Why Concordance index is increasing a lot when adding a non-significant variable?

Question

I'm testing some models using the Cox regression fitter and comparing concordance index between each other.

I can't quite understand why, adding one single variable that is not even significant (p-value > 0.005) in a model that has already 5 (all significant), the concordance index increases from 0.54 to 0.61?

Thanks a lot for your help (and sorry in advance if my question is silly for a lot of you).

Stefano