McAllister, DavidAlsallumi, Khalid2025-03-062024https://hdl.handle.net/20.500.14154/74990Background: Randomised controlled trials (RCTs) are the gold standard for determining the efficacy and safety of medical interventions. However, their representativeness is often uncertain, as many trials tend to recruit healthier, younger, and less comorbid patients, potentially limiting the generalisability of findings to the real-world population. Assessing trial representativeness is complex, and currently, there is no gold standard measure to address this. Serious adverse events (SAEs) are clearly defined by the Food and Drug Administration (FDA), and regulatory bodies obligate trial sponsors to report all SAEs, regardless of causation. SAEs may reflect the underlying health status of participants, as they include events such as hospitalisations that may not be directly related to the intervention, thus serving as a potential marker of how sicker the participants are. Consequently, SAE reporting could provide insights into trial representativeness. In this thesis, I will explore whether SAE rates can be used to measure trial representativeness, using RCTs of sodium-glucose co-transporter-2 (SGLT-2) inhibitors as an exemplar. Methods: Clinical trials of SGLT-2 inhibitors were identified from a recent systematic review and included in this thesis. I extracted data on SAE reporting, eligibility criteria, and multiple variables for trial and baseline characteristics. ClinicalTrials.gov, other trial registries, clinical study reports, and trial-relevant publications were used to extract the required data. To assess the feasibility of using SAE rate as a measure of trial representativeness, I initially explored whether SGLT-2 trials reported adequate information to calculate the SAE rates. I then compared SAE rates between trial arms to determine whether SAE rates varied depending on treatment and, by extension, whether each arm should be considered separately or together when assessing SAE rates. In the absence of a gold standard, I used different approaches to explore whether SAEs reflect trial representativeness. First, I used the pragmatism metric (PRECIS-2 tool, PRagmatic Explanatory Continuum Indicator Summary) to examine its association with SAE rates. I operationalised the PRECIS-2 tool to align with the characteristics of SGLT-2 trials and assessed the trials retrospectively by scoring the nine PRECIS-2 domains. Second, I compared SAE rates with the PRECIS-2 score based on the differences in their associations with trial and baseline characteristics (included as fair umpires) to explore whether SAE is a better metric. Lastly, I examined the association between trial eligibility criteria and the SAE rates to further assess their potential as a measure of trial representativeness. Results: A total of 146 RCTs for SGLT-2 inhibitors were identified and included in the analysis. In my literature review I found a lack of representativeness in clinical trials, which limited the generalisability of their findings. This lack of representativeness is driven by underrepresentation of older patients, racial/ethnic minorities, and the impact of strict eligibility criteria. Trials of SGLT-2 inhibitors reported sufficient information, including the number of participants who experienced SAEs, the number of subjects at risk of SAEs, and the timeframe for these events, allowing for the calculation of SAE rates. Trials registered on ClinicalTrials.gov showed better SAE reporting than other trials. The reporting of major adverse cardiovascular events (MACE) was consistently included in the reporting of SAEs. There was no significant difference in the rates of SAE between the intervention and placebo arms (incidence rate ratio [IRR] 0.89, 95% confidence interval [CI] 0.73-1.07), and between the intervention and active comparator arms (IRR 0.90, 95% CI 0.69-1.17). Trials registered on ClinicalTrials.gov had higher SAE rates than trials registered on other registries (IRR 1.94, 95% CI 1.23-3.02). Multinational trials showed higher SAE rates than national trials (IRR 1.79, 95% CI 1.32-2.41). Additionally, SAE rates were higher in trials with hard outcomes (e.g., MACE) compared to those with soft outcomes (IRR 2.86, 95% CI 1.84-4.74). A higher mean PRECIS-2 score was associated with higher SAE rates (β = 0.17, 95% CI 0.00–0.33). The score for the setting domain was also associated with higher SAE rates (β = 0.51, 95% CI 0.26–0.76). Furthermore, the SAE rate and mean PRECIS-2 score were positively associated with certain baseline/trial characteristics (umpires), such as diabetes duration and sponsorship. The trend for most umpires generally favoured SAE as a metric of trial representativeness; however, only age and type of blinding umpires strongly favoured SAE over PRECIS-2 (β = 0.97; 95% CI 0.26 to 1.70, β = 0.72; 95% CI 0.02 to 1.43, respectively). Trials with more restrictive eligibility criteria had lower SAE rates (IRR 0.75, 95% CI 0.73-0.77) than trials with more permissive criteria (IRR 1.33, 95% CI 1.30-1.37). Conclusion: Clinical trials of SGLT-2 inhibitors reported sufficient information on SAE, allowing for the calculation of SAE rates. There was no difference in SAE rates between trial arms, enabling the combination of total SAEs for both arms. Pragmatic trials, which resemble real-world practice according to the mean PRECIS-2 score, showed higher SAE rates. The trend for most umpires favoured SAE more than PRECIS-2, and trials with restrictive eligibility criteria showed lower SAE rates compared to those with permissive criteria. Therefore, SAE rates may help assess the representativeness of clinical trials.186enclinical trialsrandomised controlled trials (RCTs)trial representativenessserious adverse events (SAEs)generalizabilityapplicabilityThesis