Abstract:ABSTRACT Longitudinal studies are often subject to missing data. The recent guidance from regulatory agencies, such as the ICH E9(R1) addendum addresses the importance of defining a treatment effect estimand with the consideration of intercurrent events. Jump-to-reference (J2R) is one classical control-based scenario for the treatment effect evaluation, where the participants in the treatment group after intercurrent events are assumed to have the same disease progress as those with identical covariates in the control group. We establish new estimators to assess the average treatment effect based on a proposed potential outcomes framework under J2R. Various identification formulas are constructed, motivating estimators that rely on different parts of the observed data distribution. Moreover, we obtain a novel estimator inspired by the efficient influence function, with multiple robustness in the sense that it achieves n1/2-consistency if any pairs of multiple nuisance functions are correctly specified, or if the nuisance functions converge at a rate not slower than n−1/4 when using flexible modeling approaches. The finite-sample performance of the proposed estimators is validated in simulation studies and an antidepressant clinical trial.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the method of dealing with missing data in longitudinal studies, especially evaluating the treatment effect in the context of Control - Based Imputation (CBI). Specifically, the paper focuses on the development and application of Multiply Robust Estimators in the Jump - to - Reference (J2R) scenario.
### Background and Problems
1. **Missing Data in Longitudinal Studies**
- In long - term studies, participants may deviate from the current treatment plan for various reasons (such as loss to follow - up or switching to other therapies), resulting in missing data.
- Missing data poses a challenge to the accurate estimation of treatment effects and requires additional assumptions to handle these missing parts.
2. **Definition of Treatment Effect Estimators**
- The annex of International Conference on Harmonisation (ICH) E9(R1) emphasizes the importance of defining treatment effect estimators, especially when considering concurrent events.
- J2R is a classic control - group - based scenario, assuming that participants in the treatment group after a concurrent event have the same disease progression as individuals in the control group with the same covariates.
### Main Contributions of the Paper
1. **Proposing New Estimators**
- Based on the Potential Outcomes Framework, the paper proposes new estimators to evaluate the Average Treatment Effect (ATE).
- These estimators utilize different parts of the observed data distribution and are multiply robust, that is, if any pair of multiple nuisance functions is correctly specified, or when using a flexible modeling method, the convergence rate of the nuisance functions is no slower than \(n^{- 1/4}\), then the estimator can achieve \(n^{1/2}\)-consistency.
2. **Theoretical Contributions**
- The paper derives the Efficient Influence Function (EIF) of longitudinal data in the J2R scenario, which is a theoretical breakthrough in this field.
- The proposed estimators have triple robustness, that is, in a parametric model, if any two of the three models are correctly specified, the estimator is consistent; when using a flexible model, if the convergence rate of any model is no slower than \(n^{-1/4}\), the estimator is also consistent.
3. **Practical Applications**
- The finite - sample performance of the proposed estimators is verified through simulation studies and data from an antidepressant drug clinical trial.
- In the specific case analysis, the paper uses data from an antidepressant drug trial conducted by the Drug Information Association to demonstrate the application of the J2R method.
### Conclusion
The paper addresses the challenges of missing data processing in longitudinal studies, especially in the J2R scenario, by proposing new multiply robust estimators. These estimators not only have superior properties in theory but also show good performance in practical applications.