I don't think one very important point has been made in previous posts ...
The normally quoted (certification - ie flight manual) Vmca usually is determined with 5 degrees into the live engine(s).
Real world Vmca, however, is STRONGLY bank dependent.
Therefore, while airspeed is low, one ought not to bank away from the live engine lest the real world Vmca increase both rapidly and significantly - this effect can rush up on the unsuspecting pilot and he/she might well find him/herself to be a mere spectator in the ensuing crash (unless thrust be reduced in time to avert inverting, crashing, burning, and dying ...).
Once the speed is up to normal good margins above published certification Vmca the concern is not so significant.
Be aware that the variation in real world Vmca with bank angle generally is neither linear nor insignificant - we are not talking a knot or two here but, perhaps, many tens of knots depending on the bank angle ....
Most sims with which I have played demonstrate the effect reasonably well. A stick and rudder skills exercise I work students up to involves a takeoff liftoff failure/seizure at minimum weight, min speed, and max aft cg. Almost invariably, until the student gets on top of the rapid speed at which things happen (refer John Farley's post, above), the aircraft rolls towards the failed engine during the failure control sequence, Vmca rises rapidly, and the roll accelerates - aircraft goes in inverted. Of course we freeze the sim once the situation gets out of control as there is little, if any, training benefit to be had by letting it go any further. Point is, suggest you don't forget this Vmca - bank angle relationship.