June 2022
Volume 63, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2022
Predicting outcomes and treatment frequency following monthly intravitreal aflibercept for macular edema secondary to central retinal vein occlusion: a machine learning model approach
Author Affiliations & Notes
  • Nitish Mehta
    NYU Langone Health, New York, New York, United States
  • Yasha Modi
    NYU Langone Health, New York, New York, United States
  • Weiming Du
    Regeneron Pharmaceuticals Inc, Tarrytown, New York, United States
  • Fabiana Q Silva
    Regeneron Pharmaceuticals Inc, Tarrytown, New York, United States
  • Hadi Moini
    Regeneron Pharmaceuticals Inc, Tarrytown, New York, United States
  • Rishi P Singh
    Cleveland Clinic Cole Eye Institute, Cleveland, Ohio, United States
  • Footnotes
    Commercial Relationships   Nitish Mehta None; Yasha Modi Alimera, Allergan, Genentech, Novartis, Zeiss, Code C (Consultant/Contractor); Weiming Du Regeneron Pharmaceuticals, Code E (Employment), Regeneron Pharmaceuticals, Code I (Personal Financial Interest); Fabiana Silva Regeneron Pharmaceuticals, Code E (Employment), Regeneron Pharmaceuticals, Code I (Personal Financial Interest); Hadi Moini Regeneron Pharmaceuticals, Code E (Employment), Regeneron Pharmaceuticals, Code I (Personal Financial Interest); Rishi Singh Genentech/Roche, Alcon, Novartis, Zeiss, Bausch & Lomb, Gyroscope, Asclepix, Regeneron Pharmaceuticals, Code C (Consultant/Contractor), Apellis, NGM Biopharma , Code F (Financial Support)
  • Footnotes
    Support  Regeneron Pharmaceuticals, Inc.
Investigative Ophthalmology & Visual Science June 2022, Vol.63, 2673. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Nitish Mehta, Yasha Modi, Weiming Du, Fabiana Q Silva, Hadi Moini, Rishi P Singh; Predicting outcomes and treatment frequency following monthly intravitreal aflibercept for macular edema secondary to central retinal vein occlusion: a machine learning model approach. Invest. Ophthalmol. Vis. Sci. 2022;63(7):2673.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : To develop machine learning (ML) algorithms to predict visual and anatomic outcomes and treatment frequency in patients with macular edema secondary to central retinal vein occlusion (MEfCRVO) at 52 weeks after undergoing treatment with aflibercept.

Methods : A dataset of 198 patients with MEfCRVO treated with monthly IAI 2 mg for 24 weeks in the COPERNICUS (n=107) and GALILEO (n=91) trials was used to develop ML algorithms. In both trials, patients were switched after 6 monthly intravitreal aflibercept injection (IAI) at week 24 to pro-re-nata (PRN) dosing through Week 52. The ML algorithm was used to predict the absolute and mean change best-corrected visual acuity (BCVA) from baseline to Week 52, the proportion of patients with ≥15-letter gain at Week 52, the absolute and mean change from baseline in central subfield thickness (CST) at Week 52, and IAI frequency during Weeks 24−52. Random Forest was used to develop the ML algorithms. Algorithm performance was assessed using correlation coefficient (r) and area under the curve (AUC) for continuous and categorical variables, respectively. Predictive factors identified with ML algorithms were confirmed using univariate analyses.

Results : The ML algorithm predicted the actual observed values at Week 52 with strong correlation (r) for absolute BCVA (r=0.87), change in BCVA from baseline (r=0.76), gain of ≥15 letters (AUC=0.81), and change in CST from baseline (r=0.76). BCVA at weeks 16, 20, and 24 were predictors of absolute BCVA; BCVA at baseline, week 20 and week 24 were predictors of change in BCVA; BCVA at baseline and at Week 20 were predictors of ≥15-letter gain; and CST and BCVA at baseline were predictors of change in CST. There was no correlation (r=0.07) between predicted and observed absolute CST at Week 52. ML algorithm predicted PRN injection frequency from Week 24 through Week 52 with high accuracy (AUC=0.83). Key factors predicting injection frequency were CST at baseline and at Week 4. Univariate analyses confirmed all predictive factors described herein.

Conclusions : ML algorithms successfully predicted visual and anatomic outcomes as well as dosing frequency with high accuracy, except for absolute CST. ML may help inform patient`s and clinician`s expectations during management of MEfCRVO.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×