A perspective on incomplete data in longitudinal multi-arm clinical trials, with emphasis on pattern-mixture-model based methodology.
Year: 2014
Type: dissertation
Abstract: Missing data are common in longitudinal clinical trials. Rubin described three different missing \ndata mechanisms based on the level of dependence between the missing data process and the \nmeasurement process. These are missing completely at random (MCAR), missing at random \n(MAR) and missing not at random (MNAR). Data are MCAR when the probability of dropout \nis independent of both observed and unobserved data. Data are MAR when the probability of \ndata being missing does not depend on the unobserved data, conditional on the observed data. \nWhen neither MCAR nor MAR is valid, data are MNAR. \nThe aim of this thesis is to discuss statistical methodology required for analysing missing \noutcome data and provide valid statistical methods for the MAR, MCAR and MNAR scenarios. \nThis thesis does not focus on data analysis where covariate data are missing. Under MCAR \ncomplete and available case analyses are valid. When data are MAR multiple imputation, \nlikelihood-based models, inverse probability weighting and Bayesian models are valid. When \ndata are MNAR pattern-mixture, selection and shared-parameter models are valid. These \nmethods are illustrated by an in depth analysis of two data sets with missing data. \nThe first data set is the SAPiT trial an open label, randomised controlled trial in HIVtuberculosis \nco-infected patients. Patients were randomised to three arms; each initiating \nantiretroviral therapy at a different time. CD4+ count, an indication of HIV progression, was \nmeasured at baseline and every 6 months for 24 months. The primary question was whether \nCD4+ count trajectory over time differed for the three treatment arms. The assumption that \nmissing data are MCAR was not supported by the observed data. We performed a range of \nsensitivity analyses under both MAR and MNAR assumptions. \nThe second data set is a placebo-controlled, randomised clinical trial conducted for 8 weeks to \ndetermine the effectiveness of hypericum or sertraline in reducing depression, measured by the \nHamilton depression scale. The trial randomised 340 participants, with 28% lost to follow-up \nbefore Week 8. We performed a sensitivity analysis under different assumptions about the \nmissing data process. The missing data mechanism was not MCAR. Under MAR assumptions, \nsome of the sensitivity analyses found no difference between either of the treatment arms and \nplacebo, while some found a significant difference between sertraline and placebo, but not \nbetween hypericum and placebo. This re-analysis contributed to the literature around the \neffectiveness of St John’s Wort because it changed the conclusions of the original analysis.
Author: Anna Christina. Grobler
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Keywords: missing data, multi-arm trials, pattern-mixture models, selection model, shared parameter models, missing at random, missing not at random, missing completely at random, dropout, tuberculosis-HIV coinfection, Hypericum, sertraline, depression, sensitivity analysis, multiple imputation, inverse probability weighting, likelihood-based models, Bayesian model
Open Access status: green
OpenAlex ID: W1257612111
