Abstract
Purpose :
HAWK and HARRIER are 2-year randomized phase III trials comparing brolucizumab and aflibercept for the treatment of neovascular age-related macular degeneration. A deep learning-based analysis of OCT scans was applied to compare effect size and durability of anti-VEGF agents on macular fluids.
Methods :
In HARRIER, 743 patients were randomized to brolucizumab 6mg q12w, with the option of increasing the dosing frequency to q8w, or aflibercept 2mg at a q8w dosing. A previously described deep learning-based algorithm was applied on 19,034 OCT scans acquired monthly to derive volumes (in nl) of intraretinal fluid (IRF), subretinal fluid (SRF) and pigment-epithelial detachment (PED) for the areas around the macular center (1, 3, and 6 mm). The same approach was followed for HAWK.
Results :
In HARRIER, baseline mean (SD) IRF volumes (20 (41), 62 (122) and 82 (168) nl in the 1, 3 and 6 mm areas, respectively) decreased by >93% following the first intravitreal injection and consistently remained very low during the entire follow-up with both agents. Baseline mean (SD) SRF volumes (20 (36), 123 (177) and 364 (499) nl in the 1, 3 and 6 mm areas, respectively) decreased by >75% following the first injection with both agents. However, during follow-up, SRF volumes resolved significantly better with brolucizumab while aflibercept left fluctuating amounts of SRF up to 100nl across the 6mm area. Moreover, PED volume resolution to about 50% of its baseline mean (SD) value (69 (92), 337 (452) and 425 (583) nl in the 1, 3 and 6 mm areas, respectively) was consistently more pronounced and stable with brolucizumab. Results from HAWK were analysed in the same manner.
Conclusions :
Brolucizumab induced a durable resolution of neovascular fluid as early as after a single injection in all compartments. While superficial fluid (IRF) resolved most rapidly and similarly with both agents, brolucizumab, with a q12w/q8w regimen, was consistently superior in reducing deeper fluid volumes (SRF, PED) associated with neovascular activity.
This is a 2020 ARVO Annual Meeting abstract.