论文部分内容阅读
Randomised controlled trials(RCTs) are gold standard in the evaluation of treatment efficacy in medical investigations, only if well designed and implemented. Till date, distorted views and misapplications of statistical procedures involved in RCTs are still in practice. Hence, clarification of concepts and acceptable practices related to certain statistical issues involved in the design, conduct and reporting of randomised controlled trials is needed. This narrative synthesis aimed at providing succinct but clear information on the concepts and practices of selected statistical issues in RCTs to inform correct applications. The use of tests of significance is no longer acceptable as means to compare baseline similarity between treatment groups and in determining which covariate(s) should be included in the model for adjustment. Distribution of baseline attributes simply presented in tabular form is however, rather preferred. Regarding covariate selection, such approach that makes use of information on the degree of correlation between the covariate(s) and the outcome variable is more in tandem with statistical principle(s) than that based on tests of significance. Stratification and minimisation are not alternatives to covariate adjusted analysis; in fact they establish the need for one. Intention-totreat is the preferred approach for the evaluation of primary outcome measures and researchers have responsibility to report whether or not the procedure was followed. A major use of results from subgroup analysis is to generate hypothesis for future clinical trials. Since RCTs are gold standard in the comparison of medical interventions, researchers cannot afford the practices of distorted allocation or statistical procedures in this all important experimental design method.
Randomized controlled trials (RCTs) are gold standard in the evaluation of treatment efficacy in medical investigations, only if well designed and implemented. Till date, distorted views and misapplications of statistical procedures involved in RCTs are still in practice. Hence, clarification of concepts and acceptable practices related to certain statistical issues involved in the design, conduct and reporting of randomized controlled trials is needed. This narrative synthesis aimed at providing succinct but clear information on the concepts and practices of selected statistical issues in RCTs to inform correct applications. The use of tests of significance is no longer acceptable as means to compare baseline similarity between treatment groups and in determining which covariate (s) should be included in the model for adjustment. Distribution of baseline attributes simply presented in tabular form is however, rather preferred. Regarding covariate selection, such approach that makes use of information on the degree of correlation between the covariate (s) and the outcome variable is more in tandem with statistical principle (s) than that based on tests of significance. Stratification and minimization are not alternatives to covariate adjusted analysis; in fact they establish the need for one. Intention-totreat is the preferred approach for the evaluation of primary outcome measures and researchers have responsibility to report whether or not the procedure was followed. A major use of results from subgroup analysis is to generate hypothesis for future clinical trials. Since RCTs are gold standard in the comparison of medical interventions, researchers can not afford the practices of distorted allocation or statistical procedures in this all important experimental design method.