Mixed models for data from thorough QT studies: part 2. One-step assessment of conditional QT prolongation

Concentration-QTc Modeling of the DPP-4 Inhibitor HSK7653 in a First-in-Human Study of Chinese Healthy Volunteers

Cofrogliptin (HSK7653) is a long-acting dipeptidyl peptidase-4 inhibitor for the treatment of type 2 diabetes mellitus with a twice-monthly dosing regimen. This study included 62 participants (48 without food effect, 14 with food effect) receiving single doses of HSK7653 (5, 10, 25, 50, 100, and 150 mg) or placebo. Pharmacokinetic samples were collected over 24 hours postdosing and sampling times are aligned with 12-lead electrocardiograms (ECGs) which were derived from continuous ECG recordings. For the concentration-QT interval corrected for heart rate (C-QTc) analysis, we used linear mixed-effects modeling to characterize the correlation between plasma concentrations of HSK7653 and the change from baseline in the QT interval which was corrected by Fridericia's formula (ΔQTcF). The result showed that a placebo-corrected Fridericia corrected QT interval (ΔΔQTcF) prolongation higher than 10 milliseconds is unlikely at the mean maximum observed concentration (Cmax) (411 ng/mL) associated with the recommended therapeutic doses (25 mg twice-monthly), even at the highest supratherapeutic concentration (2425 ng/mL). Thus, HSK7653 does not significantly affect QT prolongation at either recommended doses or the highest supratherapeutic concentration.

Long-Term Effects on QT Prolongation of Pretomanid Alone and in Combinations in Patients with Tuberculosis

Concentration-QTc modeling was applied to pretomanid, a new nitroimidazooxazine antituberculosis drug. Data came from eight phase 2 and phase 3 studies. Besides pretomanid alone, various combinations with bedaquiline, linezolid, moxifloxacin, and pyrazinamide were considered; special attention was given to the bedaquiline-pretomanid-linezolid (BPaL) regimen that has demonstrated efficacy in the Nix-TB study in subjects with extensively drug-resistant or treatment-intolerant or nonresponsive multidrug-resistant tuberculosis. Three heart rate corrections to QT were considered: Fridericia's QTcF, Bazett's QTcB, and a population-specific correction, QTcN. QTc increased with the plasma concentrations of pretomanid, bedaquiline's M2 metabolite, and moxifloxacin in a manner described by a linear model in which the three slope coefficients were constant across studies, visits within study, and times postdose within visit but where the intercept varied across those dimensions. The intercepts tended to increase on treatment to a plateau after several weeks, a pattern termed the secular trend. The slope terms were similar for the three QTc corrections, but the secular trends differed, suggesting that at least some of the secular trend was due to the elevated heart rates of tuberculosis patients decreasing to normal levels on treatment. For pretomanid 200 mg once a day (QD) alone, a typical steady-state maximum concentration of drug in plasma (Cmax) resulted in a mean change from baseline of QTcN of 9.1 ms, with an upper 90% confidence interval (CI) limit of 10.2 ms. For the BPaL regimen, due to the additional impact of the bedaquiline M2 metabolite, the corresponding values were 13.6 ms and 15.0 ms. The contribution to these values from the secular trend was 4.0 ms.

Assessing QT/QTc interval prolongation with concentration-QT modeling for Phase I studies: impact of computational platforms, model structures and confidence interval calculation methods

Modeling the relationship between drug concentrations and heart rate corrected QT interval (QTc) change from baseline (C-∆QTc), based on Phase I single ascending dose (SAD) or multiple ascending dose (MAD) studies, has been proposed as an alternative to thorough QT studies (TQT), in assessing drug-induced QT prolongation risk. The present analysis used clinical SAD, MAD and TQT study data of an experimental compound, AZD5672, to evaluate the performance of: (i) three computational platforms (linear mixed-effects modeling implemented via PROC MIXED in SAS, as well as in R using LME4 package and linear quantile mixed models (LQMM) implemented via LQMM package; (ii) different model structures with and without treatment- or time-specific intercepts; and (iii) three methods for calculating the confidence interval (CI) of QTc prolongation (analytical and bootstrap methods with fixed or varied geometric mean concentrations). We show that treatment- and time-specific intercepts may need to be included into C-∆QTc modeling through PROC MIXED or LME4, regardless of their statistical significance. With the intersection union test (IUT) in the TQT study as a reference for comparison, inclusion of these intercepts increased the feasibility for C-∆QTc modelling of SAD or MAD to reach the same conclusion as the IUT analysis based on TQT study. Compared to PROC MIXED or LME4, the LQMM method is less dependent on inclusion of treatment- or time-specific intercepts, and the bootstrap CI calculation methods provided higher likelihood for C-∆QTc modeling of SAD and MAD studies to reach the same conclusion as the IUT based on the TQT study.

Interdependence of baseline correction method and covariance structure for crossover TQT studies

Thorough QT/QTc (TQT) trials are conducted to assess a drug's risk potential for prolonging the QT interval. In a randomized crossover, multiple postdose electrocardiograph (ECG) readings are collected over time, within each period leading to a repeated-measures scenario. Additionally, baseline readings at single or multiple time points are collected prior to treatment. Two active, yet seemingly separate, areas of statistical research for TQT crossover studies concern the method of baseline correction and the specified covariance structure for repeated measures analysis. Due to interdependence of the covariance structure of baseline-corrected QTc values and the definition of baseline, these two research areas cannot be considered separately. We illustrate how the covariance structure for baseline-corrected QTc values differs under various definitions of baseline and provide recommendations on the choice of baseline: Time-matched baseline averaged over periods and time-matched baseline from the first period only are preferable in terms of efficiency; in case a period-specific correction is required to account for the treatment carryover effect, period-specific time-averaged baseline with multiple measurements taken predose in each period offers greater efficiency than a single predose baseline. This research on the interdependence of baseline correction method and covariance structure enables statisticians to use observed data from previous TQT studies to plan design and analysis for future studies.

The sodium glucose cotransporter 2 inhibitor empagliflozin does not prolong QT interval in a thorough QT (TQT) study

Background

Empagliflozin is a potent, selective sodium glucose cotransporter 2 (SGLT2) inhibitor in development as an oral antidiabetic treatment. This QT interval study assessed potential effects of empagliflozin on ventricular repolarisation and other electrocardiogram (ECG) parameters.

Methods

A randomised, placebo-controlled, single-dose, double-blind, five-period crossover study incorporating a novel double-placebo period design to reduce sample size, while maintaining full statistical power. Treatments: single empagliflozin doses of 25 mg (therapeutic) and 200 mg (supratherapeutic), matching placebo and open-label moxifloxacin 400 mg (positive control). Triplicate 12-lead ECGs of 10 second duration were recorded at baseline and during the first 24 hours after dosing. The primary endpoint was mean change from baseline (MCfB) in the population heart rate-corrected QT interval (QTcN) between 1–4 hours after dosing.

Results

Thirty volunteers (16 male, 14 female, mean [range] age: 34.5 [18–52] years) were randomised. The placebo-corrected MCfB in QTcN 1–4 hours after dosing was 0.6 (90% CI: -0.7, 1.9) ms and -0.2 (-1.4, 0.9) ms for empagliflozin 25 mg and 200 mg, respectively, below the ICH E14 defined threshold of regulatory concern 10 ms. Assay sensitivity was confirmed by a placebo-corrected MCfB in QTcN 2–4 hours post-dose of 12.4 (10.7, 14.1) ms with moxifloxacin 400 mg. Empagliflozin tolerability was good for all volunteers; 23.3% experienced adverse events (AEs) with empagliflozin and 27.6% with placebo. The most frequent AE was nasopharyngitis.

Conclusions/interpretation

Single doses of empagliflozin 25 mg and 200 mg were not associated with QTcN prolongation and were well tolerated in healthy volunteers.

Trial registration

ClinicalTrials.gov: NCT01195675

Mixed models for data from thorough QT studies: part 1. assessment of marginal QT prolongation

We investigate mixed models for repeated measures data from cross-over studies in general, but in particular for data from thorough QT studies. We extend both the conventional random effects model and the saturated covariance model for univariate cross-over data to repeated measures cross-over (RMC) data; the resulting models we call the RMC model and Saturated model, respectively. Furthermore, we consider a random effects model for repeated measures cross-over data previously proposed in the literature. We assess the standard errors of point estimates and the coverage properties of confidence intervals for treatment contrasts under the various models. Our findings suggest: (i) Point estimates of treatment contrasts from all models considered are similar; (ii) Confidence intervals for treatment contrasts under the random effects model previously proposed in the literature do not have adequate coverage properties; the model therefore cannot be recommended for analysis of marginal QT prolongation; (iii) The RMC model and the Saturated model have similar precision and coverage properties; both models are suitable for assessment of marginal QT prolongation; and (iv) The Akaike Information Criterion (AIC) is not a reliable criterion for selecting a covariance model for RMC data in the following sense: the model with the smallest AIC is not necessarily associated with the highest precision for the treatment contrasts, even if the model with the smallest AIC value is also the most parsimonious model.