There are several options. You could read the formula from the ROBDD (here ¬a∧¬c ∨ a∧b∧c ∨ a∧¬b∧¬c), and draw a new ROBDD of it from scratch. Or you can reuse the part of the ROBDD where the changed variable order makes no difference (the low-branch of the a-root), and just re-draw the (sub-)ROBDD for a = true. As this (sub-)ROBDD just represents a↔b, only the b- and c-nodes flip their labels.