PSS8 - IDENTIFYING THE CONFOUNDERS THAT EXPLAIN THE DIFFERENCE IN READOUTS BETWEEN NON-INTERVENTIONAL STUDIES (NIS) AND RANDOMIZED CONTROLLED TRIALS (RCTS): THE CASE OF TREATMENTS IN NEOVASCULAR AGE-RELATED MACULAR DEGENERATION (NAMD)

Abstract

The effectiveness of intravitreal anti-vascular endothelial growth factor (anti-VEGF) agents is usually lower in real-world settings compared to randomized clinical trials (RCTs). Identifying matching cohorts might bridge the gap in clinical outcomes, but there is uncertainty about which factors to include in the matching process. This study aimed to identify potential confounders that explain the difference in readouts between non-interventional studies (NIS) and RCTs in nAMD patients using an algorithm based on ranibizumab real-world data (RWD), in reference to the CATT study1 results. A step-wise, predictive analytics approach was applied to develop a decision model where the terminal nodes represented the most influential factors on visual acuity (VA) at the end of year 1. The decision tree used electronic medical records (EMR) data from the USA. The algorithm was validated using EMR data from UK and Australia. The algorithm identified the following criteria as impactful on readouts: 1) Age>50 years, 2) Baseline VA<73 ETDRS letters, 3) presence of sub-retinal fluid and 4) administration of 3 loading doses by day-90 from drug initiation. When applying criterion 1, patients showed an increase in VA of 0.07. Application of criterion 2 increased VA by 1.08. Similarly, when criteria 3&4 were applied in addition to 1&2, VA increased by 3.22 and 5.25, respectively. VA results after applying the complete algorithm were comparable to those of the CATT trial which demonstrated a 6 letter gain. In nAMD, core attributes that matched NIS with RCT outcomes were age, baseline VA, sub-retinal fluid and loading-dose as label recommended. Our investigation demonstrates that RWD can generate clinical efficacy (VA) comparable to RCTs. These results highlight the impact of under-treatment and loading-dose on VA and build confidence in real-world evidence and its ability to support meaningful dialogue with ophthalmologists and retina specialists. 1Ophthalmology. 2016;123(8):1751-1761.