Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 442 - Disease Prediction, Statistical Methods for Genetic Epidemiology and Mis
Type: Contributed
Date/Time: Thursday, August 12, 2021 : 4:00 PM to 5:50 PM
Sponsor: Section on Statistics in Epidemiology
Abstract #318177
Title: Numerical Assessment of Robustness of Treatment Effect Estimation with Missing Data
Author(s): Zhibao Mi* and Min Zhan and Eileen Stock and Kousick Biswas
Companies: VA CSPCC at Perry Point and VA CSPCC at Perry Point and VA CSPCC at Perry Point and VA CSPCC at Perry Point
Keywords: Intent-to-Treat; MNAR; Estimation; Robustness ; Missing Imputation

Intention-to-treat (ITT) analysis is a common practice in clinical trials, which maintains integrity of trial randomization and testing power and avoids misleading artifacts. However, one of the challenges to conducting an ITT analysis is to appropriately handle the missing data which is nearly inevitable in most trial practices and causes type II error inflation and potentially biased estimates of treatment effects especially when data is missing not at random (MNAR). ITT analysis based on incomplete data relies on data imputation to reduce potential estimation bias and preserve testing power. Despite advances in imputation techniques, determining the validity of imputed values is complex, making interpretation of trial results challenging. Given the nature of data imputation as an informative guess of measures as if they would be observed, sensitivity analyses are performed with varying assumptions of the missing data to assess the robustness of the results based on imputed data from VA multicenter clinical trials. Degree of the robustness is assessed numerically by simulations with varying sample sizes, missing proportions, and treatment effects.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2021 program