2018 Volume 18 Pages 1-9
Adverse effects induced by the duplication of drugs with anticholinergic effects are a problem among elderly people who take many drugs. Various anticholinergic rating scales have been published and are applied clinically to evaluate a patient’s anticholinergic burden; however, there are some problems with these scales, such as drugs that are assessed differ ently between scales. We aimed to construct a method to more correctly distinguish between drugs with and without anticholinergic effects and to understand the properties of drugs that have anticholinergic effects. We constructed a model for identifying anticholinergic effects via a decision tree, using descriptors indicating the physicochemical properties of the drugs. The best split yielded a decision tree with 46 branches (area under the receiver operating charac teristic curve = 0.99). However, only seven branches, defined by six descriptors: ASA_P, GCUT_PEOE_0, opr_brigid, PEOE_VSA+1, GCUT_SLOGP_0, vsa_pol (related to van der Waals surface areas, partial charges, and molecule structures), were required to identify drugs with anticholinergic effects. This result suggests a relationship between the hydropho bic interactions of drugs and the muscarinic receptor. In this study, we constructed a model to predict whether drugs have anticholinergic effects, and obtained essential physicochemical information on the drugs to distinguish their anticholinergic effects. It is our hope that these findings provide useful information for predicting anticholinergic effects of drugs in clinical settings.
