I wanna use some variables (T-stage, M-stage, N-stage, Staging) in the Cox proportional hazard (or Cox regression) to investigate their impact on survival of colorectal cancer patients.

These variables are categorical and I know they are similar together. When I use univariate analysis, they are significant. But when I use multivariate analysis, they are not significant. and the standard error is big.

Do you know a method which handle this problem. I wanna investigate the impact of these variable on survival, simultaneously.