Statistical Method for Clinical Data with Missing Observations: a Case Study on a Rheumatoid Arthritis Case

概要

論文の詳細を見る
The objective of this paper is to evaluate statistical methods for analyzing clinical data with missing observations based on a real clinical trial of rheumatoid arthritis. Since the type of drop-out subjects in this case is "missing at random" and the drop-out proportion greatly varies among treatment groups, suitable adjustments are necessary in the statistical analysis to reduce bias in the estimate of drug effects due to the missing subjects. Three methods are compared using the Monte Carlo simulation: COMP method which uses only the complete case of subjects, the LOCF method which uses the last observation of each subject and the IPCW method which uses parameter estimates with inverse probability of censoring weighted. Although the real case is a three-arm trial with unequal sample sizes, a two-arm trial with equal sample sizes is assumed as the framework of simulation. The primary variable for efficacy evaluation is assumed to be the ACR20 which is recommended by the American College of Rheumatism and various parameters in the simulation are set as adaptable for the real case. The simulation revealed that the LOCF method is the best among the three methods to improve the bias and precision in the estimate of drug effects. The parameter estimates were reviewed using the LOCF method based on the above validation. As a result, the conclusion, using the COMP method derived from the trial, that the investigational drug is effective to improve symptoms of rheumatoid arthritis is enhanced by these estimates.

Statistical Method for Clinical Data with Missing Observations: a Case Study on a Rheumatoid Arthritis Case

スポンサーリンク

概要

著者

関連論文

スポンサーリンク