Sancho-Araiz, Aymara (1); F. Trocóniz, Iñaki (1); Pérez-Ruixo, Juan (2), Dosne, Anne-Gaëlle (2)
(1) Pharmacometrics and Systems Pharmacology Department of Pharmaceutical Science, School of Pharmacy and Nutrition, University of Navarra; (2) Janssen R&D, Belgium
Objectives: Population PK analysis in confirmatory clinical trials have opened the possibility of providing individual estimates of drug exposure (such as steady-state AUC values) and their relationship with key efficacy and safety endpoints, which are key to validate the adequacy of the proposed dose regimen across the entire study population through exposure-response (ER) analysis.
However, an expected limitation of such ER analyses [1, 2] is that their interpretation might be biased if the amount of subject-level PK information is too limited. Indeed, individual PK parameters exhibit “shrinkage” [3], a metric which quantifies how much individual estimates regress towards the population mean under the given sampling schedule. It is believed that true underlying ER relationships are likely biased in studies where shrinkage is high.
The objective of this work is to provide valuable recommendations regarding scenarios in which high shrinkage can distort the ER relationship, shedding light on the implication for ER modelling. This information can be instrumental in ensuring the accuracy and reliability of ER analysis in drug development and clinical decision-making.
Methods:
A simulation/re-estimation study was set up to test the impact of different levels of shrinkage on ER relationships estimated using logistic regression.
A hypothetical study where subjects received 240 mg of androgen receptor inhibitor orally once daily was simulated, with different PK sampling schemes representing different levels of AUCss-shrinkage (0-61%). The PK model was inspired from a published PK model [4].
Then, the presence or absence of an adverse event (dichotomous response) was simulated as a function of the true individual area under the concentration curve after 24 hours (AUC0-24,ss), inspired from a published logistic regression ER model [4]. The ER model was then fitted to the data for each PK sampling schedule (e.g. for different levels of AUC shrinkage). Bias and precision were compared between shrinkage levels.
Different ER simulation model parameters were used to test contribution of the proportion of treatment effect (PTE) explained by the ER relationship on the results. The presence or absence of the selected adverse event was thus simulated assuming PK was the sole driver of response, and assuming both PK and drug treatment (irrespective of PK exposure) were drivers of response.
The number of subjects simulated was 2000, with a 1:1 randomization between placebo and treatment groups. This was in line with the original study, and as the number of subjects is not expected to impact shrinkage, other sample sizes were not considered.
Results: The E-R analyses showed that:
1. No significant bias in the estimated ER relationship was observed across simulation scenarios, regardless of shrinkage level. In extreme cases it induced a bias towards a steeper exposure response relationship.
2. Higher shrinkage generally resulted in higher uncertainty for the exposure response parameters.
3. The impact of shrinkage on potential bias did not depend on PTE, i.e. results were similar regardless of whether AUC was driving a minor or a major part of the response.
Conclusions: Similarly, to a previous publication [1], our study supports that ER relationships of dichotomous endpoints based on logistic regression can be reliably assessed even in the presence of high shrinkage in the pharmacokinetic exposure metric.
References:
[1] Abstr 10663 [www.page-meeting.org/?abstract=10663]
[2] Xu XS, Yuan M, Karlsson MO, Dunne A, Nandy P, Vermeulen A. Shrinkage in nonlinear mixed-effects population models: quantification, influencing factors, and impact. AAPS J. 2012 Dec;14(4):927-36
[3] Savic RM, Karlsson MO. Importance of shrinkage in empirical bayes estimates for diagnostics: problems and solutions. AAPS J. 2009 Sep;11(3):558-69.
[4] Pérez-Ruixo C, Pérez-Blanco JS, Chien C, Yu M, Ouellet D, Pérez-Ruixo JJ, Ackaert O. Population Pharmacokinetics of Apalutamide and its Active Metabolite N-Desmethyl-Apalutamide in Healthy and Castration-Resistant Prostate Cancer Subjects. Clin Pharmacokinet. 2020 Feb;59(2):229-244. ce 1.
Reference: PAGE 32 (2024) Abstr 11183 [www.page-meeting.org/?abstract=11183]
Poster: Methodology - Model Evaluation