Emulated trial comparing the effectiveness of Bevacuzimab vs Aflibercept using an external control arm

Abstract

Purpose : There have been no randomised controlled trials comparing off-label bevacizumab to aflibercept for treating neovascular Age-related Macular Degeneration (nAMD). Recent advances in the emulation of target trials suggests that the standard-of-care arms can be approximated with real-world cohorts. Given that the treatment of retinal disorders with inhibitors of Vascular Endothelial Growth Factor accounts for 12% of Medicare expenditure, evidence for the non-inferiority of bevacizumab relative to licensed therapies could have implications for on the cost of treating nAMD.

Our hypothesis was that the visual acuity at 54 weeks of nAMD eyes treated with bevacizumab (q6w pro re nata) would not be inferior relative to aflibercept (q8w fixed-interval).

Methods : We emulated a target trial using the bevacizumab arm (n 65) of the ABC trial — a prospective, double-masked, randomised controlled trial — with external synthetic control arms derived from real-world cohorts receiving aflibercept as part of routine healthcare in the United Kingdom. Analyses were pre-specified and undertaken in accordance with the Target Trial Framework. External real-world cohorts were restricted to eyes that would have been eligible for the ABC trial and, to overcome the non-random treatment assignment in the real-world, we undertook conditional randomisation via 1:1 exact matching (EM), 1:1 propensity score matching (PSM), and inverse probability of treatment weighting (IPTW) on hypothesised confounding (age, sex, baseline read). Ordinary least squares regression was modelled on the change in visual acuity from baseline to week 54, stratified by treatment in eyes that received three loading doses within 70 days. Non-inferiority was declared if the lower confidence interval bound was ≥ than a non-inferiority margin of –4 ETDRS letters.

Results : Of 31,151 eyes that started aflibercept treatment for nAMD, 4,471 (14%) fulfilled the ABC and target trial eligibility criteria. Eyes included in the bevacizumab trial & aflibercept synthetic arms were, respectively: 131 & 8,506 for IPTW (pseudo-population); 43 & 43 for EM; and 65 & 65 for PSM. The results show, with exception of PSM, that bevacizumab is not inferior to aflibercept, on a noninferiority margin of 4 ETDRS letters.

Conclusions : We found off-label bevacizumab to be non-inferior to licensed aflibercept. We show the potential of real-world cohorts as external control arms.

This is a 2021 ARVO Annual Meeting abstract.