Controlled pattern imputation for sensitivity analysis of longitudinal binary and ordinal outcomes with nonignorable...

Area of Science:

• Statistics
• Biostatistics
• Clinical Trials

Background:

• Controlled imputation is a pattern mixture model used for sensitivity analyses in longitudinal clinical trials with nonignorable dropout.
• Existing methods lack formal development for longitudinal binary and ordinal outcomes due to distribution challenges.

Purpose of the Study:

• To propose and implement controlled imputation methods for longitudinal binary and ordinal data.
• To develop efficient algorithms for missing data imputation in these contexts.

Main Methods:

• Sequential logistic regression for binary outcomes.
• Multivariate probit model for ordinal outcomes.
• Markov chain Monte Carlo algorithms with data augmentation techniques for imputation.

Main Results:

• The proposed methods were assessed through simulation studies.
• Performance was evaluated using a schizophrenia clinical trial dataset.
• Comparisons were made against fully conditional specification, LOCF, and BOCF imputation methods.

Conclusions:

• The developed controlled imputation approaches provide viable methods for handling nonignorable dropout in longitudinal binary and ordinal data.
• These methods offer improved sensitivity analyses compared to traditional imputation techniques.