the like, and receiving physiological measurements (e.g., from a medical device or clinical database). For example, the patient data may include cardiac data from a smartwatch that can indicate periods of AF.

[0216] implementation The method of FIG. 2 may be implemented as a versatile deployment software-based ML dosage calculator configured to present dosages in short time frames (sub-seconds, seconds, or minutes) to healthcare professionals and / or patients via personal computing devices or the like.

[0217] Figure 10 illustrates a system suitable for performing any of the computer-implemented methods described herein, including a versatile deployment software implementation (also referred to herein as a digital app) and the method of Figure 2. The system 1052 includes a patient device 1054. The patient device 1054 may include one or more processors configured to receive patient data and process the patient data with an ML dosage calculator to determine a dosage of a drug to administer to the patient. The dosage calculator is derived from a PKPD model and indicates the dosage.

[0218] In the illustrated embodiment, the patient device 1054 is a smartphone. However, the invention is not limited in this respect and the patient device 1054 can take many other forms, including, but not limited to, a mobile phone, a tablet computer, a desktop computer, a voice-activated computing system, a laptop, a gaming system, a vehicle computing system, a wearable device, a smart watch, a smart television, an Internet of Things device, and a medication dispensing device.

[0219] The patient device 1054 may be configured to collect one or more parameters of patient data. The patient data may be obtained via manual data entry using a human interface device of the patient device 1054 and / or from a remote source via the network 1058.

[0220] The patient device 1054 may include memory (not shown) for storing patient data and / or the output of the ML dosage calculator. Such data may also be stored in a database 1062 as a networked or cloud-based data store.

[0221] The patient device 1054 may have one or more applications (or apps) installed on a storage medium associated with the patient device (not shown). The one or more apps may be configured to perform any of the computer-implemented methods disclosed herein. The one or more applications may be configured to assist the patient in providing patient data and / or may include a dosage calculator for processing patient data. The one or more applications may be downloaded from a network, for example, from a website or an online application store.

[0222] In this example, the system 1054 further includes a data processing device 1060 that is communicatively connected to the patient device 1054 via a network 1058. In the illustrated embodiment, the network 1058 is the Internet, although the invention is not limited in this respect and the network 1058 can be any network that enables communication between the patient device 1054 and the data processing device 1060, such as a cellular network or a combination of the Internet and a cellular network.

[0223] The data processing device 1060 complements the patient device 1054 and may perform one or more steps of any of the computer-implemented methods disclosed herein. For example, in some embodiments, the patient device 1054 may receive patient data and provide it to the data processing device 1060. The data processing device 1060 may then process the patient data with an ML dosage calculator to determine the dosage to administer to the patient. The data processing device 1060 may then provide the dosage to the patient device 1054 or the HCP device 1064 (which may also be referred to as the clinician device 1064) (either or both may indicate the dosage). In this manner, the data processing device 1060 provides networked, server-based, or cloud-based processing capabilities to the system for performing the computer-implemented methods.

[0224] The data processing device 1060 may be connected to a database 1062 that may store patient data and / or the output of the dosage calculator.

[0225] In this example, the system 1052 includes a clinician data processing device 1064 that is communicatively coupled to the patient device 1054 and the data processing device 1060 via a network 1058. The clinician data processing device 1064 may be generally similar to the patient device 1054 and provide a similar set of functions. Specifically, the clinician data processing device 1064 allows for collating or receiving patient data. The clinician data processing device 1064 is considered to be physically located within the HCP's premises during its use, such as a clinic, doctor's surgery, pharmacy, or any other medical institution, e.g., a hospital. The clinician data processing device 1064 may include one or more sensors (e.g., a blood pressure sensor) capable of collecting information about the patient and / or may be configured to control one or more separate sensors.

[0226] It is also contemplated that the clinician data processing device 1064 will typically be used by a medically trained person with appropriate data security access rights and may therefore have more advanced functionality available than via the patient device 1054. For example, the clinician data processing device 1064 may be able to access a patient's medical history, generate drug prescriptions for the patient, order medications, etc. Access to functionality may be controlled by security policies implemented by the local processor or data processing device 1060.

[0227] The data processing device 1060 and / or the clinician device 1064 may have applications installed that are compatible with or the same as the applications installed on the patient device 1054 .

[0228] It will be understood that the various steps of the computer-implemented methods disclosed herein may be performed by any of the one or more processors of the patient device 1054, the data processing device 1060, and the clinician device 1064, in any combination. For example, all steps may be performed by the clinician device 1064, and optionally the data processing device 1060, which locally receives one or more parameters of patient data from the patient device 1054 or another remote device via the network 1058. In a further example, all steps may be performed in a networked backend on the data processing device 1060, with the patient device 1054 and the clinician device 1064 acting as a human interface for collecting and presenting data. In yet another example, all steps may be performed on the patient device 1054, with the clinician device 1064 simply collecting relevant data from the patient device 1054 to inform or instruct the HCP.

[0229] It will also be understood that one or more of the components of the system 1052 may be omitted depending on the application. For example, in a clinic setting, the disclosed computer-implemented method may run only on the clinician device 1064. Alternatively, the method may run only on the patient device 1054 in a home setting.

[0230] The dosage calculator may be deployed as a digital app on any of the patient device 1054, the data processing device 1060, and the clinician device 1064. The digital app may indicate to the HCP the dose (starting dose, continuing or revised dose, timing of administration) to administer to a particular patient based on the calculated change in drug levels to achieve target PKPD metrics.

[0231] In the specific example of dabigatran, the digital app may summarize key factors (among other patient data) that influence thrombosis and bleeding and present them to the HCP. These factors may include history of thrombosis and bleeding, cancer history, fall patterns, renal and liver function patterns, platelet count patterns, hematocrit, platelets and transfusions, and patient age. The app may also recommend follow-up patterns and prompt for missing information.

[0232] output Dosage output With respect to the second and third steps of FIG. 2, which relate to determining and prescribing dosages to administer to a patient, determining and prescribing dosages can be performed in several ways.

[0233] The prescribed dosages may include starting dosages for patients beginning drug therapy, continuing dosages to ensure that patients taking the drug are receiving the appropriate level, or revised dosages that suggest that the patient's dosage should be adjusted in light of updated patient data (discussed further in the "Continuing Therapy Management" section below).

[0234] When dosage is used, it refers to both the amount administered, the form of the drug (eg, immediate release or controlled release), the time taken in relation to meals and other events, and the periodicity.

[0235] In some examples, determining the dosage may include (i) processing patient data with an ML dosage calculator to determine an ideal dosing regimen, and (ii) selecting a dosage to administer to the patient from a selection of available dosing regimens based on the ideal dosing regimen. In the specific example of dabigatran, the second step of the method of FIG. 2 may include selecting a dosage from available dosages, such as 75 mg, 110 mg, or 150 mg. For some patients, the method may determine and prescribe a higher dosage in the evening and a lower dosage in the morning so that Cmax occurs at a lower risk time (e.g., in patients who fall frequently). The method calculates an optimal total dosage to consistently achieve sufficient blood clotting prevention.

[0236] As described below, the present disclosure encompasses novel dosage forms that allow greater flexibility in selecting a dosage amount. Selecting a dosage amount can include selecting a dosage amount in increments of 1 mg, 5 mg, 10 mg, or 25 mg. Selecting a dosing regimen can also include selecting a dosage form, such as a liquid formulation or a microgranular formulation, as described below.

[0237] Determining and indicating a dosing regimen may include determining and indicating one or more of the dosage amount, dosing time, dosing frequency, and / or dosing form. In the specific example of dabigatran, the dosing frequency may include twice daily, similar to the current standard for DOACs. However, the present disclosure also encompasses other dosing frequencies, including once daily or less than once daily, which can improve patient compliance, and multiple times daily, which can reduce maximum plasma levels (Cmax). The method may determine and indicate that a particular patient should take the drug only every other day (e.g., if creatinine clearance is very low) or three times daily (e.g., if creatinine clearance is very high). Alternatively, for other patients, the method may determine and indicate whether twice-daily dosing results in an unacceptably high Cmax to achieve the desired Cmax, and therefore whether a sustained-release formulation or a lower-concentration dosage form should be administered to increase the dosing frequency.

[0238] The dosing time may include the time at which the patient takes the medication. The dosing time may be provided as a simple reminder to the patient to take the medication. The dosing time may also be specific to an event, such as an upcoming or recent surgery. The dosage form may relate to different dosage forms, such as a solid dose, a liquid dose, or a microparticulate formulation.

[0239] New dosage form The disclosed ML dosage calculators, digital apps, and related methods can encompass dosages, formulations, and regimens other than those currently approved.

[0240] As mentioned above, immediate-release doses of dabigatran are available in the EU only in three doses: 75, 110, and 150 mg. Notably, there is no explanation for this uneven distribution, despite the linear pharmacokinetics. Furthermore, recommended dosages are complex depending on the patient's weight and the child's age. Adolescents under 18 years of age weighing 30 kg require a 150 mg (bid) dose, while a person of the same age but weighing 60 kg requires a 260 mg (bid) dose consisting of a 150 mg tablet (bid) and a 110 mg tablet (bid). In the United States, only two doses, 75 mg and 150 mg, are approved, further complicating dosage titration.

[0241] Therefore, the need for other doses, such as 25, 50, 100, 200, and 250 mg, in addition to the already available doses is clear. These can be provided to the appropriate patient using the app's algorithm. Figure 11 shows the estimated plasma level time profile 1166 on day 10 for a simulated patient (85 years old, CrCl = 10 ml / min, male, heart failure, Caucasian, hemoglobin = 14.3, weight = 70 kg, no concomitant medications) taking 25 mg of dabigatran twice daily. The Liesenfeld test plasma level time profile 1168 and mean plasma level time profile 1170 on day 0 are also shown for reference. The figure shows an example in which the currently unavailable dose (25 mg) achieves an ITL of 75–150 mcg / L over a 12-hour window until the next dose in an 85-year-old individual with impaired renal clearance. Current FDA approval does not include prescribing dabigatran in patients with a CrCl less than 15 mL min-1. This figure illustrates an example where a dose other than the currently available dose may be most appropriate for a given patient. Other examples include when a lower dose may be more appropriate before or after surgery, or when the patient exhibits all indicators of prolonged drug retention in the body, or when the individual's risk of thrombosis changes over time.

[0242] All available doses are immediate-release, producing peak levels approximately 1 hour after administration without food and 3–4 hours after a meal. Because the average half-life is only 12 hours, a second dose in the evening is required to achieve stable plasma levels. Multiple doses can reduce patient compliance. Therefore, the development of sustained-release formulations is necessary. A sustained-release formulation would not only allow for once-daily dosing but would also have the advantage of reducing the maximum drug plasma level (Cmax) associated with bleeding episodes.

[0243] The disclosed method and calculator may also determine a dosing regimen that increases dosing frequency to achieve optimal trough levels without exceeding a Cmax threshold. The inventors recognize that such functionality is desirable because bleeding risk is more related to overall drug exposure and therefore to Cmax, as opposed to thrombosis risk, which is related to coverage, AUC, or blood trough levels. Overall, it may be desirable to minimize the Cmax / trough ratio and avoid Cmax exceeding a certain level. Calculating such a profile for a specific patient has not previously been practical, but the methods and devices disclosed herein make it possible.

[0244] Dabigatran has a half-life of approximately 12 to 14 hours, and therefore, twice-daily administration is recommended. However, controlled-release formulations can be developed for such purposes (e.g., to improve compliance and reduce multiple peaks and troughs in plasma drug levels) using microparticle formulations (or any other known formulations for controlling solubility and half-life) that require only a single daily dose. Each microparticle may exhibit a controlled-release profile, and the development of multiple doses may be facilitated by varying the number of microparticles in a capsule without reformulating and testing each individual dose. Reformulation is also expected to address the hygroscopicity issue that prevents approved dabigatran formulations from being placed in dosage aid boxes, etc., for patients who require such support devices.

[0245] More generally, for any drug, the disclosed PKPD models, ML calculators, apps, and methods can potentially accommodate any dosage form changes to refine and validate the PKPD model and / or ML dosage calculator following clinical trials. The disclosed PKPD models, dosage calculators, digital apps, and related methods can account for other dosages, such as those described above. Indeed, the disclosed systems and methods can perform better when more precise dosages are available.

[0246] PKPD Metrics Output Generally, for any drug, embodiments of the dosage calculation method, particularly embodiments of the digital app implementation, may also process patient data to determine and optionally output one or more PKPD metrics. The dosage calculation method may include determining one or more of the following: maximum drug concentration (Cmax), trough drug concentration, ratio of maximum drug concentration to trough drug concentration, drug concentration time profile, area under the curve of the time profile, ratio of maximum drug concentration to area under the curve of the time profile, and average daily drug concentration at steady state, coverage. The dosage calculation method may include outputting one or more of these PKPD metrics to the patient or HCP.

[0247] As further described below (in the "Individualized Target PKPD Metrics" section), the maximum drug concentration, trough drug concentration, and / or the ratio between them can be used to adjust the target PKPD metrics to ensure an optimal balance between the risks resulting from overdosing (e.g., side effects) and underdosing (e.g., no therapeutic effect). In the specific example of dabigatran, these risks manifest as risk of bleeding and risk of thrombosis, respectively. Outputting these PKPD metrics to the HCP allows the HCP to manually adjust the target PKPD metrics.

[0248] In some embodiments, the dosage calculation method can include outputting a drug concentration time profile or a risk level derived from the drug concentration time profile. The drug concentration level time profile can include data similar to that depicted in FIG. 4A, representing fluctuations in drug plasma levels over time. In such an example, the patient data can include a dosing time at which the patient received a dose of the drug. The dosing time can then be used as the start time of the drug concentration time profile.

[0249] In the specific example of dabigatran, the digital app can use the drug concentration time profile, including the plasma level time profile, to determine and indicate the patient's thrombosis risk and / or the patient's hemorrhage / bleeding risk according to the plasma level time profile and the dosing time. The digital app can convert circulating drug plasma levels into risk based on a predetermined plasma level-risk relationship, such as that shown in Figure 1.

[0250] Indicating bleeding risk (e.g., via a digital app) may include indicating to the patient times when the patient's risk of bleeding is increasing or decreasing, so they can adapt their behavior accordingly. In some examples, the HCP portion of the app may output bleeding risk, plasma level time profile, and / or circulating drug plasma levels (derived from the time profile) to the HCP before an invasive procedure. Such information provided to the HCP can guide the timing of surgery, the period for which medication should be withheld before a particular invasive procedure, such as surgery, and the time until the procedure can be performed. In some examples, the digital app may indicate the amount of time the patient or HCP should wait before performing an invasive procedure (or changing medication to a different anticoagulant). The output plasma level metric can be used to determine whether dabigatran plasma levels are suitable for surgery, for example, if a cutoff of approximately 50 ng / ml is considered acceptable, to aid clinical decision-making. It is recognized that different surgeries carry different risks, and thresholds may be adapted accordingly.

[0251] In some examples, the patient-facing portion of the digital app may indicate the risk of thrombosis. This may include advising short-term thromboprophylaxis after the patient has completed regular anticoagulation therapy for the original thrombus, for events that pose a high risk of further thrombosis, such as surgery or a flight longer than four hours. The app may also advise when to switch from prophylactic anticoagulation to therapeutic anticoagulation after suspending anticoagulation therapy for a planned procedure. This latter aspect may be presented in either the patient-facing app and / or the clinician-facing app.

[0252] In some examples, the digital app may receive patient data from other objective monitoring systems. For example, the digital app may receive activity data from a wearable device to monitor activity levels and estimate associated risk increases or decreases. Increased activity reduces stroke risk, but not beyond a level at which the risk of bleeding becomes unacceptably high. The digital app may also receive blood pressure data from a blood pressure monitoring device. Blood pressure is associated with the probability of stroke, including hemorrhagic stroke. The digital app may adjust or correct bleeding or thrombosis risk based on blood pressure and / or activity data. Incorporating home blood pressure monitoring data into patient data and providing it to the HCP may also help the HCP determine whether additional treatment is needed (which is not routinely done for patients receiving DOACs).

[0253] Individualized target PKPD metrics In addition to those already mentioned above, blood clotting tendency can be affected by various patient parameters that change over time and between individuals, such as hydration status, infection, inflammation, menstruation, blood pressure, and blood flow rate through blood vessels. Stagnation of blood flow can be seen during prolonged immobility, especially after surgery. Medical procedures or surgeries involving trauma or needle punctures can increase the risk of bleeding. As already mentioned, blood vessel walls can weaken as a result of aging, amyloid deposition, particularly in the brain, or inflammatory processes or medications, such as steroids or antidepressants. While initiation of thrombus formation can occur within the normal physiological range, the anticoagulant properties of the blood clotting cascade prevent prolongation.

[0254] The inventors have recognized that the process of thrombus formation or thrombosis occurs over a longer time course than the process of bleeding. The balance between thrombosis and bleeding tendencies will vary between individuals based on their specific physical condition. This tendency can be altered by the dosing of DOACs. The longer-term process of thrombosis is more closely related to total exposure to DOACs, while bleeding tendency is more closely related to maximum drug concentrations. Therefore, considering both these parameters and their ratios provides a path to a more personalized approach to achieving the best risk-benefit profile for anticoagulant therapy. For example, one approach may be to minimize the ratio of maximum plasma level (Cmax) to the area under the curve of the plasma level-time profile. As mentioned above, current formulations can only suppress Cmax by prescribing multiple doses per day, which poses inherent compliance challenges. A once-daily controlled-release solution is desirable but not yet available.

[0255] Upsetting the risk balance and exacerbating inappropriate anticoagulant dosing is the asymmetry in prescribing tendencies among HCPs, which reflects a psychological need to avoid responsibility. HCPs may perceive the occurrence of thrombosis, which can subsequently lead to embolism, as less responsible than the occurrence of bleeding (inaction vs. inaction). As a result, HCPs may tend toward underdosing to avoid the risk of potential overdosing, but at the population level, this leads to poorer outcomes. This reflects a psychological deficit rather than rational prescribing.

[0256] A specific example embodiment of the dosage calculation method for dabigatran can account for individual blood clotting and bleeding risk factors by determining a target plasma level metric as an individualized plasma level metric. In some examples, the target plasma level metric can include a trough plasma corresponding to the ITL. The individualized plasma level metric can include an adjustment to the ITL. As described below, the individualized plasma level metric can also include the ratio of the maximum plasma level (Cmax) to the area under the curve of the plasma level time profile that is below the bleeding risk threshold.

[0257] In embodiments employing an individualized target plasma level metric, the patient data may include one or more goal-dependent patient parameters. The goal-dependent patient parameters may include one or more of reported side effects (see "Monitoring Side Effects" below), the patient's thrombosis history, the patient's bleeding history, the patient's cancer history, the patient's stroke history, the patient's liver function metrics, the patient's cardiac function metrics, the patient's brain status, the patient's smoking history, the patient's alcohol history, the patient's blood pressure, the patient's mobility status, the patient's menstrual status, the patient's inflammatory status, the patient's infection status, the patient's concomitant medications, blood coagulation metrics, and the patient's hydration status. The dosage calculation method may determine the individualized plasma level metric based on one or more of the target-dependent patient parameters. The target-dependent patient parameters may be monitored on an ongoing or periodic basis, and the individualized target plasma level metric may be refined on an ongoing basis as part of ongoing treatment management, as described below.

[0258] In some examples, the method may determine individualized adjustments to the target plasma level metric based on specific calculations related to one or more specific goal-dependent patient parameters, particularly those indicative of a time-limited risk due to a risk event (fall, surgery, etc.). Other goal-dependent patient parameters that confer risk of bleeding or hemorrhage may be chronic or lifestyle-dependent in nature, such as inherited blood clotting disorders, alcoholism, etc.

[0259] In some examples, the method may output personalized adjustments to the patient or HCP. For example, the method may indicate a personalized ideal therapeutic trough level for the individual.

[0260] Risk Event In some examples, a dosage calculation method can include determining an individualized plasma level metric based on a bleeding risk event and / or time since a hemorrhage risk event. For example, the method can include applying a maximum adjustment to a nominal target plasma level metric immediately after the risk event and periodically tapering the adjustment toward the nominal target plasma level metric as time since the event increases. The method can suggest a new dosing regimen for each variation in the adjustment.

[0261] For example, immediately after surgery, patients may be at high risk for bleeding for the first 48 hours, after which the risk may rapidly decrease. During this period, anticoagulants may be avoided, except for patients at high risk for thrombosis. After this initial period, patients may be at highest risk for thrombosis due to inflammation and decreased mobility associated with healing. At this stage, the method may determine an individualized target plasma level metric as the upper limit of the ITL or a plasma trough level above the ITL (e.g., 150 ng / ml). The dosage calculation method may include determining and indicating a dosage to administer to the patient, which may provide the individualized target plasma level metric. After a fixed period (e.g., several days) after surgery, the dosage calculation method may include reducing the individualized target trough level to 120-130 ng / ml and, accordingly, determining and outputting a dosage to administer. This may continue until the individualized target level metric decreases to the ITL.

[0262] Other thrombotic risk events include thrombosis, one or more episodes of atrial fibrillation, dehydration, chemotherapy, trauma, surgical procedures (especially large joint orthopedic procedures such as hip replacement), and indwelling central venous catheters (e.g., Hickman lines).Hemorrhagic risk events include hemorrhagic stroke, trauma, surgical procedures, needlestick injury, or patient fall.

[0263] Some events pose both bleeding and thrombosis risks (e.g., falls and surgery). In such examples, the method may include determining an individualized plasma level metric as the ratio of Cmax to the area under the curve of the plasma level time profile, which is below the bleeding risk threshold. This may be combined with maintaining plasma trough levels within the ITL. In such examples, the method may include determining and recommending a dosing regimen that combines a controlled-release formulation or a dosing frequency of more than twice per day (e.g., 8 times per hour or 6 times per hour) with a lower dosage (e.g., 75 mg or less, if a suitable dosage form is available). Although a higher dosing frequency may reduce compliance, it can be appropriately managed in carefully managed patients (e.g., cancer patients) and care settings (e.g., post-surgery or nursing homes). As discussed above, cancer patients may be at risk for both thrombosis and bleeding. Therefore, the method may also include determining a similar individualized plasma level metric (ratio and / or trough level) and dosing regimen for cancer patients. Additionally, some patients have chronic bleeding or thrombosis risk and may experience thrombotic or bleeding risk events. The present method can also account for such dual risk by individualizing rates and drug dosing regimens in a similar manner.

[0264] Chronic or lifestyle-related risk parameters Patient falls may also provide an indicator of risk for further falls. Determining whether the bleeding risk from falls outweighs the benefits of anticoagulation therapy presents challenges. A rough guideline is that nine or more falls per year may be a contraindication to anticoagulation therapy, but this varies depending on the nature of the falls. Dosage calculation methods may decrease the target trough plasma level or the ratio of Cmax to the area under the curve of the time profile based on the frequency and / or timing of falls, as indicated by the patient's fall history. Data from devices measuring fall propensity may be input into the system. A patient app may provide customized advice regarding when risks from falls are highest and lowest, or when those risks are related to other activities, mitigation measures, and medication timing.

[0265] The risk of bleeding is partly age-related. The reasons are not fully understood, but may be related to age-related changes in the endothelium, particularly changes in collagen. Dosage calculation methods may reduce the target trough plasma level or the ratio of Cmax to the area under the curve of the time profile based on the patient's age.

[0266] Further issues for women relate to menstruation and contraception. Estrogen increases the rate of thrombosis, so estrogen-based pills are usually discontinued (incidentally, the menstrual cycle can also affect INR readings, making warfarin a difficult medication for menstruating patients). Patients then have heavy menstrual bleeding, become anemic, and may then receive blood transfusions. Some accidentally become pregnant. For the majority of people, optimal management is overlooked. Most people do not know if their menstruation is abnormal. Data on menstrual patterns could be entered into the system, and a patient app could provide customized advice on contraception and menstruation (e.g., advice on appropriate treatment for heavy menstrual bleeding). The app could also notify HCPs of problems.

[0267] Other target-dependent patient parameters that may increase bleeding risk and therefore require a reduction in the ratio of Cmax to area under the curve of the target trough plasma level and / or time profile include high blood pressure (risk of hemorrhagic stroke), poor liver function, presence of cerebral amyloid, poor renal function, high alcohol consumption, and use of aspirin, clopidogrel, NSAIDs, steroids, antidepressants, or any other medication that may increase bleeding risk. In some examples, the dosage calculation method may determine a bleeding risk score based on the presence, timing, and / or severity of bleeding risk events and one or more of the aforementioned bleeding risk factors. The dosage calculation method may use the HASBLED score to determine the bleeding risk score.

[0268] Alcohol consumption may also increase AF and dysthymia, which may increase the likelihood of stroke and thrombosis.

[0269] The dosage calculation method may determine an individualized target plasma level metric by decreasing the target trough plasma level if the bleeding risk score is above a bleeding risk threshold.

[0270] Other goal-dependent patient parameters that may increase thrombosis risk and therefore require an increase in target trough plasma levels include genetic profile, active cancer or inflammatory state, and the patient's hydration status. In some examples, the dosage calculation method may determine a thrombosis risk score based on the presence, timing, and / or severity of a thrombosis risk event and one or more of the aforementioned thrombosis risk parameters.

[0271] The dosage calculation method may determine an individualized target plasma level metric by increasing the target trough plasma level if the thrombosis risk score exceeds a thrombosis risk threshold.

[0272] In some examples, the dosage calculation method may receive a bleeding risk score and / or a thrombosis risk score via data input from an HCP.

[0273] The dosage calculation method may determine an individualized target plasma level metric as the ratio of Cmax to the area under the curve of the plasma level time profile that is below the bleeding risk threshold when the thrombosis risk event exceeds the thrombosis risk threshold and the bleeding risk score exceeds the bleeding risk threshold.

[0274] In some examples, the method may include assigning a patient to one of a plurality of subgroups. Each subgroup may correspond to one or more of the above-described goal-dependent patient parameters. The method may include assigning a patient based on time since a risk event and / or the presence of specific patient risk factors. Each subgroup may have a corresponding individualized target plasma level metric.

[0275] More generally, for any drug, the digital app and associated methods may determine a target PKPD metric as an individualized PKPD metric based on patient data. In some examples, the app and method may determine (i) a first risk associated with too much medication based on patient data, (ii) a second risk associated with too little medication based on patient data, and (iii) an individualized target PKPD metric based on the first risk and the second risk. For example, the patient data may indicate susceptibility and / or resistance to a particular side effect, and a first risk may be calculated accordingly. The patient data may also indicate a risk of under-prescribing the drug to the patient (e.g., thrombosis risk in the specific example of dabigatran, or hypertension risk in the example of an antihypertensive drug), and a second risk may be calculated accordingly.

[0276] Ongoing Care Management As a patient's condition changes, pharmacokinetics and pharmacodynamics may also change, potentially resulting in under- or overdosing. The digital app and associated method may receive updated patient data, including measured PKPD metrics obtained from the patient's physiological testing, process the updated patient data with an ML dosage calculator to determine an updated dosage, and indicate the updated dosage. The measured PKPD concentration may be a drug concentration. The physiological testing may include any of a blood test, a urine test, a cerebrospinal fluid analysis, a biopsy, or any other physiological testing. The measured PKPD metric / parameter may be a drug concentration from the physiological testing.

[0277] In the specific example of dabigatran, ideally, patients should be reviewed at least annually, and in high-risk patients, perhaps every 3-6 months, or more frequently in particularly high-risk patients (e.g., patients with renal failure, patients experiencing signs of bleeding during preparation for surgery, patients taking drugs with absorption interactions, or patients at particularly high risk of stroke).

[0278] A comprehensive DOAC monitoring approach may include one or more of the following items assessed at each follow-up: 1. Current health condition (e.g., measurement of blood pressure and blood samples to check for new anemia, liver and kidney function). 2. Lifestyle assessment (including nature of sport). 3. Risk of falling. 4. Individualized bleeding and thrombosis risk. 5. Compliance and Costs. 6. Bleeding / thromboembolic events. 7. Adverse events. 8. Complete drug screening for drug interactions. 9. Reassess the appropriateness of therapy. 10. Laboratory parameters requiring retesting. 11. Ongoing follow-up appointments for patients

[0279]

[0003] Embodiments of the digital app and associated methods may include providing the HCP with a summary of key patient data parameters tracked during treatment. The digital app may track any of the patient data parameters described herein for presentation to the HCP. Patient data parameters may be tracked via manual data entry from the patient, HCP, or clinic, from clinical records, or any other input method. In the specific example of dabigatran, parameters that may be tracked include, among others, age, weight, renal function, liver function, bleeding tendency, any new thrombotic events, and risk.

[0280] Embodiments of the digital app and associated methods may include tracking the aforementioned patient data parameters as updated patient data (e.g., periodically (daily, weekly, monthly)), processing the updated patient data parameters to determine an updated dosage to administer to the patient, and presenting the updated dosage. For example, a change in patient data underlying a PKPD model (such as renal function or a drug in a dabigatran plasma level model) may result in a different calculated dosage to achieve a target PKPD metric. As a further example, in the case of dabigatran, a change in patient data indicating an increased risk of bleeding and / or thrombosis may result in an individualization of the target plasma level metric (described in the previous section), thereby changing the recommended dosage to administer to the patient. Presenting the updated dosage may include presenting the updated dosage to the HCP at a periodic review meeting and / or presenting the updated dosage to the HCP or other healthcare professional (e.g., pharmacist, nurse, etc.) via an alert if the change in the updated dosage exceeds an alert threshold.

[0281] Embodiments of the digital app and associated methods may include tracking or monitoring patient data, including side effects such as excessive bleeding, GI dysfunction, vomiting, skin rashes, and patient behavior patterns (and others listed below under "Monitoring Side Effects"). Patient data may be provided to HCPs as part of a healthcare review or may be alerted in serious cases such as bleeding.

[0282] More generally, patient data tracking allows for periodic assessment of underlying changes in the patient's physiology due to time, age, disease progression, or drug or lifestyle or other environmental factors to maintain personal calibration of dosage to provide the ideal therapeutic level for that patient.

[0283] Monitoring blood coagulation tests In the specific example of dabigatran, ideally, a continuous monitoring program should include measurement of drug trough plasma levels. However, measurement of plasma drug concentrations both in hospitals and by home health care providers is rare, expensive, and infrequent.

[0284] To directly assess anticoagulant efficacy, specific quantitative measurements of dabigatran, such as plasma dabigatran drug concentration, as well as blood coagulation measurements such as antifactor Xa levels, dilute thrombin time (dTT), and ecarin time (ECT) exist. Currently, dilute prothrombin time and ecarin clotting time (ECT) are used, particularly in high-risk patients, to indicate that bleeding risk is low enough to allow surgery, with a typical threshold of 50 ng / ml blood concentration. One problem is that these measurements have not been shown to directly relate to clinical outcomes, and no consensus on standardized therapeutic ranges has been established. Furthermore, these more specific tests are not always available to all healthcare providers.

[0285] Other coagulation tests include activated partial thrombin time (aPTT), chromogenic assay (ECA), and INR test, the latter of which is usually considered inappropriate for dabigatran but is a standard test for warfarin and vitamin K antagonist (VKA) compounds.

[0286] Unlike VKA anticoagulants, these tests lack PKPD hysteresis and can therefore be directly correlated to circulating dabigatran plasma levels over time.

[0287] Several papers (Stangier [4], Jaffer [5]) have related these measurements to the steady-state circulating plasma level of dabigatran (Cp) as shown in the following formula: For INR:

number

number

number

number

number

number

[0288] Additionally, new assays are under development that may allow for better calculation of dabigatran drug plasma levels.

[0289] For the specific example of dabigatran, an embodiment of a digital app or related method may include receiving patient data including patient coagulation metrics (also referred to herein as patient coagulation metrics) from coagulation test results. The patient coagulation metrics may include a drug plasma level (also referred to as a drug concentration). The patient coagulation metrics may include any of the coagulation metrics described above, and the embodiment may include calculating a measured drug plasma level based on the coagulation metrics. In some examples, the embodiment includes calibrating the dosing calculator and / or plasma level prediction model by comparing the measured drug plasma level with the corresponding plasma level metric calculated by the dosing calculator. Following calibration, the embodiment may include processing the patient data with the calibrated dosage calculator to determine an updated dosage of dabigatran to administer to the patient.

[0290] More generally, for any drug, the present embodiment may include calibrating the ML dosage calculator by comparing measured PKPD metrics with corresponding PKPD metrics calculated by the ML dosage calculator.

[0291] The challenge of measuring drug or coagulation levels via blood testing is relating them to the time of administration; therefore, trough levels are typically employed. This can be logistically challenging compared to taking samples at any time. Embodiments of the present disclosure enable personalized calculation of coagulation parameters and drug profiles by recording the time of medication administration and blood sample collection (which may be any time) as patient data. Embodiments of a digital app and associated method may include receiving the time of drug administration, the time of blood sample, and the measured drug plasma level as patient data, and combining the measured drug plasma level, the time of drug administration, and the time of blood sample with the plasma level-time profile estimated by the dosage calculator to fit the measured drug plasma level to a measurement-derived maximum plasma level (Cmax) or a measurement-derived trough plasma level (Ctrough). In other words, the digital app and associated method can convert a blood test taken at any time into a measurement-derived maximum plasma level or a measurement-derived trough plasma level. It will be appreciated that this time transformation can be applied more generally to measured PKPD metrics of any drug.

[0292] Embodiments of the digital app and related methods may also calibrate a plasma level prediction model and / or a dosage calculator based on measured drug plasma levels or coagulation metrics. Embodiments of the digital app and related methods may include receiving patient data including the time of drug administration and the time of the blood sample, estimating a drug plasma concentration at the time of the blood sample using a dosage calculator based on the dosage and the time of drug administration, and calibrating the plasma level prediction model and / or the dosage calculator based on the difference between the estimated drug plasma concentration and the measured drug plasma concentration. In some embodiments, the embodiments may include calibrating the plasma level prediction model and / or the dosage calculator based on the difference between the maximum plasma level estimated by the dosage calculator and the maximum plasma level derived from the measurement and / or the difference between the trough plasma level estimated by the dosage calculator and the trough plasma level derived from the measurement. In this manner, dosage can be adjusted to obtain a corrected ITL. Both the trough level and the maximum coagulation level can be used to recommend an optimal dose for a particular individual.

[0293] In embodiments where the digital app and associated methods are used with patients at high bleeding and clotting risk (e.g., surgery patients, cancer patients, elderly patients, patients prone to falls), periodic blood testing monitoring may be employed to ensure the accuracy of the dosage calculator and / or the underlying plasma level prediction model. The method may include updating the dosage calculator, the plasma level prediction model, the individualized target plasma level metric (e.g., the ratio or trough level described in the previous section), and / or the dosage prescribed for administration to the patient based on the calculated or measured drug plasma level.

[0294] The availability of coagulation test data also allows for personalized calibration of dose prediction models. Plasma level prediction models and dosage calculators can predict the circulating drug in an average patient at any time after drug ingestion. This will result in predicted coagulation test results. By comparing actual coagulation test results with predictions, the dose prediction model can be adjusted proportionally for individual characteristics, or vice versa.

[0295] If the dose-response can be reliably and individually calibrated, medications can be adjusted to accommodate acute or long-term changes in patient data, such as adjusting to accommodate temporary or long-term antiplatelet therapy administered after coronary stent placement (e.g., via individualized target plasma level metrics as described above). Drugs such as aspirin or NSAIDs may also be inadvertently taken for headache or pain without the patient realizing that their additive effects with dabigatran increase the risk of major bleeding.

[0296] In some embodiments, the digital app and associated methods can be used in conjunction with low-cost, home testing of dabigatran plasma levels. Combining the app and methods with such testing can calibrate the model and optimize / minimize the frequency of future required testing. In some embodiments, a single such assessment can be sufficient to improve predictions. Measuring drug levels at home is simpler and less expensive than clinical testing, and potentially can be performed at a convenient time during treatment (e.g., immediately prior to administration, i.e., at trough levels) rather than at any time during the day or during an emergency before surgery.

[0297] Other patient data parameters that may be monitored by the app or related methods may include signs and symptoms of bleeding, complete blood count, and a comprehensive metabolic panel (evaluating liver function tests, albumin, total bilirubin, and serum creatinine, among other things).

[0298] It will be understood that personal calibration via physiological testing is optional. In many instances (e.g., low-risk patients), HCPs may rely on dosage recommendations and / or PKPD metrics provided by the app and associated methods disclosed herein.

[0299] Monitoring side effects As described above, the app and related methods may include receiving patient data as self-reported side effects. In response to detecting a side effect above a respective sensitivity level, the digital app and related methods may take several actions, including alerting a clinician, advising the patient to schedule a medical appointment, recommending a physiological test for calibration (as described above), adjusting the first risk level, adjusting the second risk level, adjusting the individualized target PKPD metric, recalculating the dosage to administer to the patient based on the adjusted first risk level, the adjusted second risk level, and / or the adjusted individualized target PKPD metric, and presenting the updated dosage to the clinician or patient. By monitoring and responding to side effects, the app and methods can provide important feedback to the HCP regarding the need to obtain physiological measurements and advise dosage changes appropriate for that particular patient.

[0300] In the case of dabigatran, the digital app and associated methods may use suitable questions to monitor dabigatran side effects, including those that occur when the dosage is too high (see Table 2). Gastrointestinal disorders and bleeding, in particular, may be closely monitored. The first risk level may include a risk level for bleeding, and the second risk level may include a risk level for thrombosis. [Table 2]

[0301] FIG. 12 illustrates an exemplary patient monitoring process according to one embodiment of the present disclosure.

[0302] Following diagnosis 1272 by a healthcare provider (HCP) 1274, e.g., the HCP, a first step of the method includes receiving 1276, in a digital app 1278, patient data including demographic details entered by the patient 1280. The method may also include receiving 1282, in the digital app, the patient data from the HCP 1274 (e.g., via manual data entry or clinical notes, etc.).

[0303] After receiving the patient data, the method proceeds with processing the patient data 1284 using a dosage calculator to determine a starting medication regimen for administration to the patient 1280. Following calculation of the starting medication regimen, the method includes presenting the medication regimen 1286 to the HCP 1274. The HCP 1274 may take into account the starting medication regimen calculated by the app 1278 and provide an initial prescription to the patient 1280. The prescription may be provided 1287 to the digital app 1278 as patient data.

[0304] The patient 1280 begins treatment. During treatment, the method includes receiving updated patient data 1288 from the patient 1280, such as side effects or events, and receiving updated patient data 1290 from the HCP following physiological testing of PKPD metrics, such as blood coagulation metrics, drug plasma levels, and / or blood test results including CrCl. The method processes the updated patient data 1292 to determine updated dosages to administer to the patient. Processing the updated patient data 1292 may include calibrating an ML dosage calculator based on the physiological testing, as described above. Processing the updated patient data 1292 may include calculating an updated dosing regimen based on significant changes in patient data parameters that drive the calculator, such as changes in PKPD metrics. Processing the updated patient data 1292 may also include individualizing target PKPD metrics and calculating an updated dosing regimen, as described above. The method proceeds to presenting 1294 the updated dosing regimen to the HCP 1274.

[0305] The HCP may take into account the updated dosing regimen calculated by the app 1278 and provide an updated prescription to the patient 1280. The updated prescription may be provided 1296 to the digital app 1278 as further updated patient data. The lower loop shown in the figure may be repeated as treatment progresses and, optionally, as the patient makes further or periodic visits to the HCP 1274 (e.g., for further physiological testing).

[0306] Patient teaching and education Digital apps and related methods may provide patients with instructions, guidance, and education to support ongoing medication use. In the case of dabigatran, the lack of ongoing review may leave many patients feeling unsure about their medication, leading to poor adherence. For example, the impact of thrombosis and DOAC prescriptions on patients extends far beyond the immediate effects. Patients live with a great deal of fear and are unsure of what to do if further bleeding occurs. The management of venous thromboembolism, for example, has two phases: active and preventive. The first phase aims to prevent clot progression and early recurrence. This lasts for one month and, in some cases, may be extended to three months. After that, treatment becomes more flexible. Patients also have limited understanding of how to take medications, and many, like a course of antibiotics, interpret a prescription from the hospital as sufficient and are unaware that they need to request a refill. Some patients overlook the need to take certain medications with food.

[0307] Many patients do not know what to do if they experience another blood clot and mistakenly believe that if they have a blood clot in their leg, they should take preventative measures, such as moving their leg. However, the opposite is true: they should elevate their leg and rest. Many patients have local complications from the clot-affected vein, and post-phlebitic limb syndrome is very common. It is often inappropriately managed. Few people are aware of the nature and timing of compression stockings and bandage application. Many patients who have experienced a pulmonary embolism misinterpret their subsequent chest pain and shortness of breath. Misinterpretation of a benign cause can lead to anxiety and avoidance of exercise, resulting in deconditioning and increased risk of future cardiopulmonary events.

[0308] Similar concerns and uncertainties regarding the prescription of dabigatran may be experienced by patients taking the drug for preventative purposes, such as those with AF.

[0309] The fact that DOACs are not monitored in the same way that warfarin is monitored in a warfarin clinic means that educational opportunities are limited. Overall, it is estimated that 20-25% of patients do not take the dose originally prescribed.

[0310] Thus, as outlined below, the methods and devices of the present disclosure are intended to provide instructions, guidance, and education to patients to support their drug therapy by improving reliability and compliance.

[0311] Dose Administration Instructions Embodiments of the digital app and related methods may include providing medication dosages to patients via a patient device. The app may provide the patient with dosage information, including how and when to take the medication in relation to meals, how to address missed doses, and / or whether other medication-related issues exist. For example, taking a DOAC with food can slow drug absorption by up to three hours, and the app may instruct the patient accordingly. Patient compliance can be an issue with some medications (e.g., dabigatran and anticoagulants) because missing a dose can lead to reduced efficacy. The digital app can emphasize the need to take medications regularly as recommended by the HCP by using visual, auditory, and / or tactile prompts, alerts, alarms, reminders, and other alert mechanisms from the patient device. If a dose is missed, the app and related methods may use a dosage calculator and target PKPD metrics to calculate a new schedule of medication dosages for the temporary period and / or adjusted dosages (increases or decreases). The app and associated methods may include receiving (from an HCP or patient) a new candidate concomitant medication, comparing the concomitant medication to a reference concomitant medication list, and providing concomitant medication instructions, which may include revised dosages of the medications, instructions to avoid taking the concomitant medication, and / or instructions to consult an HCP.

[0312] Side effect guidance As described above, the app and related methods may include receiving patient data as a self-reported side effect. The digital app and related methods may provide guidance to the patient regarding the reported side effect or any other side effect to allay the patient's concerns. As described above, the app may alert the patient to seek medical attention in the event of a serious side effect.

[0313] Lifestyle and Risk Guidance The digital app and related methods may also incorporate and provide a comprehensive educational program, including advice on self-monitoring of side effects and other lifestyle guidance (as described above). In the case of dabigatran, this may include messages regarding preventive measures and lifestyle modifications to avoid potential stroke (such as smoking cessation, reducing alcohol consumption, travel time and habits, and hydration status). As described above in the output section, the app and methods may determine periods of bleeding risk and / or thrombosis risk based on the plasma level-time profile and suggest timing of risky activities (sports, long travel, etc.) relative to medication dosing and maximum plasma level (Cmax) as indicated by the plasma level-time profile. In this way, the app and methods can minimize the risk of thrombosis and bleeding while allowing patients to maintain their quality of life. Also, as described above, the app and methods may pre-calculate and suggest adjusted doses before and after high-risk procedures (e.g., surgery and long travel).

[0314] Educational advice may include instructions regarding management of the phlebitis limb (including duration of elevation, type of bandage or stocking and application procedure), as well as follow-up and management of any undesirable consequences.

[0315] The app and method may advise on the risks of various activities, particularly those with a high risk of falls, such as contact sports and skiing. The app and method may provide personalized risks for informed patient decision-making. For example, the app may advise on strategies to reduce risk during long journeys.

[0316] The app and method may provide advice to patients on lifestyle changes to reduce bleeding risk, including advice on improving balance and types of exercise to prevent falls, while still ensuring an optimal quality of life that may be affected if activity is curtailed too much.

[0317] The app and method track respiratory symptoms in patients who have experienced a pulmonary embolism, helping to identify whether chest pain and shortness of breath are merely residual effects of the original embolism or a new cause for concern. Prospective tracking overcomes patient memory issues.

[0318] For women, the app and method may advise on contraception and on managing heavy menstrual bleeding with anticoagulants. The app and method may incorporate a bleeding score, which helps determine whether menstruation is abnormal and whether further investigation is needed (e.g., to detect new cancers).

[0319] For patients who fall, the app and method can track the number and type of falls to help with risk management.

[0320] The examples described herein, including the ML dosage calculator, PKPD model, and method for calculating personalized target PKPD metrics, are illustrative. Therapeutic algorithms, including calculation of ideal and personalized PKPD metrics, can also be expressed via explicit rules (e.g., if...then...), Bayesian or other statistical inference derived from population data, or machine learning (e.g., deep learning). The type of algorithm can be selected for performance, suitability for the context, and governing regulatory framework. In some examples, the dosage calculator and / or PKPD model can be continuously refined as new data is periodically accumulated or revised for regulatory approval, in accordance with governing requirements and patient risk.

[0321] It will be understood that throughout this specification any reference to "near," "before," "just before," "after," "just after," "higher," or "lower," etc., can refer to the parameter in question being less than or greater than a threshold value, or between two threshold values, depending on the context. [1]:Paul A.Reilly et al,‘‘The Effect of Dabigatran Plasma Concentrations and Patient Characteristics on the Frequency of Ischemic Stroke and Major Bleeding in Atrial Fibrillation Patients:The RE-LY Trial(Randomized Evaluation of Long-Term Anticoagulation Therapy),Journal of the American College of Cardiology,Volume 63,Issue 4,2014. [2]:Liesenfeld et al,‘‘A Population pharmacokinetic analysis of the oral thrombin inhibitor dabigatran etexilate in patients with non-valvular atrial fibrillation from the RE-LY trial.’’ J Thromb Haemost.2011 Nov;9(11):2168-75. [3]:Hawkins et al,‘‘Variation of the Hemoglobin Level with Age and Sex’’,Blood,1954,9(10,999-1007. [4]:Stangier J et al,The pharmacokinetics,pharmacodynamics and tolerability of dabigatran etexilate,a new oral direct thrombin inhibitor,in healthy male subjects.’’ Br J Clin Pharmacol.2007 Sep;64(3):292-303. [5]:Jaffer IH et al,‘‘Comparison of the ecarin chromogenic assay and diluted thrombin time for quantification of dabigatran concentrations.’’J Thromb Haemost.2017 Dec;15(12):2377-2387.

Claims

1. 1. A method of generating a dosage calculator for determining a dosage of a drug to administer to a patient, the method comprising: receiving population data, including simulated population data calculated using a pharmacokinetic-pharmacodynamic PKPD model of the drug; and training a machine learning dosage calculator using the population data. method.

2. the machine learning dosage calculator is configured to calculate the dosage to administer to the patient by processing patient data and target PKPD metrics for the patient. The method of claim 1.

3. the population data includes drug data for a plurality of patients, the drug data for each patient including patient data, dosage data, and PKPD metrics; 3. The method according to claim 1 or 2.

4. the method comprising calculating the simulated population data for the drug using the PKPD model; 10. A method according to any one of the preceding claims.

5. Calculating simulated population data using the PKPD model defining simulated patient data for a patient population including a plurality of simulated patients and a corresponding plurality of simulated dosages; calculating PKPD metrics for each of the plurality of simulated patients by processing the simulated dosage and simulated patient data using the PKPD model to determine simulated PKPD metrics; determining the simulated population data including the simulated patient data, the simulated dosage, and the simulated PKPD metrics. The method of claim 4.

6. the simulated patient data includes one or more of renal function metrics, liver function metrics, patient age, patient ethnicity, patient gender, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, a pharmacogenomic profile, and a patient medication list; The method of claim 5.

7. defining the simulated patient data; receiving a population distribution for each parameter type of the simulated patient data; generating each simulated patient by stochastic selection of each parameter type according to a respective said population distribution; defining a number of discrete values ​​for each parameter type; generating each simulated patient data set as a different combination of one discrete value from each parameter type; 7. The method according to claim 5 or 6.

8. generating each simulated patient includes generating simulated patient data representing all combinations of one discrete value from each parameter type; The method of claim 7.

9. the PKPD model comprises a plasma level prediction model and the PKPD metric comprises a plasma level metric; The method according to any one of claims 5 to 8.

10. the PKPD model includes a time-based differential equation model for modeling the time dependence of the concentration of the drug in an effective region of the body as a function of patient data; 10. A method according to any one of the preceding claims.

11. the PKPD model comprises a two-compartment model for modeling the time dependence of central compartment drug concentration as a function of patient data; 10. A method according to any one of the preceding claims.

12. the population data includes actual population data; 10. A method according to any one of the preceding claims.

13. training the machine learning dosage calculator using the population data; training the machine learning dosage calculator using simulated population data to generate a partially trained machine learning dosage calculator; training the partially trained machine learning dosage calculator using real population data.

10. A method according to any one of the preceding claims.

14. further comprising validating the machine learning model using additional population data.

10. A method according to any one of the preceding claims.

15. the further population data is different from the population data; 15. The method of claim 14.

16. locking the machine learning dosage calculator to prevent further adjustments to the machine learning dosage calculator.

10. A method according to any one of the preceding claims.

17. calibrating the machine learning algorithm for the patient based on measured PKPD metrics obtained from physiological testing of the patient.

10. A method according to any one of the preceding claims.

18. the machine learning dosage calculator is configured to directly calculate the dosage for administration to the patient by processing patient data and target PKPD metrics; 10. A method according to any one of the preceding claims.

19. the patient data includes one or more of: renal function metrics, liver function metrics, patient age, patient ethnicity, patient sex, patient weight, patient body mass index, patient hemoglobin level, patient left ventricular function, a pharmacogenomic profile, and patient medication list; 20. The method of claim 18.

20. The drug is an anticoagulant, preferably a direct oral anticoagulant, preferably dabigatran; 10. A method according to any one of the preceding claims.

21. 1. A computer-implemented method for determining a dosage of a drug to administer to a patient, the method comprising: receiving patient data relating to a patient; processing the patient data with a dosage calculator to determine the dosage of the drug to administer to the patient, the dosage calculator comprising a machine learning algorithm trained using population data, including simulated population data obtained from a pharmacokinetic-pharmacodynamic PKPD model; and indicating the dosage. method.

22. receiving updated patient data; processing the updated patient data with the dosage calculator to determine an updated dosage; and indicating the updated dosage.

22. The method of claim 21.

23. the updated patient data includes measured PKPD metrics obtained from physiological testing of the patient; 23. The method of claim 22.

24. wherein the PKPD metric comprises drug concentration; 24. The method of claim 23.

25. calibrating the dosage calculator by adjusting the dosage calculator and / or the PKPD model using the measured PKPD metrics.

25. The method of claim 23 or 24.

26. processing the patient data in a dosage calculator to determine a dosage of dabigatran to administer to the patient; calculating the dosage to administer to the patient by processing the patient data and the patient's target PKPD metrics with the dosage calculator. The method according to any one of claims 21 to 25.

27. the method including determining the target PKPD metric as an individualized target PKPD metric based on the patient data.

27. The method of claim 26.

28. processing the patient data with the dosage calculator to determine PKPD metrics; the PKPD metric and and indicating one or more of the patient risks based on the PKPD metrics. The method according to any one of claims 21 to 27.

29. the PKPD metric comprises a drug concentration at an effective site in the patient's body; 29. The method of claim 28.

30. the patient data includes one or more dosing times at which the patient received a dose of the drug; the PKPD metric comprises a time-dependent drug concentration at an effective site in the patient's body based on the one or more dosing times; 30. The method of claim 28 or 29.

31. The PKPD metric is: a time profile of the drug concentration at an effective site in the patient's body; a maximum drug concentration at an effective site in the patient's body; the trough drug concentration at the effective site in the patient's body; the mean drug concentration at the effective site in the patient's body; the area under the curve of said time profile; the ratio of the maximum drug concentration to the trough drug concentration; the ratio of the maximum drug concentration to the area under the curve of the time profile; or Physiological measurements, including one or more of: The method according to any one of claims 28 to 30.

32. 1. A dosage calculator for determining a dosage of a drug to administer to a patient, said dosage calculator comprising: receiving patient data relating to a patient; processing the patient data with a dosage calculator to determine the dosage to administer to the patient, the dosage calculator comprising a machine learning algorithm trained using population data, including actual population data and / or simulated population data obtained from a pharmacokinetic-pharmacodynamic PKPD model; and indicating the dosage. Dosage calculator.

Images