The duration and type of dual antiplatelet therapy (DAPT) prescribed to patients after percutaneous coronary intervention (PCI) involves carefully balancing reduced ischemia and increased bleeding risk for individual patients. Whereas multiple bleeding risk scores exist, the performance of these models to predict long-term bleeding in the setting of DAPT across different settings and populations is unclear. Therefore, we performed a systematic review and meta-analysis to compare the performance of current bleeding risk prediction scores for predicting major long-term bleeding events in patients on DAPT post-PCI. Based on a search of MEDLINE (January 1, 1946 to March 3, 2017) and EMBASE (January 1, 1974 to March 3, 2017) for studies published in the English language, we identified 10 published studies of 11 risk unique risk prediction models across a wide variety of settings. Area under the receiver operating characteristic curve (AUC) was used to measure discrimination, when available. Our findings reveal that the prediction models created to date demonstrate only modest accuracy, with the reported AUCs ranging from 0.54 to 0.89; aggregated AUC 0.68 (95% confidence intervals 0.65 to 0.72). Although only 5 studies (50%) reported measures of calibration, the reported models were reasonably well calibrated. Only 3 models (33%) were externally validated. Meta-regression demonstrated lack of influence by age (p = 0.99) or length of follow up (p = 0.42). Sensitivity analysis did not significantly change the results. Novel prediction models are warranted to aid in maximizing the benefit of DAPT after PCI while minimizing harm.
Copyright © 2018. Published by Elsevier Inc.