Gravar-mail: Implications of the Propensity Score Matching Paradox in Pharmacoepidemiology