The Problem of Confounding in Studies of the Effect of Maternal Drug Use on Pregnancy Outcome