Dr Susanna Dodd PhD

My NIHR-funded PhD research investigated the use of causal methods to account for deviations from randomised treatment in the analysis of randomised trials. My review of published trials (Dodd, White, & Williamson, 2012) demonstrated that, although the vast majority of trials are subject to treatment deviations, appropriate methods to account for these deviations in trial analysis are very rarely implemented; instead trialists use simple analysis methods (such as per protocol or as treated analyses) when seeking to estimate treatment efficacy, thereby introducing confounding and selection biases by naively analysing according to treatment received. More appropriate causal methods exist which seek to overcome these inherent biases, but they remain underused, primarily due to their complexity and unfamiliarity. The aim of my research has therefore been to explore, explain, demonstrate and compare the use of causal methodologies in the analysis of trials, in order to increase the accessibility and comprehensibility by non-specialist analysts of the available, but somewhat technical, statistical methods to appropriately adjust for treatment deviations. A practical application of appropriate causal methods is demonstrated in the analysis of a trial featuring typical complications when assessing treatment in chronic disease with longitudinal treatment and follow up periods, trials of which are often subject to deviation from randomised intervention (Dodd, Williamson & White 2017, SMMR). The work has culminated in publication of a framework of guidance for trialists and trial statisticians who are seeking to account for treatment deviations in the design, conduct and analysis of a trial, highlighting issues that must be considered from the trial planning stage regarding data collection and analysis, summarised as a list of recommendations (Dodd, White & Williamson, 2017 Trials).

I was involved in the ORBIT (Outcome Reporting Bias In Trials) project, helping to develop the study classification system for missing or incomplete outcome reporting in reports of randomised trials, and in the outcome classification project associated with the COMET (Core Outcome Measures in Effectiveness Trials) initiative. This project led to the development of a classification system for outcomes included in systematic reviews, core outcome sets and trial reports, with the aim of promoting efficient searching, reporting and classification of trial outcomes. Current work relates to the profiling of outcomes in core outcome sets for practice. I continue to work in numerous COS uptake projects, assessing the representation of COS in trials, systematic reviews, health technology assessments, clinical and regulatory guidelines.

Nonadherence is common in general medical practice and clinical trials, with significant economic and clinical consequences. It complicates trial analysis and interpretation, as the underlying random assignment mechanism, which forms the basis for unbiased hypothesis testing, no longer reflects treatment received. Adherence information is crucial to inform efficacy (or causal) analysis, but although several adherence measurement techniques exist, each has considerable limitations such that no method is considered a gold standard. Trials typically rely on patient-reported adherence data, but such methods are susceptible to distortion, as patients are reticent to fully disclose their true adherence. Substantial amounts of adherence research have been conducted regarding its causes, impact and solutions, but research into improving the accuracy of adherence data, particularly when provided by patients, is severely lacking. I am therefore interested in research that aims to determine how to obtain acccurate patient-reported adherence information in order to reliably inform causal analyses.