Can serious adverse event rates be used as a metric of trial representativeness for pharmacological interventions? Glucagon-like peptide-1 receptor agonist trials as an exemplar

Almazam, Saleh Ali M (2025) Can serious adverse event rates be used as a metric of trial representativeness for pharmacological interventions? Glucagon-like peptide-1 receptor agonist trials as an exemplar. PhD thesis, University of Glasgow.

Abstract

Background: Randomised controlled trials (RCTs) are considered the gold standard for assessing the efficacy and safety of interventions because they introduce the concept of randomisation, which effectively deals with unmeasured confounders. Trial representativeness is crucial for determining the generalisability of trial findings to the target population. However, various groups, such as older people and those with multiple long-term conditions, are often under-represented in trials. Moreover, there is no gold standard measure of trial representativeness. Current measures are complex, subjective and time-consuming, suggesting the need for new and better measures of trial representativeness. Serious adverse events (SAEs) are likely to be reported in trials and routine care. Moreover, they are reasonably objective, tangible and predictable trial outcomes. Consequently, the SAE rate has been proposed as a potential metric of trial representativeness. However, little is known about its feasibility and validity in measuring trial representativeness. This thesis investigates whether the SAE rate can be used to measure trial representativeness, using glucagon-like peptide-1 receptor agonists (GLP-1 RA) trials as an exemplar.

Methodology: This thesis builds on an ongoing systematic review of novel antidiabetics. GLP-1 RA trials were used as an exemplar for the analysis. Data sources used in this thesis were ClinicalTrials.gov, clinical study reports (CSRs), study protocols, and journal publications. I conducted several approaches to examine whether the SAE rate can be used as a proxy for trial representativeness. First, I examined SAE reporting in RCTs in the published literature to explore the feasibility of using the SAE rate as a metric of trial representativeness. Second, I explored SAE capturing in GLP-1 RA trials to enable the calculation of SAE rates. Third, I compared SAE rates between intervention and control arms to determine if combining SAE rates of trial arms is feasible to increase statistical precision. Fourth, I assessed GLP-1 RA trials using the PRECIS-2 (PRagmatic Explanatory Continuum Indicator Summary) tool to enable fair comparison with the SAE rate and explore the challenges of using this tool. I protocolised and operationalised the PRECIS-2 tool to score GLP-1 RA trials as objectively as possible. Fifth, I compared the SAE rate with the PRECIS-2 tool based on the differences in their associations with several markers of trial representativeness that serve as fair umpires to examine the validity of the SAE rate as a metric of trial representativeness. Finally, I examined the association between eligibility criteria and the SAE rate to examine whether eligibility criteria, a possible driver of trial representativeness, is associated with the SAE rate.

Results: Firstly, I found that SAEs were reported for most GLP-1 RA trials. However, SAE timeframes were not explicitly reported for nearly half of the trials. Major adverse cardiovascular event (MACE) trials had inconsistent reporting of MACE counts in SAE total counts. Secondly, I found no difference in SAE rates between intervention and control arms. Furthermore, the retrospective assessment of GLP-1 RA trials using the PRECIS-2 tool was challenging and time-consuming. The missingness of information required to score recruitment and organisation domains was high. I found no correlations between the domains of the PRECIS-2 score, except for modest correlations between eligibility criteria and recruitment domains and between setting and primary outcome domains. Moreover, I found no association between the SAE rate and the PRECIS-2 score. Additionally, I found that all fair umpires did not strongly favour the SAE rate over the PRECIS-2 tool. However, the direction of the difference in associations for half of the umpires favoured the SAE rate. Finally, I found that trials with permissive eligibility criteria were positively associated with higher SAE rates. Trials with increased continuous eligibility criteria scores were also positively associated with increased SAE rates.

Conclusion: Most GLP-1 RA trials reported sufficient SAE data indicating the ability to calculate SAE rates. SAE rates were similar across trial arms, suggesting the feasibility of combining SAE rates from the intervention and control arms. The SAE rate was not associated with the PRECIS-2 score, and none of the fair umpires strongly favoured the SAE rate. However, the directions of half of the fair umpires favoured the SAE rate and trials with permissive eligibility criteria were associated with increased SAE rates. Therefore, the SAE rate may be useful as a quick-to-measure proxy of the restrictiveness of eligibility criteria. However, given that the triangulation of evidence is inconsistent in supporting the use of the SAE rate as a standalone metric of trial representativeness, examination of the SAE rate would need to be considered carefully in combination with other metrics before making overall judgements about trial representativeness.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: R Medicine > R Medicine (General)
• Colleges/Schools: College of Medical Veterinary and Life Sciences > School of Health & Wellbeing > Public Health
• Supervisor's Name: McAllister, Professor David
• Date of Award: 2025
• Depositing User: Theses Team
• Unique ID: glathesis:2025-84908
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 17 Feb 2025 14:31
• Last Modified: 17 Feb 2025 14:31
• Thesis DOI: 10.5525/gla.thesis.84908
• URI: https://theses.gla.ac.uk/id/eprint/84908