Estimating Treatment Effects with Missings Not At Random in the Estimand Framework using Causal Inference

TL;DR

This paper introduces a causal inference approach to estimate treatment effects in randomized trials with MNAR missing data, avoiding imputation by modeling missingness with directed acyclic graphs and potential responses.

Contribution

It presents a novel method for unbiased treatment effect estimation under MNAR missing data without imputation, using causal modeling and formal robustness results.

Findings

Unbiased treatment effect estimation without imputation under MNAR.

Modeling missingness with directed acyclic graphs improves accuracy.

Average conditional log-odds ratio is robust if adjusted by confounders.

Abstract

The analysis of randomized trials is often complicated by the occurrence of intercurrent events and missing values. Even though there are different strategies to address missing values it is still common to require missing values imputation. In the present article we explore the estimation of treatment effects in RCTs from a causal inference perspective under different missing data mechanisms with a particular emphasis on missings not at random (MNAR). By modelling the missingness process with directed acylcic graphs and patient-specific potential response variables, we present a new approach to obtain an unbiased estimation of treatment effects without needing to impute missing values. Additionally, we provide a formal that the average conditional log-odds ratio is a robust measure even under MNAR missing values if adjusted by sufficient confounders.