5.30pm – 6.00pm:      Mouna Akacha (Warwick University)

Missing Data in Longitudinal Studies: A Review

In clinical trials it is very common for sets of repeated measurements to be incomplete. Missingness usually occurs for reasons outside of the control of the investigators and may be related to the outcome measurement of interest, hence complicating the data analysis.

Over the last decades several approaches for the analysis of studies with missing data have been discussed. All of these methods rest on assumptions which concern the relation between the missingness process and the outcome variable of interest. These assumptions are to a certain extent untestable.

In this talk we will give a review on potential problems that arise with missing data and on methods to handle missing data. We will discuss the underlying assumptions and limitations of these approaches. Finally, we will illustrate the challenges we encountered in analyzing two clinical trials with missing data.

Assessing the sensitivity of meta-analysis to selection bias: a multiple imputation approach

Evidence synthesis, both qualitatively and quantitatively through meta-analysis, is central to the development of evidence-based medicine. Unfortunately, meta-analysis is often complicated by the suspicion that the available studies represent a biased subset of the evidence, possibly due to publication bias or other systematically different effects in small studies. A number of statistical methods have been proposed to address this, among which the trim-and-fill method and the Copas selection model are two of the most widely discussed.

Here, we adopt a logistic selection model, and show how treatment effects can be rapidly estimated via multiple imputation. Specifically, we impute studies under a missing at random assumption, and then re-weight to obtain estimates under non-random selection. Our proposal is computationally straight forward. It allows users to increase selection while monitoring the extent of remaining funnel plot asymmetry, and also visualise the results using the funnel plot. We illustrate our approach using a small meta-analysis of benign prostatic hyperplasia.