Model-based meta-analysis (MBMA) is a new type of quantitative analysis method, random effects in statistics will be found and put into the model as well as covariates. The final model has a strong predictive ability, can be conducted to make clinical decisions and evaluation. For its complexity, MBMA study is difficult to perform which restricted its application. By summarizing current published MBMA studies as well as the authors' experience, this article introduced the principle, method and status of MBMA.
Despite the rapid development of meta-analysis technology, there were currently no consolidation technology for longitudinal data. The meta-analysis model based on the generalized linear mixed-effects model can fully encapsulate the correlation between various time points and accurately estimate the final combined effect, which is an ideal model for longitudinal-data meta-analysis. Through example data, this paper used SAS software to realize longitudinal-data meta-analysis and provided programming codes.
Covariates are factors which have significant impacts on the primary analysis prior to the subjects being treated. Covariates adjustment should be considered in the design and analysis stages of the clinical trial. Through controlling in the design stage is the optimal resolution; randomization, stratified randomization and restricted covariant values could be used to balance the covariates between comparative treatments. During data analysis stage, analysis of covariance, stratified analysis, linear or generalized linear model can be conducted for covariate adjustment according to different types of outcome and covariate. For confirmatory clinical trial, covariates should be defined in advance in the protocol and statistical analysis plan with the main statistical model.
Randomized controlled trial has been the "gold standard" for clinical trials, in which randomization serves as a fundamental principle of clinical trials and plays an important role in balancing covariates. The allocation probability in traditional design is fixed, while that in adaptive randomization can alter during the experiment according to the specified plan to achieve the purposes of balancing the sample size, maximizing the benefit of patient, or balancing covariates etc. In this study, the adaptive randomization methods applied in clinical trials are discussed to explore their advantages and disadvantages for providing reference for the randomization of clinical trials.
This paper introduced the fundamental theory, method advantages, application scenario and R software implementation method of the covariate-adjusted receiver operating characteristic (ROC) curve. Compared with the traditional univariate ROC curve, the covariate-adjusted ROC curve has distinct methodological advantages and wider application scenarios, which can help to evaluate the ability of markers to predict the targeted outcome more scientifically. It merits more widespread and prior adoption in practical research.