gms | German Medical Science

MAINZ//2011: 56. GMDS-Jahrestagung und 6. DGEpi-Jahrestagung

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V.
Deutsche Gesellschaft für Epidemiologie e. V.

26. - 29.09.2011 in Mainz

Comparison of sample size reestimation procedures: Bias versus variance in variance estimation

Meeting Abstract

Search Medline for

  • Tim Friede - Universitätsmedizin Göttingen, Göttingen
  • Meinhard Kieser - Universität Heidelberg, Heidelberg

Mainz//2011. 56. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 6. Jahrestagung der Deutschen Gesellschaft für Epidemiologie (DGEpi). Mainz, 26.-29.09.2011. Düsseldorf: German Medical Science GMS Publishing House; 2011. Doc11gmds047

DOI: 10.3205/11gmds047, URN: urn:nbn:de:0183-11gmds0471

Published: September 20, 2011

© 2011 Friede et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.en). You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.


Outline

Text

Kieser and Friede [1] proposed the use of the total variance estimator as an estimator for the within group variance to estimate the sample size in a blinded fashion from data of an internal pilot study [2]. The total variance estimator or one-sample variance is an appropriate estimator under the null hypothesis of no effect but is biased under the alternative by overestimating the within group variance. Xing and Ganju [3] and Ganju and Xing [4] proposed an estimator for blinded variance estimation utilising information about randomisation blocks that is unbiased under any alternative. In this paper we compare the blinded sample size reestimation procedure based on the one-sample variance with the blinded procedure using the randomization block information proposed by Xing and Ganju and a recently proposed estimator by Shih [5]. For the latter procedure the variance in the final analysis is adjusted using the correction term by Miller [6] to achieve a type I error rate close to the nominal level. We show that the variance of the one-sample variance is smaller than the variance of the estimator by Xing and Ganju as long as the standardised treatment difference is smaller than the square root of 2 for randomised blocks of length 2 and give a more general result for arbitrary block lengths. Through simulation studies the advantages of the simpler one-sample variance are demonstrated in terms of power and expected sample size for practically relevant scenarios.


References

1.
Kieser M, Friede T. Simple procedures for blinded sample size adjustment that do not affect the type I error rate. Statistics in Medicine. 2003;22:3571-81.
2.
Wittes J, Brittain E. The role of internal pilot studies in increasing the efficacy of clinical trials. Statistics in Medicine. 1990;9:65–72.
3.
Xing B, Ganju J. A method to estimate the variance of an endpoint from an on-going blinded trial. Statistics in Medicine. 2005;24:1807-14.
4.
Ganju J, Xing B. Re-estimating the sample size of an on-going blinded trial based on the method of randomization block sums. Statistics in Medicine. 2009;28:24-38.
5.
Shih WJ. Two-stage sample size reassessment using perturbed unblinding. Statistics in Biopharmaceutical Research. 2007.
6.
Miller F. Variance estimation in clinical studies with interim sample size reestimation. Biometrics. 2005;61:355-61.