June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Big data analytics in small sample sizes: collective analysis of visual acuity and contrast sensitivity endpoints
Author Affiliations & Notes
  • Luis A Lesmes
    Adaptive Sensory Technology, Inc, San Diego, California, United States
  • Michael Dorr
    Adaptive Sensory Technology, Inc, San Diego, California, United States
  • Yukai Zhao
    Center for Neural Science and Department of Psychology, New York University, New York, New York, United States
  • Zhong-Lin Lu
    Center for Neural Science and Department of Psychology, New York University, New York, New York, United States
    Division of Arts and Sciences, New York University Shanghai, Shanghai, Shanghai, China
  • Footnotes
    Commercial Relationships   Luis Lesmes Adaptive Sensory Technology, Code E (Employment), Adaptive Sensory Technology, Code I (Personal Financial Interest), Adaptive Sensory Technology, Code P (Patent); Michael Dorr Adaptive Sensory Technology, Inc, Code E (Employment), Adaptive Sensory Technology, Inc, Code I (Personal Financial Interest), Adaptive Sensory Technology, Inc, Code P (Patent); Yukai Zhao None; Zhong-Lin Lu Adaptive Sensory Technology, Inc, Jiangsu Juehua Medical Technology, Ltd., Code I (Personal Financial Interest), Adaptive Sensory Technology Inc., Jiangsu Juehua Medical Technology, Ltd., Code P (Patent)
  • Footnotes
    Support  NIH Grant EY017491
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 3340. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Luis A Lesmes, Michael Dorr, Yukai Zhao, Zhong-Lin Lu; Big data analytics in small sample sizes: collective analysis of visual acuity and contrast sensitivity endpoints. Invest. Ophthalmol. Vis. Sci. 2023;64(8):3340.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Early-stage clinical trials in retina are marked by small sample size, noisy visual function endpoints, and variability in the treatment response. To address these issues and improve signal detection of vision changes in early-stage studies, we present a proof-of-concept for collective endpoint (CE) analysis, using Hierarchical Bayes Joint Modeling (HBJM) of visual acuity (VA) and contrast sensitivity (CS).

Methods : In a small sample (N=14), we analyzed within-subject VA and CS changes induced by Bangerter foils (Zhao et al. 2021), measured with quantitative VA (qVA) and quantitative CSF (qCSF) testing across three conditions defined by the number of completed qVA/qCSF rows: short (5 qVA/15 qCSF), medium (15 qVA/25 qCSF), and long (45 qVA/50 qCSF). HBJM was applied to trial-by-trial data to generate a five-dimensional joint posterior that combines parameters for VA (threshold,range) and CS (peak gain, peak frequency, and bandwidth). In addition to VA and CS changes estimated from one-dimensional analysis (VA1D, CS1D), and VA and CS changes derived from CE analysis (VA-CE, CS-CE), we evaluated signal detection for the full 5D collective endpoint (CE5D). For data from each eye, signal-noise features were evaluated by signal magnitude (mean change), noise magnitude (variance of change), and significance testing for individualized vision changes (p-values).

Results : Figure 1 (a,b) shows consistent change signals for VA and CS, (~1 dB=10 log10 unit), across all test lengths, and across 1D and CE analyses. Figure 1 (c,d) shows how the recovery of strong covariances by CE analysis reduces the variance of VA/CS estimates, relative to 1D analysis.

Figure 2 shows how statistical power for individualized significance testing is improved by the optimal combination of signals and reduction of noise provided by CE analysis. At the individual level, median p-values for VA1D and CS1D changes were not significant for short or medium testing (p>.05), but CE5D changes were statistically significant for short (p<.025), medium (p<.0025), and long testing (p<.00005).

Conclusions : The signal detection of vision changes in small sample sizes can be improved by a collective analysis that combines signals and reduces noise across multiple endpoints. Further studies are needed to determine the potential value of this data science framework for detecting vision gains in early stage studies in retina.

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.

×