Use of continuous exposure variables when examining dose-dependent pharmacological effects – Application to the association between exposure to higher statin doses and the incidence of diabetes