In most instances of blinded sample size re-estimation, the timing of the interim analysis that determines whether the sample should be increased is preplanned. Yet such an approach is not necessarily the best for a sponsor. There could be benefits to continuously monitoring a clinical trial to see if a sample size re-estimation is necessary. In such cases, what are the costs to the sponsor in terms of power and sample size? New simulations by Cytel’s Martin Kappler and Ursula Garczarek assess the costs and consequences of this approach.

If one could limit the costs of continuous monitoring, sponsors could rely on even greater flexibility through blinded sample size re-estimations. While traditionally a blinded SSR was used in response to the overwhelming numbers of inadequately powered trials, more recently they have been deployed for strategic investing or ensuring that patients are channeled towards therapies that are most likely to succeed if adequately powered. There is a more fundamental use for such designs, and that is to overcome common statistical uncertainties like the misspecification of parameters in the control arm.

Sample sizes are sometimes determined with wrong assumptions on standard deviations. This can be revealed during the progress of a clinical trial. Therefore, an interim look which considers a blinded sample size re-estimation could solve for this difficulty. Yet this leaves sponsors in a situation where they must guess when to plan the perfect interim look. The continuous monitoring approach avoids this guessing game.

The benefits of continuous monitoring are added flexibility. The cost can take the form of impact on Type I error, power and increases to average sample size. A new poster by Kappler and Garczarek, to be presented at this year’s PSI, reveals the full extent of this cost.

Abstract

Authors: Martin Kappler, Ursula Garczarek, Cytel Inc.

Blinded sample size re-estimation (bSSR) is applied in a lot of clinical trials as an insurance against misspecifications of nuisance parameters (such as standard deviation, or event or hazard rate in the control group) during the determination of sample size. The type-1 error control of bSSR has already been studied for several situations and endpoints (e.g., Kieser 2003, Friede 2013). However, to our knowledge interim analyses for the bSSR are always assumed to happen at a preplanned timepoint while in fact a continuous blinded monitoring of the nuisance parameter is possible throughout the trial and often is done by the clinical study team. Therefore, the bSSR calculations could be done in regular intervals regardless of a pre-planned bSSR. When the need for a sample size increase or decrease becomes obvious, a protocol amendment could be prepared to either introduce a previously unplanned bSSR or to prepone a planned bSSR. As this additional flexibility may have detrimental effects to operational characteristics, this approach is simulated for a continuous endpoint in several settings for trials with and without bSSR to assess the impact on type-1 error, power and average sample size.

References
Friede T, Kieser M. Blinded sample size re-estimation in superiority and noninferiority trials: bias versus variance in variance estimation. Pharm Stat. 2013;12(3):141-6

Kieser M, Friede T. Simple procedures for blinded sample size adjustment that do not affect the type I error rate. Stat Med. 2003; 22(23):3571-81

Share Via