Application of Calibration Estimator in Comparison of TVR-MACE Survival Rate Between IntraVascular UltraSound Guided PCI and Angiography Guided PCI Within Complex Lesion Population

• Activity Number: 359 - Contributed Poster Presentations: Section on Medical Devices and Diagnostics
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Cheng Zhang*
• Keywords: calibration estimator; raking; propensity score; IVUS PCI; complex lesions; generalized boosted model

Abstract

To power the IMPROVE trial, which aims to compare the treatment effect of IntraVascular UltraSound (IVUS) guided PCI to that of angiography guided PCI within the complex lesion population, we estimate the accumulated incidence rate of all-cause mortality TVR-MACE at 1 year using Medstar WHC retrospective data. Complex lesion criteria are defined by the presence o Type C, ISR, CTO, Long Lesions, Bifurcation, Severe calcification or LMCA. To account for confounders in retrospective study, we first weigh each unit with the inverse of propensity score. A generalized boosted model is applied to model propensity, resulting in an AUC of .76. To account for the potential local intuition bias, such as a higher than average rate of ISR, we further adjust the base weight by the calibration equations formulated using literature prevalence of high risk factors using raking estimators. The accumulated incidence rate based on calibration estimators adjusts for both confounders and local institution bias. We also formulated various calibration estimators based on different combinations of high risk lesion factors to examine the robustness of calibration estimators on sample size calculation.