You need more information to eliminate pathological cases such as $T = \text{diag}(1,1,1,....,1,\epsilon)$ and $A = \text{diag}(1,1,1,...,\epsilon^{-1})$. Both matrices have condition number $\epsilon^{-1}$, but \begin{equation} T^{-1}A = \text{diag}(1,1,1,...,1,\epsilon^{-2}) \end{equation} has condition number $\epsilon^{-2}$.