Anzeige
Erschienen in:
01.09.2016 | Research Methods and Statistical Analyses (Y Le Manach, Section Editor)
An Overview of Challenges and Approaches to Minimize Bias in Randomized Controlled Trials in Perioperative Medicine
Erschienen in: Current Anesthesiology Reports | Ausgabe 3/2016
Einloggen, um Zugang zu erhaltenAbstract
Purpose of Review
Randomized controlled trials (RCT) are recognized as the most robust design to study the relationship between exposure and outcomes. The conventional RCT design is commonly used in pharmacological trials. Some surgical interventions are not be well suited to a conventional RCT design and may be associated with methodological challenges. Approaches have been proposed in non-pharmacological trials to overcome some of these challenges and minimize the risk of bias.
Recent Findings
Imbalance in prognostic factors between intervention groups, lack of allocation concealment, unblinding, non-intention-to-treat analysis, and losses to follow-ups can all threaten the validity of RCT results to various degrees. Procedure-based trials raise also specific challenges since physician expertise and training can affect the intervention, exposing to potential differential-expertise bias. Lack of statistical power can also affect the confidence in a trial’s result. Small sample sizes also usually mean small number of events for comparison between interventions, resulting in less statistically robust findings.
Summary
Minimizing risk of bias and achieving adequate statistical power are crucial to producing high quality and meaningful results. Non-pharmacological trials pose certain methodological challenges, and several approaches have been proposed to address the risk of bias. Large sample sizes are also usually required to achieve sufficient statistical power to provide answers to meaningful clinical questions. However, small perioperative trials remain frequent and result interpretation based solely on P values might not always appropriately inform on the confidence in a trial’s results. The Fragility Index can be used to further inform on the confidence of statistically significant result.