[Statistical methods in the evaluation of the therapeutic efficacy--their problems and solutions]

In order to evaluate the therapeutic efficacy accurately, it is necessary to conduct a well-designed clinical trial and to draw conclusions after considering both the statistical significance and the clinical significance. The well-designed clinical trial needs to meet at least the following three conditions: (1) the acquisition of the minimal sample size to obtain statistically significant results, (2) the existence of an adequate control (standard) treatment group, and (3) the good comparability between treatment groups. In order to secure the good comparability between treatment groups, it is desirable to conduct a stratified randomized controlled clinical trial in the form of a multi-clinic cooperative study in case it is difficult to secure the enough sample size in one institute. The concept and problems of the life table methods which are frequently used in the clinical trial were discussed. The merits and limitations of the multivariate analyses, especially the Cox multiple regression life table method, were also discussed in this paper.