Reader Comments

Post a new comment on this article

The sample size is inadequate

Posted by janakadesilva on 22 Aug 2009 at 12:14 GMT

We read with interest the paper by Eddleston and colleagues reporting the RCT conducted to assess the effectiveness of pralidoxime in acute organophosphorus (OP) poisoning. The study has been meticulously planned and the paper well written. However, we are concerned that these positive features mask a major flaw in the study – inadequate sample size. By the authors’ calculations the trial required a total of 1500 patients, but the number of patients finally recruited was 235. The observed event rate in the trial for the primary end point was similar to the rates used in the sample size calculation.

The conclusions in the main paper and the abstract are contradictory. The authors conclude the abstract by stating, “Despite clear reactivation of red cell acetylcholinesterase in diethyl organophosphorus pesticide poisoned patients, we found no evidence that this regimen improves survival or reduces need for intubation in patients with organophosphorus insecticide poisoning.” This lack of evidence is perfectly understandable given the inadequate number of subjects recruited for the trial. The trial had a power of only 16% (going up to 25% for a sided significance test) if the pre-trial assumptions about the event rates were correct. But the authors seem to be puzzled by these results as indicated by the next sentence in the abstract, “The reason for this failure to benefit patients was not apparent.” This is an example of missinterpreting ‘negative’ trial results – equating absence of evidence to evidence of absence. The authors conclude the paper by stating “Our trial provides evidence that routinely following the WHO recommended high-dose pralidoxime regimen in all patients does not improve survival in OP insecticide self-poisoned patients.” This claim is unfounded given the power of their study.

We think it is also important to address the wider issue raised by studies such as this; that is, the issue of inadequate recruitment in clinical trials. There are times, as in the present trial, when the achieved sample size is far short of the calculated sample size due to reasons beyond the researchers’ control. In such situations the data cannot and should not be used to answer the original research question unless there is a major discrepancy between the assumptions in the original sample size calculation and the collected data. If not, the whole purpose of sample size calculation is lost.

H J de Silva and A Pathmeswaran*, Departments of Medicine and *Public Health, Faculty of Medicine, University of Kelaniya, P O Box 6, Ragama, Sri Lanka

No competing interests declared.