Blinded Sample Size Recalculation in Noninferiority Trials: A Case Study in Dermatology
- PDF / 10,200,487 Bytes
- 9 Pages / 612 x 792 pts (letter) Page_size
- 68 Downloads / 195 Views
599
Blinded Sample Size Recalculation in Noninferiority Trials: A Case Study in Dermatology
Tim Frirdr, PhD University Medical Center Gottingen. Department of Medical Statistics, Gdttingen. Germany
Holgrr Stommrr, MSc Phannalog Institut fiir klinische Fonchung GmbH, Munich, Germany
Key Words Sample size recalculation; Noninferiority test; Clinical trial Corrrspomdrmtr Address Dr. Tim Friede. Department of Medical Statistics. University Medical Center Giittingen, 37073 Gottingen, Germany (email: tim.friede@med. uni-goettingen.de).
When designing a dim& trial, a number of design faaturcs induding the sample size have to be decided u p . nte smnplt size dadation lLmally nquircs some discusdon ofrelevant ef' fed sizes and infarnation abact mdsmra pa?umetmsuhasstandmddcviationsQrovcIyIu event mtm, with nuisancepmamdasbeing estimated jkom prwioup studics. using a novd
endpoint or moving into a new indication, no or ady vay limited infonnation might be availabIe and thc sample size &adation is thercfoe subject to CanSidaaMc whrtainty. f n t d pilot study dm@s that allow sample size m-
INTRODUCTION When designing a clinical trial, a number of features including inclusion and exclusion criteria, and primary and secondary endpoints have to be decided upon, and calculation of the sample size plays an integral part in this planning process. Whereas the significance level and the desired power are usually set to some standard values, such as a two-sided significance level of 5% and a power of 80%or 90%,it usually requires some discussion as to what a clinically relevant effect is and what are likely values of the nuisance parameters (eg, standard deviations, overall event rate). The latter are often informed by estimates from previous studies, with data available either in-house or from published summaries. However, when using novel endpoints or moving into new indications, such data or published information might not be available. In this article we report design considerations for a recently completed randomized, active controlled trial in dermatology where a modified version of an established scale was used in a new indication, and no prior information on the size of the standard deviation was available. To make the trial robust against possible underestimation of the standard deviation in the planning phase, which in
timation midcourse of the ongoing study haw been propas& to make trials mow robust to misspxi~cutions ofrmisanccparametm in the pkanring phase. In this mtide we present the design ofa d y completed randomized active contrdlad trial in dennatdogy as a case study. Furthennore, we demonstrate how type 1 C~TOTrate contrd can be achieved when testing for noninjkiotity and explore operating chmactuisficssudt as power and sample size distributions thraqgh simulations motivated by the case study. Finally, relevant regulatory guidefi~onsampZesiecn?estimationarerefewedto.
turn would lead to an underpowered trial, the study design included an internal pilot study (1)and a blinded sample size review (2,3). In the
Data Loading...
