Jixian Wang and Kai Grosch
Novartis Pharma AG, Basel, Switzerland
Objectives: To examine confounding biases in the estimation of dose-PK and PKPD relationships in randomized concentration controlled (RCC) trials with different design and analysis approaches using.
Methods: We propose using instrumental variables (IV) [1, 2, 3] for estimation of parameters in PKPD and dose-exposure models to eliminate or reduce confounding biases for RCC trials. Performance of a number of approaches including the IV approach with different designs for RCC trials were examined from theoretical aspects and via simulations. An approach to detect confounding bias based the IV approach in combination with bootstrap was also proposed.
Results: Simulations showed that in general the IV approach can eliminate confounding bias for both RCC trials but it is less robust to confounded treatment heterogeneity. A good trial design of is the key to ensure high performance of the IV approach. With the bootstrap approach, confounding bias can also be detected even when the number of exposure ranges is very low (e.g. 2). The IV approach is less efficient than the simultaneous modeling approach but it is based on much weaker assumptions, hence provides a robust alternative to it. The IV estimate for the dose-proportionality parameter had almost no bias when the trial is well designed.
Conclusion: Using randomized exposure range as IV can eliminate the confounding bias for a well-designed RCC trial. Issues and approaches for causal effect determination with response dependent dose adjustment [4] should be examined carefully when using RCC trials to analyze PKPD relationships.
References:
[1] Nedelman, J.R. (2005) On some "disadvantages" of the population approach. The AAPS Journal 7, Article 38.
[2] Wang J. (2012) Dose as instrumental variable in exposure–safety analysis using count models, J. Biopharma. Stat., 22: 565-581.
[3] Wang J. Determining causal effects in exposure-response relationships with randomized concentration controlled trials. J. Biopharma. Stat., tentatively accepted.
[4] Wang, J. (2013) Causal effect estimation and dose adjustment in exposure-response relationship analysis. Developments in Statistical Evaluation in Clinical Trials. van Montfort et al (Edit). Springer, New York.
Reference: PAGE 22 () Abstr 2949 [www.page-meeting.org/?abstract=2949]
Poster: Estimation methods