Concentration-QTc analysis versus Intersection Union Test: sample size considerations for a thorough QT study
Aziz Ouerdani (1), Julius Nangosyah (1), Annemie Buelens (1), Tristan Baguet (1), Dymphy Huntjens (1*), Juan Jose Perez-Ruixo (1), Oliver Ackaert (1), Nele Goeyvaerts (1)
(1) Janssen Research and Development, Beerse, Belgium, * Currently at Priothera SAS, Saint-Louis, France
Objectives: Dedicated Thorough QT (TQT) studies are conducted to assess the proarrhythmic potential of non-antiarrhythmic drugs based on evaluating the corrected QT (QTc) interval prolongation [1]. The Intersection Union Test (IUT) is the traditional test used in a TQT study, however, Concentration-QTc (C-QTc) modeling has the potential to reduce the sample size while maintaining a robust assessment of QTc interval prolongation [2, 3]. A TQT study is considered negative when the upper bound of the two-sided 90% confidence interval (CI) for the mean difference in baseline-corrected QTc (?QTc) between test drug and placebo excludes 10 milliseconds (ms), at all selected time points for IUT and similarly at the highest clinically relevant exposure for C-QTc analysis [1, 4]. The objective of this simulation study was to determine the sample size of a TQT study with a parallel design, when C-QTc analysis or IUT are considered as primary analysis.
Methods: Because of the long half-life of the compound, the TQT study considered had a parallel design with 1:1 randomization to supratherapeutic dose or placebo, and rich paired PK-ECG sampling. TQT studies were simulated considering 6, 12, 18, 24, 30, 36 and 48 subjects per arm, and assuming at the target Cmax a mean placebo-corrected ?QTc (??QTc) of 0 ms, 3 ms and 5 ms to assess the power (true negative rate) and 10 ms to assess the type I error (false negative rate). Plasma concentrations were simulated from a 2-compartment disposition population PK model with time-dependent apparent clearance. ΔQTc values were simulated from a C-QTc model developed on time-matched PK and Holter ECG data from the First-in-Human study, including an intercept, concentration slope and treatment-specific intercept, but ignoring time effects. Inter-individual variability (IIV) and residual error on PK and ΔQTc were considered. For the IIV of ΔQTc intercept and concentration slope, two scenarios were explored: sampling from the Empirical Bayes Estimates (EBEs) of the developed C-QTc model or assuming a fixed standard deviation (SD) to mimic the lower variability observed in internal TQT studies. For each simulated TQT study, IUT was performed considering 8 pre-specified time points around tmax and a C-QTc model was fitted and for the latter, the two-sided 90% CI for the mean ΔΔQTc was evaluated at the geometric mean Cmax.
Results: For all scenarios, the power to exclude 10 ms and type I error were larger for C-QTc analysis compared to IUT. The assumed IIV of ΔQTc had a large impact on power. For the fixed SD scenario, the power of C-QTc analysis was >80% for a study with 6, 12 and 18 subjects per arm (or higher) for a mean ??QTc effect of 0 ms, 3 ms and 5 ms, respectively. In comparison, the power of IUT was >80% for a study with 24 and 48 subjects per arm for a mean ??QTc effect of 0 ms and 3 ms, respectively. For the EBEs scenario, the power of C-QTc analysis was >80% for a study with 36 subjects per arm for a mean ??QTc effect of 0 ms, however, 48 subjects per arm was insufficient to achieve a power of 80% for small mean effects of 3 ms or 5 ms. The latter was observed for IUT as well, already for a mean ??QTc effect of 0 ms. For C-QTc analysis, the type I error decreased with increasing sample size and was <5% for ≥12 subjects per arm, which is considered acceptable. For IUT, the type I error was 0%-0.7% confirming that the procedure is conservative regarding the intended 5% level [5].
Conclusions: A simulation study was conducted to determine the sample size for a parallel design TQT study, comparing C-QTc and IUT as primary analysis and hence complementing published C-QTc simulation studies [2, 3, 6]. Assuming a ΔQTc variability as observed in internal TQT studies, it was shown that a TQT study with 12 subjects in both supratherapeutic dose and placebo arms entailed a power of >80% for the C-QTc analysis to exclude 10 ms for a mean ??QTc effect of 3 ms, while controlling the type I error below 5%, similar to Garnett et al. [3]. In comparison, a total of 48 subjects per arm would be needed to entail a similar power for IUT, showing that the sample size of a parallel TQT study can be reduced if C-QTc would be considered as primary analysis. An increased IIV resulted in a lower power for both C-QTc and IUT.
References:
[1] ICH E14 (2005). The clinical evaluation of QT/QTc interval prolongation and proarrhythmic potential for non-antiarrhythmic drugs.
[2] Ferber G, Zhou M, Darpo B (2015). Detection of QTc effects in small studies--implications for replacing the thorough QT study. Ann Noninvasive Electrocardiol., 20(4), 368-77. doi: 10.1111/anec.12227.
[3] Garnett C, Needleman K, Liu J, Brundage R, Wang Y (2016). Operational Characteristics of Linear Concentration-QT Models for Assessing QTc Interval in the Thorough QT and Phase I Clinical Studies. Clin Pharmacol Ther., 100(2): 170-8. doi: 10.1002/cpt.361.
[4] ICH E14 Q&As (R3) (2016). The clinical evaluation of QT/QTc interval prolongation and proarrhythmic potential for non-antiarrhythmic drugs (R3) – questions and answers.
[5] Zhang J and Machado SG (2008). Statistical Issues Including Design and Sample Size Calculation in Thorough QT/QTc Studies, Journal of Biopharmaceutical Statistics, 18:3, 451-467, doi: 10.1080/10543400802020938.
[6] Tsamandouras N, Duvvuri S, Riley S (2019). Impact of Phase 1 study design on estimation of QT interval prolongation risk using exposure-response analysis. J Pharmacokinet Pharmacodyn., 46(6):605-616. doi: 10.1007/s10928-019-09661-4.