Pieter Colin

Evidence generation for antihyperglycemic therapies in pediatric type 2 diabetes mellitus (T2DM) remains challenging, with most randomized trials failing to demonstrate statistically significant reductions in glycated hemoglobin (HbA1c), partly due to substantial between‐subject variability. To better understand the sources and implications of this variability, we conducted a model‐based longitudinal meta‐analysis of individual participant data from seven pediatric T2DM clinical trials submitted to the European Medicines Agency in support of marketing authorization applications. The analysis incorporated 3295 HbA1c observations from 809 participants and quantified the contributions of baseline disease severity, disease progression, placebo response, and treatment effects to longitudinal HbA1c dynamics. Insulin use and longer duration of diabetes were associated with higher baseline HbA1c and/or faster disease progression, while the placebo effect was substantial and highly variable across participants. Clinical trial simulations using the final model demonstrated that trials with fewer than 200 participants are unlikely to achieve 80% power to detect placebo‐corrected treatment effects smaller than −0.70%‐points over 26 weeks. Enrichment strategies targeting patients with lower variability, as well as model‐based estimators—including Bayesian approaches leveraging historical information—substantially reduced the required sample size. These findings indicate that current pediatric T2DM trials are typically underpowered due to underestimation of HbA1c variability and highlight opportunities for more efficient trial designs. By quantifying disease‐ and treatment‐related drivers of HbA1c trajectories and demonstrating the potential of model‐informed strategies to improve power, this work provides a framework to enhance pediatric T2DM drug development and support regulatory decision‐making.

Mixed effects models are a backbone of pharmacometrics, and NONMEM software, with the first‐order conditional method with interaction having become the de facto industry standard for model estimation. Documentation exists for the general mathematical methodology for estimation, but many technical and implementation details are lacking. OpenPMX aims to enable nonlinear mixed‐effects modeling and estimation in a transparent and efficient manner, with open source licensing allowing for broad application and development. Model parameter estimation bias and root mean squared error (RMSE) obtained using OpenPMX were compared to that using NONMEM for five population models and datasets with varying degrees of complexity. For each model and dataset, repeated simulation and estimation were performed, and the per‐dataset difference in precision for parameter estimates was calculated for OpenPMX versus NONMEM. We found that the bias and RMSE of OpenPMX are comparable to the industry standard NONMEM, and in some cases slightly better. The project has low complexity, few dependencies, and is open source, with all technical details open for inspection, auditing, and scientific collaboration.

Long‐acting injectable medicinal products (LAIs) prolong drug release and thereby aim to enhance adherence and patient outcomes. European regulatory guidelines require the conduct of single‐ and multiple‐dose trials to exclude differences in drug release between non‐steady and steady state conditions. The complexity of these trials may however hamper the development of LAIs. This study aimed to examine whether drug release is different after single‐ and multiple‐dose administration using clinical pharmacokinetic (PK) data of a sample of five regulatory‐approved LAIs. Single‐ and multiple‐dose data were extracted from an internal regulatory database. Population pharmacokinetic models with different absorption structures were developed using nonlinear mixed‐effect modeling based on the single‐dose data of every LAI. The best‐fitting models were used to predict the pharmacokinetic profiles after multiple‐dose administration. The absorption of LAIs after single‐dose administration was best described with (parallel) first‐order absorption structures (with and without lag‐time). After multiple‐dose administration, the mean model accuracy was 93% (minimum to maximum: 70%–122%), and 7 out of 10 observed pharmacokinetic variables (i.e., area under the plasma concentration—time curve, minimum and maximum concentration) met the pre‐specified acceptance criteria. In conclusion, multiple‐dose PK characteristics can be predicted using models developed from single‐dose PK data, which indicates that drug release may not be very different between dosing conditions in this sample of regulatory‐approved LAIs. Nevertheless, additional studies on other LAIs are required to test the generalizability of our findings and to increase our understanding of the limitations of the proposed model‐based approach vis‐à‐vis the current evidentiary standard.

Background and Objective End-stage kidney disease (ESKD) patients undergoing haemodialysis (HD) require a dosing regimen that balances the low endogenous clearance with the additional dialyser clearance. This study aimed to expand a previously proposed general-purpose pharmacokinetic model for piperacillin/tazobactam with a new population of ESKD patients undergoing intermittent high-flux haemodialysis. Methods Inter- and intradialytic blood samples were collected in ESKD patients undergoing intermittent high-flux haemodialysis, in HD or haemodiafiltration (HDF) mode, who received piperacillin/tazobactam during routine care. The previous general-purpose model was expanded to reflect changes in the pharmacokinetics in the new patient population. A covariate search was performed focussing on factors that explained variability between patients in endogenous and dialysis clearance. Simulations were performed to determine the probability of target attainment with current dosing recommendations in this specific population. Results In 20 ESKD patients, 195 piperacillin/tazobactam concentrations were determined. The general-purpose model was successfully expanded, wherein endogenous piperacillin/tazobactam clearance in patients with/without residual diuresis was 63% (95% confidence interval [CI] 49.5–73.0%) and 78.6% (95% CI 66.3–86.4%) lower compared with the general population, respectively. Extraction ratios of piperacillin and tazobactam ranged from 64 to 80%. Differences in probability of target attainment (PTA) for piperacillin were observed between patients with normal kidney function and ESKD patients undergoing haemodialysis with current dosing recommendations. Conclusion We successfully expanded a general-purpose model to reflect the piperacillin/tazobactam pharmacokinetics in ESKD patients undergoing intermittent haemodialysis using high-flux dialysers. The current dosing recommendations provide inconsistent probability of target attainment in ESKD patients compared with the general population.

Background: Remimazolam is not approved for use in pediatric patients. The pharmacokinetics of remimazolam have been reported to be similar to those of adult patients after scaling for body size. This article reports on the pharmacokinetics and pharmacodynamics of pediatric patients aged 6 to 18 yr and a subsequent model-based optimization of the used dosing regimen. Methods: Thirty-one patients were included in the trial and stratified across four treatment arms: bolus administration, infusion, bolus plus fentanyl, or infusion plus fentanyl. The University of Michigan (Ann Arbor, Michigan) Sedation Scale (UMSS) was used to assess the depth of sedation. Blood samples were drawn to measure the concentrations of remimazolam and its metabolite CNS7054. Population pharmacokinetic pharmacodynamic modeling was performed in NONMEM (GloboMax LLC, USA). Results: A population pharmacokinetic model was developed for remimazolam and CNS7054. The elimination clearance of remimazolam was 0.70 l · min ^–1 · 70 kg ^–1 . A proportional odds model combined with a simplified Minto model described the observed UMSS well. The EC50 of remimazolam for a UMSS score of 3 or greater was 777 ng · ml ^–1 in the absence of fentanyl, and decreased to 655, 533, and 287 ng/ml for concomitant fentanyl steady state concentrations of 1, 2, or 4 ng · ml ^–1 , respectively. Simulations confirmed that the studied dosing regimen resulted in 9.2 to 22.0% of patients not reaching a UMSS score of 3 or greater at the end of the induction. Model-based optimization resulted in higher per-kilogram dosages and the removal of the maximum allowable dose. Simulations indicated that the percentage of patients achieving a UMSS score of 3 or greater can be expected to be high (88 to 97%). Conclusions: This study has shown that the pharmacokinetics of remimazolam are likely different between children 6 yr or older and adults (after correcting for size). In addition, the exposure–response relationship shows that to effectively use remimazolam for procedural sedation in children 6 yr or older, the dosing regimen should be modified to allow for higher remimazolam exposures.

Aims: For new medicines, drug companies obtain regulatory approval on the strategy to generate evidence in the paediatric population, which can be supported by extrapolation of evidence obtained in a reference population. This study investigated whether paediatric marketing authorization applications (PMAAs) supported by extrapolation based on exposure‐matching were more successful—i.e. approval of the targeted paediatric population—and efficient—i.e. duration of the drug development—compared to PMAAs not supported by extrapolation. Methods: Data was extracted from completed paediatric investigation plans (PIPs), associated drug labels and public assessment reports published on the European Medicines Agency website. Assessment reports were evaluated to assess whether PMAAs were supported by extrapolation based on exposure‐matching. Wilcoxon rank‐sum tests were used to compare PMAAs supported and not supported by extrapolation based on exposure‐matching for outcomes of interest. Results: Exposure‐matching supported the benefit/risk assessment of 39.6% of the PMAAs. Targeted and approved minimum age of the paediatric population were comparable for PMAAs where extrapolation based on exposure‐matching supported the benefit/risk assessment (2.0 vs . 2.0 years, P ‐value = .72), but not for PMAAs not supported by extrapolation (0.2 years vs . 0.5 years, P ‐value = .05). Completion of drug development was 5.4 years vs . 4.3 years ( P = .04) in PMAAs supported by extrapolation based on exposure‐matching compared to those not supported by extrapolation, respectively. Conclusions: PMAAs supported by extrapolation based on exposure‐matching succeeded more often in obtaining marketing approval in the targeted paediatric population than PMAAs not supported by exposure‐matching, but were also less efficient.

Background: Synergistic effects between opioids and remimazolam on Bispectral Index (BIS) and Modified Observer’s Assessment of Alertness and Sedation (MOAAS) score were previously described. This study aimed to characterize the influence of remifentanil on the sedative properties of remimazolam as measured by MOAAS, BIS, and tolerance to laryngoscopy or tetanic stimulation (TOL or TOTS) and to determine target concentrations that maximize MOAAS 2 or 3. Methods: A three-period, crossover, dose-ranging clinical trial was performed in 24 healthy volunteers. In all periods, remimazolam was administered using a step-up and step-down target controlled infusion protocol (50 to 2,000 ng/ml). Stable remifentanil target concentrations of 0.5 ng/ml and 0.1 to 4.0 ng/ml were maintained in periods 2 and 3, respectively. Remifentanil, remimazolam, and CNS7054 (metabolite) concentrations and MOAAS, BIS, TOL, and TOTS were collected in each step of the target controlled infusion protocol. Data were analyzed using nonlinear mixed-effects models, where P ≤ 0.01 was considered significant. Results: Remifentanil reduced the apparent clearance of CNS7054 with a half-maximum inhibition at 8.0 ng/ml (95% CI, 5.5 to 13.4 ng/ml). A pharmacodynamic interaction was detected on all endpoints. Simulations indicate that the probability of observing a MOAAS 2 or 3 is highest at remimazolam target concentration of 275, 250, or 200 ng/ml combined with 0, 0.1, or 0.5 ng/ml remifentanil resulting in probabilities of 45%, 45%, and 44%, respectively. Additionally, simulations indicate that the highest probability of observing TOTS and TOL was 93.3% and 85.5%, respectively, at the highest studied target concentrations. Conclusions: A pharmacokinetic and pharmacodynamic drug–drug interaction between remimazolam and remifentanil was quantified in this clinical trial. Appropriate target concentrations for MOAAS and BIS could be estimated, but for TOL and TOTS, the trial design did not allow to fully characterize the exposure–response relationship.