Analysis of the methodological points and practice status of matching-adjusted indirect comparison application
Objective To provide methodological guidance for the application of matching-adjusted indirect comparison(MAIC).Methods The methodology literature on MAIC was examined to clarify key steps and methodological points,and MAIC application literature in the non-small cell lung cancer field published after January 2016 was systematically reviewed to compare and analyze the current status and problems of MAIC.Results MAIC consisted of five key steps:data source and sample selection,matching variable screening,individual weight calculation,matching validity evaluation,and relative efficacy calculation.The systematic review revealed that studies primarily employed literature reviews to screen data sources,used statistical analysis and other scientific methods to screen matching variables,employed software for individual weight calculation,evaluated matching validity by reporting effective sample size(ESS),calculated relative efficacy using Cox,logistic,and other models,conducted sensitivity analyses to evaluate the uncertainty caused by different data sources and matching variable combinations,and the studies demonstrated good overall reporting standardization but significant differences in particular aspects.Concerning the connection between MAIC and pharmacoeconomic research,studies included mainly used target drugs as the reference group of survival data extrapolation,and proportional hazards(PH)assumptions were considered when utilizing hazard ratios(HR)in extrapolation.Conclusion There are some deficiencies in the method application and reporting standards of MAIC research,such as lack of explanation of data source selection criteria and matching variable screening criteria,insufficient reporting of weight distribution,and inadequate consideration of PH assumptions.It is recommended that future MAIC research systematically screen data sources and report covariate distribution evaluation,covariate status evaluation,weight distribution,uncertainty measurement,etc.Additionally,considering PH assumptions after calculating HR is suggested.
NETL
NSTL
万方数据
