Trial sequential analysis : Sample size calculation for reliable meta-analyses

Abstract

Meta-analyses have agreat impact on medical decision-making. Random errors are, however, often the reason for misinterpretation of interventional effects in meta-analyses. The aim of this article is to introduce authors and readers of meta-analyses to the problem of random errors. The article presents trial sequential analysis (TSA) as asuitable and user-friendly method that adjusts for the risk of random errors in meta-analyses. The practical application of TSA is illustrated and exemplified using regional anesthesiology procedures versus conventional pain therapy with respect to the prevention of persistent postoperative pain after breast cancer surgery or thoracotomy. The results were compared with those from conventional meta-analytical methods. Conventional meta-analytical methods showed asignificant advantage for patients after breast cancer surgery as well as after thoracotomy for regional anesthesia procedures with respect to the reduction of persistent postoperative pain. By means of TSA it could be concluded for thoracotomy that the evidence of this meta-analysis was sufficient. In contrast, the TSA for breast cancer surgery showed that based on the current data set and on the basis of relevant assumptions, it is potentially a false indication of an effect. There is currently no evidence that regional anesthesia leads to asignificant reduction of persistent postoperative pain. The TSA is asuitable tool to minimize the risk of random errors and for a more reliable assessment of the evidence for the results of ameta-analysis.