| dc.description.abstract | The COVID-19 pandemic and the strain it placed on healthcare systems highlighted a key role for real-world data in quickly generating health evidence. A growing demand for large-scale health analytics has also resulted in the existence of large health data networks and opportunities for the incorporation of both randomised and non-randomised evidence in the health regulation decision-making process. This study collected non-randomised evidence on COVID-19 treatment effects on short-term mortality among patients hospitalised with COVID-19 from the European Health Data and Evidence Network (EHDEN) and synthesised these data with randomised evidence.
Randomised evidence was collected through literature search among two published NMAs, COVID-NMA, and clinicaltrials.gov for trials which studied the target treatments in hospitalised COVID-19 patients aged 18 years or older. Results of 21 randomised trials including 15,246 patients were combined through Bayesian network meta-analysis. Non-randomised evidence was collected as a retrospective, multinational comparative cohort study with a new user, active comparator design through EHDEN. Included patients were aged 18 years or over at cohort entry, had at least 365 days of continuous observation time prior to cohort entry, had 0 prior exposures to index treatment in the 365 days prior to index, had at least 1 COVID-19 diagnosis or positive test results in the 30 days prior to or on index, and were hospitalised on index. Although target treatments included aspirin, baricitinib, heparin, remdesivir, and tocilizumab, only evidence for remdesivir and tocilizumab was available. 475 patients from one data partner were included.
This study generated inconclusive evidence for the comparative effectiveness of remdesivir and tocilizumab in reducing short-term mortality among hospitalised patients with COVID-19. Randomised evidence showed no difference between remdesivir and tocilizumab, while non-randomised evidence showed remdesivir to significantly reduce short-term mortality compared to tocilizumab.
The non-randomised results indicate a need for more streamlined observational data collection pathways, but show that it is feasible to collect non-randomised evidence through EHDEN and that this non-randomised evidence can be compared with randomised evidence. Future studies investigating non-randomised evidence collection may build upon the infrastructure developed for this study at participating hospitals. | |
