Abstract
Abstract
Introduction: The aim of a sample size calculation is to enrol sufficient participants to detect a clinically relevant treatment effect, without including so many patients that resources are wasted. If the sample size is too small significant differences can be missed and a too large sample size would be unethical
... read more
and a waste of money and time. Optimizing the sample size is therefore a part of good clinical practice.
A sample size calculation is based on a priori values for nuisance parameters such as the control event rate or the standard deviation in the control group. Since these a priori assumptions greatly affect sample size calculations, it’s important that the used values are reliable In paediatric clinical trials, little is known about the agreement between the a priori assumptions and the realized values after completion of the trial. Therefore this thesis focuses on the problems of sample size calculation in paediatric trials.
Methods: PubMed was searched for articles published between the first of January 2006 and the first of October 2010. For each included trial we searched whether there was a registration or not.
For each article that provided the required characteristics the sample size was recalculated and the two values were compared. The agreement between the a priori estimates and the realized values after completion of the trial were assessed. Furthermore the anticipated effect size was compared to the realized effect size and the theoretically possible minimal significant effect size.
Results: Of all 52 articles, 16 (31%) was registered in an online database.
Twenty-seven (82%) articles with a dichotomous outcome variable reported all parameters required for a sample size calculation. Only nine (47%) articles with a continuous outcome variable reported all necessary parameters for sample size calculation. The articles that didn’t mention all parameters all failed to mention at least an assumption of the standard deviation.
Discusssion: From our results we conclude that the journals not belonging to the ICMJE member journals didn’t all follow the demand to researchers to register their trials in front of the study is started. Therefore many trials are still not registered. We believe it should be considered to create one international free accessible database, where all trials should be registered in a standardized way.
Sample sizes are frequently completely missing or based on the wrong assumptions. We believe a part of the problems could be solved through more collaboration between researchers and statisticians, because there seems to be a shortage of statistical knowledge.
show less