Cox modelisation: variable omitted because of collinearity

[eric some]

I am trying to build a Cox model. When I run my initial model, I got this message regarding many variables: "MEA_MI_CANB omitted because of collinearity". At the end I got this one :"flat region resulting in a missing likelihood". Can somebody explains me what it means and how can I solve this issue of collinearity?. Thanks.
Eric

[Nick Cox]

Wrong way round: Stata solved a problem for you by omitting a collinear variable. The follow-up should be that you work out why the variable omitted is collinear. There's no point in trying to force it in.

[eric some]

Dear Nick,
Thanks so much. Very helpful