Global sensitivity analysis for repeated measures studies with informative drop-out: A semi-parametric approach

Area of Science:

• Biostatistics
• Clinical Trials
• Psychiatric Research

Background:

• Repeated measures studies with drop-out necessitate both testable and untestable assumptions for valid inference.
• Existing sensitivity analysis methods, like Scharfstein et al. (2014), rely on restrictive parametric models.
• These parametric assumptions limit the applicability and flexibility in empirical research.

Purpose of the Study:

• To develop and present a more flexible, semi-parametric approach for sensitivity analysis in repeated measures studies with drop-out.
• To relax the restrictive distributional assumptions of prior methodologies.
• To enhance the robustness of statistical inference in the presence of missing data.

Main Methods:

• Proposed a semi-parametric sensitivity analysis methodology.
• Relaxed the fully parametric distributional assumptions of previous approaches.
• Applied the methodology to a randomized trial for schizoaffective disorder treatment.

Main Results:

• The semi-parametric approach offers greater flexibility compared to fully parametric models.
• Demonstrated the utility of the proposed method in a real-world clinical trial setting.
• Provided a more robust framework for analyzing data with drop-outs.

Conclusions:

• The developed semi-parametric approach is a valuable advancement for sensitivity analysis in longitudinal studies.
• This method improves the reliability of conclusions drawn from data with missing observations.
• Facilitates more accurate treatment effect evaluation in clinical trials, particularly in psychiatric research.