Limitation
Our meta-analysis has several limitations. First, the thresholds for defining polypharmacy status differed in each study, and subjects with different baseline characteristics may have significant bias despite statistical adjustments. Second, the included studies only assessed the extent of patients’ polypharmacy at baseline and were unable to adjust drug use for subsequent prescription changes. Fortunately, the burden of co-morbidity in polypharmacy populations is often chronic, and the number of drugs rarely changes significantly. In addition, the study did not take into account non-prescription drugs and health products that patients were taking, which could lead to a potential influence on the results. Third, drug compliance largely determines how patients actually take their medication. Unfortunately, the frequent dose changes of warfarin prevented the study from further evaluating the compliance of the two classes of oral anticoagulants. Finally, due to the limited amount of data, we were unable to perform head-to-head comparisons among NOACs, and more studies in the future may reveal which NOACs would be more suitable for this population.