Using pharmacokinetics to individualize hemophilia therapy

Abstract

Prevention and treatment of bleeding in hemophilia requires that plasma clotting factor activity of the replaced factor exceeds a defined target level. Most clinical decisions in hemophilia are based on implicit or explicit application of pharmacokinetic measures. The large interindividual variability in pharmacokinetics of factor concentrates suggests that relying on the average pharmacokinetic characteristics of factor concentrates would not allow optimizing the treatment of individual patients; for example, adjusting the frequency of infusions and targeting a specific clotting factor activity level on a case-by-case basis. However, individual pharmacokinetic profiles are seldom assessed as part of routine clinical care. Population pharmacokinetics provide options for precise and convenient characterization of pharmacokinetics characteristics of factor concentrates, simplified individual pharmacokinetic profiling, and individualized dosing. Population pharmacokinetics allow for the incorporation of determinants of interpatient variability and reduces the need for extensive postinfusion plasma sampling. Barriers to the implementation of population pharmacokinetics are the need for concentrate-specific pharmacokinetic models, Bayesian calculation power, and specific expertise for production, validation, and appraisal of forecasted estimates. Population pharmacokinetics provide an important theoretical and practical contribution to tailoring the treatment of hemophilia. The need remains for prospective exploration of the clinical impact of tailoring hemophilia treatment based on individual pharmacokinetics, and for the systematic validation of existing software solutions and concentrate-specific models.

Figure 1.

Comparing use of average PK parameters to individualized PK profiles. The figure represents the % time spent over 0.01 IU/mL for 10 hypothetical patients. In blue, the time spent by each patient was calculated when they were treated with a standard half-life factor VIII; below each patient label, the U/kg administered to obtain the displayed results. On average, the population of patients was treated with 30 UI/kg and spent 68% of their time above 0.01 IU/mL, with a minimum of 45 and a maximum of 90. The orange columns show the % time spent over 0.01 IU/mL obtained by switching all patients to the same dosage of an extended half-life concentrate; the average time spent over 0.01 IU/mL increases to 100, ranging from 68% to >100%. The green columns show the % time spent over 0.01 IU/mL obtained by switching all patients to the extended half-life concentrate adjusting their dose to their PK (the 5 patients identified by the green arrows had their dose decreases, those identified by the red arrows had their dose increased); the average time spent over 0.01 IU/mL increases to 100, ranging from 95% to >100%. Note that we have hypothesized not to decrease the dose of patient number 3 because it is treated with 20 IU/kg. The average dose in the population would remain 30 IU/kg.

Figure 2.

Derivation of a population PK model by hierarchical multivariable regression. Dots represent postinfusion plasma measurements of clotting factor activity; each different color identifies a different patient; when more than 1 time point is available for a patient, the association among them is considered in the regression model. The red dashed line represents the average regression line; the blue dotted regression lines represent the upper and lower boundaries of the regression.

Figure 3.

Individualized plasma clotting factor activity over time profile: case 1. The red line with hollow points shows the measured plasma clotting factor activities used to estimate the PK profile for the patient (green dashed line). The solid line shows the predicted PK profile for the simulated regimen. The further away the predicted (solid) line is from the measured (hollow point) line and from the estimated (dashed) predicted individual PK profile, the lower our confidence in the precision of the calculation. (A) The threshold after 2000 IU on Monday-Wednesday-Friday. (B) After 2000 units every third day. (C) Every 3.5 days (Sunday morning and Thursday night). The calculations include the weekly average consumption, which is 6000 U with the current regimen, 5000 with solution B, 4000 with solution C. The plots were produced with the calculator function of the WAPPS-Hemo software, a freely available, registration based online tool providing individualized PK profiles using a Bayesian population approach (www.wapps-hemo.org). The interactive graph relative to this example is available at https://demo.wapps-hemo.org/PatientResult.aspx?PIR=Tb1LL6RgexRq8cZ2K5-pEA.

Figure 4.

Individualized plasma clotting factor activity over time profile: case 2. The red line with hollow points shows the measured plasma clotting factor activities used to estimate the PK profile for the patient (green dashed line). The solid line shows the predicted PK profile for the simulated regimen. The further away the predicted (solid) line is from the measured (hollow point) line and from the estimated (dashed) predicted individual PK profile, the lower our confidence in the precision of the calculation. (A) Individual PK profile on the current dose of 1000 IU on Monday and Thursday. (B) The effect of increasing the dose to 2000 IU. (C) The dose that would be required on the Friday infusion to obtain a trough >0.025. (D) The booster dose that would be required on Sunday to obtain a trough >0.025. The plots were produced with the calculator function of the WAPPS-Hemo software, a freely available, registration-based online tool providing individualized PK profiles using a Bayesian population approach (www.wapps-hemo.org). The interactive graph relative to this example is available at https://demo.wapps-hemo.org/PatientResult.aspx?PIR=pVyUp4UHO0ib9GzrozEx2g, and the interactive graph relative to case 3 is available at https://demo.wapps-hemo.org/PatientResult.aspx?PIR=QYLy46pQOGEEn9PewiBvHA.