Hi, everyone! I used limma package in R to find different expression genes in a TCGA dataset, and then I used the univariate cox regression to find different expressed genes releated to the overall survival, using the P value<0.05, I got a gene list, then I used the multivariate cox regression on the gene list and find out 5 genes with P value<0.05. And then something weird happened. One of these 5 genes is a risk factor with HR>1 in multivariate cox regression, but it is a protective factor in univariate cox regression with HR<1,how could this happen? Which one is correct?