Lung cancer : journal of the International Association for the Study of Lung Cancer
-
Review Comparative Study
Pitfalls and biases in the reporting and interpretation of the results of clinical trials.
There are numerous biases (not all statistical) that can occur in the reporting and interpretation of the results of clinical trials. In this paper we review some of the major sources of such biases and propose some solutions for these problems. ⋯ We also discuss the correct interpretation of a P-value and show the need for both estimates of the treatment effect and confidence intervals for reliable interpretation of the results. Finally, we consider the important but less well discussed problems of the interpretation of both classically 'negative' and 'positive' trials and the impact of the early stopping of the trial on the estimate of the treatment effect.
-
If meta-analyses of randomized trials were better in quality and were correctly interpreted, then there would be far less reasons for criticisms. The large number of meta-analyses to date are ample evidence of the usefulness of this technique in giving a synoptic appraisal of apparently discordant trials, but their shortcomings should be known. The first step needed to optimize the quality of meta-analyses is to improve the quality of randomized clinical trials. ⋯ Meta-analyses based on individual data provided by each trialist allow a better appreciation of the quality of a trial, an increase of the statistical power, and for some covariates to be taken into account; such is not generally the case for those based on published data. Trialists should collect and conserve relevant data on ongoing trials and ensure that these trials are registered. These data would serve for the conduct of future unbiased meta-analyses, the duration of which would be reduced considerably, and unnecessary trials would be avoided.