Assessments of harms in clinical trials

Hodkinson, Alexander Assessments of harms in clinical trials. PhD thesis, University of Liverpool.

	Text PhD Thesis.docx - Unspecified Access to this file is embargoed until Unspecified. Available under License Creative Commons Attribution. Download (4MB)
	Text HodkinsonAle_Aug2015_2023762.pdf - Unspecified Available under License Creative Commons Attribution. Download (5MB)

Abstract

Introduction and Aims Healthcare interventions are usually associated with a risk of harmful events that must be balanced against the potential favorable outcomes. However reliable evidence on harms for interventions is often inadequate, and hampered by the many challenges that stem from the reporting, analysis and interpretation of harms data in clinical trials. This thesis addresses some of these issues. Methods Reporting of harms data is assessed in a systematic review of reviews and a case study investigating the additional value of harms data reported in clinical study reports (CSRs). A framework for searching and identifying relevant sources of harms data is outlined, and then explored further in a survey assessing current practices in clinical trial units (CTUs). Signal detection methods are introduced, and evaluated using simulated data to assess their performance when detecting safety signals in CTU databases. Results The systematic review highlights that the reporting of harms in RCTs is inconsistent, and often inadequate. In the case study, CSRs presented data on harms, including SAEs which are not reported or mentioned in publications, they also provide more detail about the design, conduct and analysis of the trial which facilitate the assessment of risk of bias in evidence synthesis. A wide range of sources for harms data have been identified, each with distinct strengths and limitations discussed. Selection of appropriate sources depends on the research question, and whether a hypothesis generating or hypothesis testing approach should be taken. Relevant sources have been identified for each approach, with examples of their exploitation in CTUs evaluated in the survey. The simulation study has shown that some of the current available signal detection methods are not able to control the false discovery rate well, and are only able to detect few safety signals for small or sparse data. Conclusions The work carried out within this thesis provides some recommendations to address the reporting, conduct, and analysis of harms in clinical trials. Wider adoption of recommendations made by the CONSORT-harms guideline will enhance the quality of reporting and improve subsequent evidence synthesis. Recent initiatives to promote open access to clinical trials data including CSRs is a major step towards supporting better data transparency. It is important to identify and consider different sources that are most likely to yield robust data on harms of interest, rather than relying on studies that cannot reliably detect harm. The survey identified published literature and systematic reviews as the most common source being used in the trial safety monitoring within CTUs. Signal detection methods are potentially unsuitable for use in CTUs. Further tools and guidelines for enhanced signal detection are needed in clinical trials.

Item Type:	Thesis (PhD)
Additional Information:	Date: 2015-08 (completed)
Subjects:	?? Q1 ?? ?? R1 ?? ?? RA0421 ??
Depositing User:	Symplectic Admin
Date Deposited:	10 Dec 2015 14:38
Last Modified:	17 Dec 2022 00:49
DOI:	10.17638/02023762
URI:	https://livrepository.liverpool.ac.uk/id/eprint/2023762