Proof that different integral prescriptions lead to different scattering matrices?

So I have read that if you calculate feynman diagrams with an $iepsilon$ prescription, that is with propagators of the form $frac{i}{p^2-m^2+iepsilon}$ then you get the standard $mathcal{iM}$ scattering matrix however if instead you use a PV prescription then you get the $mathcal{K}$-matrix, related to $mathcal{M}$ by $mathcal{M}=frac{1}{mathcal{K}^{-1}-irho}$. is this true, I cant seem to find a proof of this?