Positive Control in a “Thorough QT/QTc Study”

In a recent issue of the Journal of Clinical Pharmacology, Chatterjee et al1 reported the first “thorough QT/QTc study” on GLP-1 analog liraglutide. Although the study provided important scientific information on the effects of liraglutide on the QT/QTc intervals, treatment of the positive control deserves additional discussion. In this study the authors claimed that the upper bound of the 95% one-sided confidence intervals for the maximum mean difference of time-matched, baseline-subtracted QTc values between the positive control moxifloxacin and placebo, which was larger than 10 ms, confirmed the internal validity of the study to detect prolongation of the QTc interval. As implied in the International Conference on Harmonization (ICH) E14 guidance,2 the lower bound of 95% confidence intervals for the mean moxifloxacin effect is necessary to validate the assay sensitivity, but it was not reported in this article. Based on central limit theorem, the 95% confidence interval for the mean ΔΔQTc (difference of QTc values between time-matched, baseline-subtracted drug and placebo) of a large sample can be calculated with the following formula: Upper bound of the 95% one-sided confidence interval = mean Lower bound of the 95% one-sided confidence interval = mean where n is the sample size, SD is the standard deviation of ΔΔQTc, t = mean ΔΔQTc/(SD√n), and tn – 1.0.95 is the t value corresponding to a cumulative area of 95% of the t distribution with n — 1 degrees of freedom. Thus, for the upper bound of the 95% confidence interval to be larger than 10 ms, one could simply have a smaller sample or make the QTc measurements more variable. Clearly, an upper bound of the 95% confidence interval above 10 ms for the mean ΔΔQTc of moxifloxacin does not confirm the sensitivity of the study to detect an increase in QT/QTc. Furthermore, the advantages and disadvantages of testing moxifloxacin with a parallel design while testing liraglutide with a crossover design were not discussed. The variance of ΔΔQTc is the summation of the variance of the baseline-adjusted QTc values (ΔQTc) for drug and the ΔQTc for placebo (ie, the variance of the procedure), corrected by the covariance between placebo and drug treatments. A concurrent positive control is used to show whether the contribution of the variance from the procedure to the variance of the mean ΔΔQTc is sufficiently small for detection of a small drug effect. Retrospective examination of the data from this study shows that the variance of baseline-adjusted, individually corrected QTc values (ΔQTci) for placebo was 72% larger during the crossover period than during the parallel treatment period. Since the correlation of the ΔQTci values between placebo and liraglutide appeared positive and the sample size in the crossover period was larger as compared with the parallel period, the contribution from the procedure to the variance of the mean liraglutide effect was likely small. However, for the purpose of establishing assay sensitivity, choosing a study design for the positive control that differs from the design for the investigational drug should be carefully justified before the initiation of the study. It is worth noting that the positive control in a “thorough QT/QTc study” has limited value if one wishes to decrease the chance of false negatives (ie, a drug that truly prolongs the QT/QTc interval is not found to prolong QT/QTc in a thorough QT/QTc study). Table I summarizes several different scenarios where the positive control is used. If the upper bound of the 95% confidence interval for the mean test drug effect is less than 10 ms, the contribution to variance of the mean test drug effect from the procedure must have been made sufficiently small; thus, the information from the positive control does not add much. If, on the other end, the upper bound of the 95% one-sided confidence interval for the mean test drug effect is larger than 10 ms, additional QT evaluation will be necessary, which may be either a larger thorough QT/QTc study or an expanded electrocardiogram evaluation during later stages of drug development. In summary, for a thorough QT/QTc study where a positive control is used, the lower bound of the 95% confidence interval of the mean QTc effect of the positive control should be reported. Furthermore, without prior knowledge of the variance of drugs and the placebo and the covariance between them, the positive control should be tested in a way that is similar to that for the investigational drug. Nevertheless, the value of the positive control in a negative thorough QT study is limited.