First we review standard ways to use Girsanov theorem to reduce variance in option pricing. Then we introduce a new adaptative variance reduction method. We show how this method can be efficiently implemented for very general payoff and we discuss extensions of the method.