June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Potential Clinical Benefits of a Composite Automated Score for Dry Eye Redness
Author Affiliations & Notes
  • Rachel Zilinskas
    Statistics and Data Corporation, Arizona, United States
  • Adam Hamm
    Statistics and Data Corporation, Arizona, United States
  • John David Rodriguez
    Ora Clinical, Massachusetts, United States
  • Benjamin Liu
    Ora Clinical, Massachusetts, United States
  • Mark B Abelson
    Ora Clinical, Massachusetts, United States
  • Footnotes
    Commercial Relationships   Rachel Zilinskas Statistics and Data Corporation, Code E (Employment); Adam Hamm Statistics and Data Corporation, Code E (Employment); John Rodriguez Ora Clinical, Code E (Employment); Benjamin Liu Ora Clinical, Code E (Employment); Mark Abelson Ora Clinical, Code E (Employment)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 3987. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Rachel Zilinskas, Adam Hamm, John David Rodriguez, Benjamin Liu, Mark B Abelson; Potential Clinical Benefits of a Composite Automated Score for Dry Eye Redness. Invest. Ophthalmol. Vis. Sci. 2023;64(8):3987.

      Download citation file:

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

  • Supplements

Purpose : Dry eye redness is an important primary endpoint in clinical trials. This endpoint is typically recorded manually by expert human graders. Though considered to be the “gold standard” in determining these measures, there is often variability present both within and between human graders. This study aims to quantify the potential benefit of using an automated score relative to expert human graders.

Methods : Using a real world data sample where both human grades, automated redness, and automated horizontality of conjunctival vessels were measured, 1,2 we first showed the differences in variability between human grades and automated scores. A linear model was used to designate a composite score, incorporating both the automated redness and horizontality metrics. Another linear model was fit on a separate test data set to determine equivalent values of grader scores in terms of the automated composite score. Using parameters from a sample size calculation for a trial powered (alpha = 0.05 , beta = 0.10) to show a 0.117 difference in grader redness scores with a standard deviation of 0.31, we repeatedly sampled a corresponding predicted value for the difference and standard deviation in terms of the proposed automated composite score and conducted a sample size calculation.

Results : From the real world sample data analysis, the coefficient of variation for the human grader scores was 50% while the corresponding automated composite score had a coefficient of variation of only 30%. Using 1,000 simulated values for the sample size calculation, 99.5% of the resulting sample size calculations with automated composite scores were less than the sample size determined for the trial using grader-determined redness as the primary outcome. The average sample size from the simulations was approximately 20% of the size for the real trial at matched alpha and beta levels.

Conclusions : We have previously shown and validated a composite score for redness using automated redness and horizontality1,2. This composite score exhibits less variability than scores determined by human graders and shows meaningful potential benefit in terms of reducing the necessary sample size in a trial with dry eye redness as the primary endpoint. These results along with the practical benefits of using an automated score provide support for implementation in a clinical trial setting.
1 Rodriguez et al. (2013). Clinical Ophthalmology
2Liu et al. ARVO 2023

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.


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.