The Quantitative Evaluation Of Clinical Trials Of Bias Source Method Research

Background:The basic requirement of Evidence Based Medicine (EBM) was that results from Randomized Controlled Trial (RCT) must be validated before the extrapolation to clinical practice. Instead of the rigorous and scientific approach, most strength of evidence evaluation tool was established just based on intuition.Objective:To generate an evaluation method to assess the level of evidence in RCT. Monte Carlo simulation method was chosen to figure out the source of bias and the accompany confounding effects.Method:The simulation model was integrated by3sub-models:1, the dose-effect model used for fit the treatment-outcome patterns (the dose-effect model was based on published papers and had a mathematical algorithm to transform the treatment effect to clinical endpoint).2, the source of bias sub-model was used to generate the confounding factors. The source of bias might be divided into two categories. One is patient oriented and the other is investigator oriented. The source of bias includes placebo or nocebo effect, measurement error, effect of concomitant treatment, the regression to mean phenomenon, lost to follow-up and the randomization process.3, sub-model used to describe the distribution of baseline covariates.Results: The relative difference between true (without bias) and observed (with bias) odds ratio was treated as the measurement of trial validity. The measurement error, placebo effect and concomitant treatment had the greatest influence on study result. Randomization was correlated to the above factors. And the randomized process also had the moderated effect to the validity. The class of importance of each bias source was robust in the sensitivity analysis even the critical parameters were changed.Conclusion:Diverse source of bias could influence the result in a randomized control clinical trial. Different source of bias had their own confounding effect and the consequences could not be ignored. The available evaluation tools were commonly focused on the randomization, allocation concealment and adequate blinding. Some important biases were missed. There is the need and essentiality to develop innovative methods to assess the quality of evidence in a more appropriate manner.