That would apply if EACH engine had its own individual probability specified. Then the chance would become the sum of the two probalities eg Engine 'A' has a 1:1 000 000 chance of failure, engine 'B' has a 1.5:1 000 000 then overall probability would be 2.5:1 000 000.
However, this is not the case. Within the probability for the FLEET, some fail early, some fail much later but the overall probability remains consistant. This is Ignoring trends from a spate of failures or unusually good reliability. Failure probability is quoted on fleetwide basis which is largely independent of engine location. It's a bell curve that considers ALL engines in operation. A statistic that applies to a population is not capable of supplying a prediction for any particular member of the population.
For any specific airframe, that airframe could have an increased probability of failure IF it is unfortunate enough to be fitted with engines that are each on the unreliable side of the bell curve. However, the airframe could equally have a much reduced probability of failure if it ends up fitted with engines that happen to be all on the more reliable side of the curve.
Changing the number of engines on the airframe doesn't affect this **overall** probability and so that is why there is this seemingly weird 'number of engines doesn't affect reliability' situation. Keep in mind that adding more engines could result in the airframe getting a batch of 'super reliable' ones then the chance of a failure would reduce. Until it received a batch of 'super unreliable' ones when the chance of a failure would increase. Over time, and over a large number of installations, the 'average' chance of failure would tend towards the fleet average.
Oh, why did I get involved in this discussion?? 
But it is a great debate and I have been following responses very closely, with interest. 
Tins is on the money! In a theoretical world, if the demonstrated failure rate were < 1:100,000, when the fleet total exceeds 100,000 hours, it is really irrelevent whether the engine which will fail is on a single engine or multi engine aircraft. For that reason, there should be no greater risk of engine failure on a S/E or M/E aircraft.
However, I accept if one has a four engine aircraft, the risk may be higher that the engine which is going to fail will be on that four engine aircraft and not on the S/E aircraft.
Now I'm completely muddled!
But I still suggest the failed engine is not relevent - it's how many you have left after the failure that counts. Unless of course, the aircraft is a BN2A MKIII Trislander, in which case the two remaining are guaranteed to take you to the scene of the crash!
Woomera