It's the timing to use the meta-analysis when the reviewed researches have poor cumulation and when we would like to combine some researches with small effects. The limitations of meta-analysis are publication bias, selection bias, the missing original data of primary studies, the differences in when and where the trials were conducted, the different relative factors among studies, and the problems of insufficient reviewed articles or total samples. There are some statistic methods of integrating the data. They were used to calculate the effect size and variance. The main directions are fixed effect model and random effect model.