Estimands and estimators of two-level methods using return to baseline strategy for longitudinal clinical trials with incomplete daily patient reported outcomes.

Publisher: Taylor & Francis Country of Publication: England NLM ID: 9200436 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1520-5711 (Electronic) Linking ISSN: 10543406 NLM ISO Abbreviation: J Biopharm Stat Subsets: MEDLINE

Publication: 2005- : Abingdon, Oxford : Taylor & Francis
Original Publication: New York, N.Y. : Marcel Dekker, c1991-

Research Design*; Humans ; Likelihood Functions ; Longitudinal Studies

Returning to baseline (RTB) has been a practical method for handling missing data. Here we consider longitudinal clinical trials with daily patient reported outcomes (PROs), where efficacy endpoints are often defined as the average daily values in a cycle (such as a month or a week). The conventional method treats data at cycle level and ignores daily values. In this paper, we build a two-level constrained longitudinal data analysis (cLDA) model on daily values and propose two-level RTB method to impute daily values. Standard multiple imputation (MI) approach and likelihood-based approach are proposed and evaluated by simulations.

Keywords: Missing not at random; multiple imputations; reference-based imputation; return to baseline

Date Created: 20210624 Date Completed: 20230626 Latest Revision: 20230626

20230626

10.1080/10543406.2021.1934855

34162312