Gravar-mail: Potential Bias in Medication Adherence Studies of Prevalent Users