"These relations persisted after controlling for confounding variables, such as personal characteristics and other drug use." . . . . .