A dynamic model of drug initiation: Implications for treatment and drug control