Blinded and unblinded sample size reestimation in crossover trials balanced for period

Abstract

The determination of the sample size required by a crossover trial typically depends on the specification of one or more variance components. Uncertainty about the value of these parameters at the design stage means that there is often a risk a trial may be under- or overpowered. For many study designs, this problem has been addressed by considering adaptive design methodology that allows for the re-estimation of the required sample size during a trial. Here, we propose and compare several approaches for this in multitreatment crossover trials. Specifically, regulators favor reestimation procedures to maintain the blinding of the treatment allocations. We therefore develop blinded estimators for the within and between person variances, following simple or block randomization. We demonstrate that, provided an equal number of patients are allocated to sequences that are balanced for period, the proposed estimators following block randomization are unbiased. We further provide a formula for the bias of the estimators following simple randomization. The performance of these procedures, along with that of an unblinded approach, is then examined utilizing three motivating examples, including one based on a recently completed four-treatment four-period crossover trial. Simulation results show that the performance of the proposed blinded procedures is in many cases similar to that of the unblinded approach, and thus they are an attractive alternative.

Keywords: blinded; crossover trial; internal pilot study; sample size reestimation.

© 2018 The Authors. Biometrical Journal Published by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Figures

Figure 1

The distribution of σ^e2 is shown for each of the reestimation procedures for several values of τ, and several values of nint, for Example 1. Precisely, for each scenario, the median, lower, and upper quartile values of σ^e2 across the simulations are given. The dashed line indicates the true value of σe2

Figure 2

The distribution of N^ is shown for each of the reestimation procedures for several values of τ, and several values of nint, for Example 1. Precisely, for each scenario, the median, lower, and upper quartile values of N^ across the simulations are given. The dashed line indicates the true required value of N

Figure 3

The simulated familywise error‐rate (FWER) is shown under the global null hypothesis for each of the reestimation procedures when nint∈{16,32}, as a function of the within person variance σe2, for Example 1. The Monte Carlo error is approximately 0.0007 in each instance. The dashed line indicates the desired value of the FWER

Figure 4

The simulated power is shown under the global alternative hypothesis for each of the reestimation procedures when nint∈{16,32}, as a function of the within person variance σe2, for Example 1. The Monte Carlo error is approximately 0.0013 in each instance. The dashed line indicates the desired value of the power

Figure 5

The simulated familywise error‐rate (FWER) is shown under the global null hypothesis for each of the reestimation procedures when nint∈{16,32}, as a function of the clinically relevant difference δ, for Example 1. The Monte Carlo error is approximately 0.0007 in each instance. The dashed line indicates the desired value of the FWER

Figure 6

The simulated power is shown under the global alternative hypothesis for each of the reestimation procedures when nint∈{16,32}, as a function of the clinically relevant difference δ, for Example 1. The Monte Carlo error is approximately 0.0013 in each instance. The dashed line indicates the desired value of the power

Figure 7

The influence of the considered inflation factor upon the power of the re‐estimation procedures under the global alternative hypothesis is shown for several values of nint, for Example 1. The dashed line indicates the desired value of the power