Johan E. Wallin

Prediction-corrected visual predictive checks for diagnosing nonlinear mixed-effects models.

Informative diagnostic tools are vital to the development of useful mixed-effects models. The Visual Predictive Check (VPC) is a popular tool for evaluating the performance of population PK and PKPD models. Ideally, a VPC will diagnose both the fixed and random effects in a mixed-effects model. In many cases, this can be done by comparing different percentiles of the observed data to percentiles of simulated data, generally grouped together within bins of an independent variable. However, the diagnostic value of a VPC can be hampered by binning across a large variability in dose and/or influential covariates. VPCs can also be misleading if applied to data following adaptive designs such as dose adjustments. The prediction-corrected VPC (pcVPC) offers a solution to these problems while retaining the visual interpretation of the traditional VPC. In a pcVPC, the variability coming from binning across independent variables is removed by normalizing the observed and simulated dependent variable based on the typical population prediction for the median independent variable in the bin. The principal benefit with the pcVPC has been explored by application to both simulated and real examples of PK and PKPD models. The investigated examples demonstrate that pcVPCs have an enhanced ability to diagnose model misspecification especially with respect to random effects models in a range of situations. The pcVPC was in contrast to traditional VPCs shown to be readily applicable to data from studies with a priori and/or a posteriori dose adaptations.

Population Pharmacokinetic-Pharmacodynamic Model for Neutropenia with Patient Subgroup Identification: Comparison across Anticancer Drugs

Purpose: Cancer chemotherapy, although based on body surface area, often causes unpredictable myelosuppression, especially severe neutropenia. The aim of this study was to evaluate qualitatively and quantitatively the influence of patient-specific characteristics on the neutrophil concentration-time course, to identify patient subgroups, and to compare covariates on system-related pharmacodynamic variable between drugs. Experimental Design: Drug and neutrophil concentration, demographic, and clinical chemistry data of several trials with docetaxel (637 patients), paclitaxel (45 patients), etoposide (71 patients), or topotecan (191 patients) were included in the covariate analysis of a physiology-based pharmacokinetic-pharmacodynamic neutropenia model. Comparisons of covariate relations across drugs were made. Results: A population model incorporating four to five relevant patient factors for each drug to explain variability in the degree and duration of neutropenia has been developed. Sex, previous anticancer therapy, performance status, height, binding partners, or liver enzymes influenced system-related variables and α1-acid glycoprotein, albumin, bilirubin, concomitant cytotoxic agents, or administration route changed drug-specific variables. Overall, female and pretreated patients had a lower baseline neutrophil concentration. Across-drug comparison revealed that several covariates (e.g., age) had minor (clinically irrelevant) influences but consistently shifted the pharmacodynamic variable in the same direction. Conclusions: These mechanistic models, including patient characteristics that influence drug-specific parameters, form the rationale basis for more tailored dosing of individual patients or subgroups to minimize the risk of infection and thus might contribute to a more successful therapy. In addition, nonsignificant or clinically irrelevant relations on system-related parameters suggest that these covariates could be negligible in clinical trails and daily use.

A tool for neutrophil guided dose adaptation in chemotherapy

Chemotherapy dosing in anticancer treatment is a balancing act between achieving concentrations that are effective towards the malignancy and that result in acceptable side-effects. Neutropenia is one major side-effect of many antitumor agents, and is related to an increased risk of infection. A model capable of describing the time-course of myelosuppression from administered drug could be used in individual dose selection. In this paper we describe the transfer of a previously developed semi-mechanistic model for myelosuppression from NONMEM to a dosing tool in MS Excel, with etoposide as an example. The tool proved capable to solve a differential equation system describing the pharmacokinetics and pharmacodynamics, with estimation performance comparable to NONMEM. In the dosing tool the user provides neutrophil measures from a previous treatment course and request for the dose that results in a desired nadir in the upcoming course through a Bayesian estimation procedure.

Population pharmacokinetics of tacrolimus in pediatric hematopoietic stem cell transplant recipients: new initial dosage suggestions and a model-based dosage adjustment tool.

The population pharmacokinetics of tacrolimus was described in 22 pediatric hematopoietic stem cell transplant recipients, and a model-based dosage adjustment tool that may assist with therapy in new patients was developed. Patients received tacrolimus by continuous intravenous (IV) infusion (0.03 mg·kg−1·d−1) starting 2 days before transplantation, with conversion to oral therapy 2-3 weeks after transplant. Population pharmacokinetic analysis was performed using NONMEM. A Bayesian dosage adjustment tool that searches for individual parameter estimates to describe concentration measurements, counterbalanced by the final population model, was created in Excel. Typical clearance was 106 mL·h−1·kg−0.75, typical distribution volume was 3.71 L/kg, and typical bioavailability was 15.7%. Tacrolimus clearance decreased with increasing serum creatinine, and bioavailability decreased with postoperative day. A Bayesian dosage adjustment tool capable of suggesting an initial infusion rate based on patient covariate values and devising a further individualized dosage regimen as drug concentration measures become available was developed. Predictions from the model showed that current IV dose recommendations of 0.03 mg·kg−1·d−1 may potentially produce toxic drug concentrations in this patient population, whereas current oral conversion of 4 times the adjusted IV dose may lead to subtherapeutic concentrations. A more suitable infusion rate to obtain a steady state concentration of 12 ng/mL was predicted to be 0.035 mg·kg−0.75·d−1. An additional loading dose of 0.07 mg·kg−1·d−1 (total dose: 0.07 mg·kg−1·d−1 + 0.035 mg·kg−0.75·d−1) during the first 24 hours of therapy should allow rapid achievement of steady state concentrations. A conversion factor of 6 from IV to enteric therapy may be more suitable. Such dosage recommendations may be site specific. The appropriateness of targets was not investigated in this study. The Bayesian dosing adjustment tool and suggested dose recommendations need to be evaluated in a prospective study before they can be applied in the clinical setting.

Population pharmacokinetics of tacrolimus in pediatric liver transplantation : early posttransplantation clearance

Background: Tacrolimus is an immunosuppressant with a narrow therapeutic window, with considerable pharmacokinetic variability. Getting sufficient concentrations in pediatric liver transplantation is imperative, but it has proven difficult in the immediate posttransplantation period in particular. A predictive pharmacokinetic model could be the basis for development of a novel initial dose schedule, and therapeutic drug monitoring with Bayesian methodology. Methods: The predictive capacity of 2 previously developed population pharmacokinetic models of tacrolimus in pediatric liver transplant recipients was tested in 20 new patients using Bayesian forecasting. Predictive performance was poor in the immediate posttransplant period with tacrolimus pharmacokinetics changing rapidly. A new population pharmacokinetic model, focusing on the immediate posttransplant period, was subsequently developed in 73 patients. Results: An increase in the apparent clearance of tacrolimus in the first few weeks after transplant was evident. Typical apparent clearance of tacrolimus was 0.148 L·h−1·kg−0.75 immediately after transplantation, increasing to a maximum of 1.37 L·h−1·kg−0.75. Typical apparent distribution volume was 27.2 L/kg. Internal and external validation studies confirmed the predictive capabilities of the developed model. Simulation studies reveal that in 60% of subjects the current initial standard dose without subsequent dosage adjustments overshoot the desired trough concentration range of 10–20 ng/mL. An alternative dosing schedule was developed based on allometric scaling with an initial loading dose followed by a maintenance dose increasing with time. Conclusions: A population pharmacokinetic model for tacrolimus was developed, to better describe the early posttransplantation phase. This model has the potential to aid therapeutic drug monitoring and was also used to suggest a revised dosing scheme in the intended population.

Model-Based Neutrophil-Guided Dose Adaptation in Chemotherapy: Evaluation of Predicted Outcome with Different Types and Amounts of Information

One of the most employed approaches to reduce severe neutropenia following anticancer drug regimens is to reduce the consecutive dose in fixed steps, commonly by 25%. Another approach has been to use pharmacokinetic (PK) sampling to tailor dosing, but only rarely have model-based computer approaches utilizing collected PK and/or pharmacodynamic (PD) data been used. A semi-mechanistic model for myelosuppression that can characterize the interindividual and interoccasion variability in the time-course of neutrophils following administration of a wide range of anticancer drugs may be used in a clinical setting for model-based dose individualization. The aim of this study was to compare current stepwise procedures to model-based dose adaptation by simulations, and investigate if the overall dose intensity in the population could be increased without increasing the risk of severe toxicity. The value of various amounts of PK- and/or PD-information was compared to standard dosing strategies using a maximum a posteriori procedure in NONMEM. The results showed that when information on neutrophil counts was available, the additional improvement from PK sampling was negligible. Using neutrophil sampling at baseline and an observation near the predicted nadir increased the number of patients in the target range by 27% in comparison with a one-sided 25% dose adjustment schedule, while keeping the number of patients experiencing severe toxicity at a comparable low level after five courses of treatment. High interindividual variability did not limit the benefit of model-based dose adaptation, whereas high interoccasion variability was predicted to make any dose adaptation method less successful. This study indicates that for successful model-based dose adaptation clinically, there is no need for drug concentration sampling, and that one extra neutrophil measurement in addition to the pre-treatment value is sufficient to limit severe neutropenia while increasing dose intensity.