Optimal design and QT-prolongation detection in oncology studies
Sylvain Fouiliard and Marylore Chenel
Department of Clinical Pharmacokinetics, Institut de Recherches Internationales Servier,
Objectives: QT interval prolongation is considered as a biomarker of torsade de pointe (TdP) in cardiac safety assessment in drug development. Thus, specific QT/QTc studies are usually performed in healthy volunteers, allowing an accurate estimation of such noisy data as QT interval length. This is however not possible in the specific context of oncology, where patients only can receive the drug.
Population approach may help the description of PKPD relationship while taking into account all sources of variability (e.g. circadian rhythm, inter-individual variability, inter-occasion variability, residual error) [1], but the clinical constraints of phase I/II studies in oncology limit electrocardiogram (ECG) schedules.
Based on both known population PK and QT circadian rhythm models, assuming the PKPD relationship and testing different effect sizes, we propose a design strategy in order to assess cardiac safety by optimizing ECG sampling times.
Our aim is to propose a cardiac safety assessment method, based on both optimal sampling design and population PKPD modelling. The ultimate goal is to estimate the power of detection of any potential effect of SX compound on QT interval length.
Methods: First, a population PK model of drug SX was developed based on both oral and IV phase I studies (45 patients). This model was used to simulate the concentration of the drug based on the administration schedule of further studies.
Secondly, a population model describing the circadian rhythm of QTc was developed using data from two former QT/QTc phase I studies including a total of 160 healthy volunteers under placebo. Both model buildings were performed using NONMEM VI with the FOCE-I method.
Based on preliminary experience in PKPD of QT length, we assume a linear relationship between drug concentration and QTc effect. A range of values of the drug effect on QTc was tested (from 5 ms to 100 ms).
At last, the ECG record times planned in the trials to come were evaluated using PopDes 3.0 design evaluation feature and were compared to the optimal ones obtained by D-optimality criteria (with Fedorov exchange optimization algorithm) [2]. Planned ECG schedules are on days 1 (predose, 1h and 4h after 1st dose, 1h and 4h after 2nd dose), and on days 2, 4, 14 and 22 (predose and 1h after 1st dose each day). In parallel, the precision of estimation of the drug effect parameter was used in the computation of the power of detection of a significant drug effect.
Results: SX concentration-time data were fitted with a 3-compartment model with a first-order absorption. Inter-individual variability was added on clearance CL, bioavailability F, absorption rate Ka and on inter-compartmental constants Q2 and Q3. The residual error was multiplicative. The circadian rhythm of QTc was modelled as a mesor and a sum of three cosine terms (one amplitude and one lag-time per cosine term), representing three periods of 24, 12 and 6 h, with inter-individual variability on every parameter except the second amplitude term, and an additive residual error. The drug effect is assumed to be proportional to the mesor.
The proposed design lead to a good estimation of every parameter of the model, according to the RSEs given by the population Fisher information matrix. Whatever the tested value of the drug effect, the statistical power of detection of a significant QT effect (i.e. that may cause a QT-prolongation adverse event) was found to be over 90 %. Comparison with optimal designs (under various time constraints) showed possible improvements for future studies.
Conclusions: This work proposes a modelling and simulation based strategy in order to make sure QT prolongation risk is correctly assessed in the context of clinical trials in oncology. Although the assumptions made on PKPD relationship is not negligible and will be assessed throughout further trials, the first results show a good power of detection of a QT-prolongation related adverse event with a feasible ECG recording design in oncology patients.
References:
[1] Piotrovsky, V. "Pharmacokinetic-pharmacodynamic modeling in the data analysis and interpretation of drug-induced QT/QTc prolongation". AAPS J 7.3 (2005): E609-E624.
[2] Ogungbenro K, Dokoumetzidis A, Aarons L. "Application of optimal design methodologies in clinical pharmacology experiments". Pharm Stat. (2009) Jul-Sep;8(3):239-52.
