Randomized exposure-controlled trials; impact of randomization and analysis strategies – from a toxicity perspective

PDF of poster

Objectives: It has previously been shown[1] that the most beneficial parallel group trial design and analysis strategy from a statistical power perspective is a randomized dose-controlled trial (RDCT) and a model-based analysis (MBA) with an independent variable as highly correlated to the clinical endpoint as possible. It has been argued on the basis of a traditional statistical analysis approach that there are situations where a randomized concentration-controlled trials (RCCT) is a better trial design compared to an RDCT[2]. One of the perceived advantages is the lower risk of toxicity. Given that a higher statistical power leads to smaller group size it is not at all clear that an RCCT with traditional analysis will result in fewer toxicity events than an RDCT with model based analysis. The aim of this study was therefore to investigate, through simulations, how the randomization scheme and analysis strategy can influence the number of toxicity events in a clinical trial.

Methods: A simulation framework was set up with a dose leading to a plasma concentration (one-compartment model at steady state), leading to a change in a biomarker (E_max model) and the biomarker leading to a response in a dichotomous clinical endpoint (linear-logistic model). There was also a link between the plasma concentration and a toxicity response (T_max model, adverse event T ≥ 0.75·T_max). Three factors were varied across simulations: TC₅₀, variability in TC₅₀ and variability in CL. Two randomization schemes were simulated; RDCT and RCCT, and two analysis methods were used; group-wise analysis (GWA) and a MBA. In the MBA three independent variables were used; dose, concentration and biomarker. Each combination of randomization scheme and analysis strategy was simulated with a number of subjects that yielded an 80 percent statistical power to detect a treatment effect in the clinical endpoint.

Results: An RDCT with model based analysis using the biomarker typically required less half the number of subjects of an RCCT with GWA. These represent the most and least efficient design/analysis options, respectively. For situations with low variability in CL, the former also results in a lower number of adverse events. As variability in CL increases, the number of adverse events in an RDCT increases relative an RCCT. At higher variability in CL the design/analysis option resulting in the lowest number of adverse events is an RCCT with MBA, about half that of an RCCT with GWA.

Conclusions: This demonstrates that under reasonable conditions and maintaining equal power across trial designs and analysis options, an RDCT may well result in fewer adverse events than an RCCT. A MBA will result in a higher statistical power and therefore smaller group sizes compared to a GWA, given the same randomization scheme.

References:
[1] Jonsson EN. On the design and analysis strategy for randomized exposure-response trials. 11th EUFEPS Conference on Optimising Drug Development: Integrating New Concepts and Tools, Basel, Switzerland, 2003
[2] Kraiczi H, et al. Randomized concentration-controlled trials: Motivations, use, and limitations. Clin Pharmacol Ther 2003;74:203-214.