Robust QT prolongation assessment using model-averaging
Anne-Gaëlle Dosne (1), Martin Bergstrand (1), Mats Karlsson (1), Didier Renard (2), Günter Heimann (2)
(1) Uppsala University, Sweden, (2) Novartis Pharma AG, Basel, Switzerland
Objectives: TQT studies are pivotal safety studies which assess whether a drug prolongs the QT interval by 10 ms or more. Model-based estimation of the drug-induced QT prolongation at the estimated mean maximum drug concentration could increase efficiency over the currently used intersection-union test. However, robustness against model misspecification needs to be guaranteed in pivotal settings. The objective of this work was to develop an efficient, fully pre-specified model-based inference method for thorough QT studies where type I error is controlled.
Methods: The proposed estimator of the concentration-response relationship consisted of the weighted average of a parametric (linear) and a nonparametric (monotonic I-splines) estimator. Three alternatives were tested for estimating the weight of each estimator, based on the global Mean Integrated Square Error (MISE, as adapted from [1]), on the local MISE ([1]) or on the Bayesian Information Criteria (BIC). TQT studies were simulated to assess the performance of the methods under 24 scenarios with varying drug effect models (linear, Emax, sigmoid Emax and quadratic) and noise levels (sd of QTc 3.5-15 ms, 50 or 100 IDs).
Results: Model-averaging using global MISE weights was found to be an adequate method for TQT analysis. The proportion of studies wrongly concluding to the absence of QT prolongation was below 5% in all but 2 scenarios. Bias in estimated QT prolongation was small (+0.33 ms on average) and conservative 83% of the time under a true drug effect of 10 ms. Relative increases in power compared to the nonparametric method were 30% on average, while decreases in power compared to the parametric method were mostly below 10%.
Conclusions: An efficient, fully pre-specified model-based inference method for TQT studies where type I error is controlled was developed. This methodology could also easily be applied to QT assessment outside of TQT studies, for example in early phase I studies.
Acknowledgements: This work was supported by the DDMoRe (www.ddmore.eu) project and the FP7-HEALTH-2013-602552.
References:
[1] Yuan Y, Yin G, Dose-Response Curve Estimation: A Semiparametric Mixture Approach. Biometrics, 2011. 67: p. 1543–1554.
[2] Ramsay JO, Monotone Regression Splines in Action. Statistical Science, 1988. 4(5):p. 421-465.
