July 2019
Volume 60, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2019
Detecting functional change using permutation tests on overlapping clusters of visual field locations
Author Affiliations & Notes
  • Stuart Keith Gardiner
    Legacy Research Institute, Devers Eye Institute, Portland, Oregon, United States
  • Shaban Demirel
    Legacy Research Institute, Devers Eye Institute, Portland, Oregon, United States
  • Steven L Mansberger
    Legacy Research Institute, Devers Eye Institute, Portland, Oregon, United States
  • Cindy Albert
    Legacy Research Institute, Devers Eye Institute, Portland, Oregon, United States
  • Footnotes
    Commercial Relationships   Stuart Gardiner, None; Shaban Demirel, None; Steven Mansberger, None; Cindy Albert, None
  • Footnotes
    Support  NIH EY020922; NIH EY019674
Investigative Ophthalmology & Visual Science July 2019, Vol.60, 2465. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Stuart Keith Gardiner, Shaban Demirel, Steven L Mansberger, Cindy Albert; Detecting functional change using permutation tests on overlapping clusters of visual field locations. Invest. Ophthalmol. Vis. Sci. 2019;60(9):2465.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Global perimetric indices such as Mean Deviation (MD) can miss localized changes; pointwise analyses are more sensitive but also more variable. We previously showed that averaging within each of ten predefined non-overlapping clusters of visual field locations allows earlier detection of subsequently-confirmed progression. However, this may still be insensitive to defects that straddle borders between clusters. Here, we extend the approach to capture more of the localized changes using overlapping, anatomically-based clusters.

Methods : 495 eyes of 254 participants with suspected or confirmed glaucoma were tested for a mean of 15 visits (range 5-27) over a mean of 11 years (2-19), using the 24-2 test grid. For a chosen sector width θ°, 360 clusters were defined as the visual field locations whose axons enter the optic nerve head in sectors 0° to θ°, 1° to (θ+1)°, 2° to (θ+2)°, etc., based on the map of Garway-Heath et al (Ophthalmology 2000). “Cluster deviations” were defined as the mean total deviation within each cluster that had ≥2 locations. Change was “detected” at visit V if at least N cluster deviations worsened with p<pCrit by linear regression over visits 1-V, and this was confirmed using visits 1-(V+1). The pCrit for each series was derived by Monte-Carlo sampling, using 475 random permutations of the series (but the same visit dates) to obtain specificity 95% with accuracy ±1%. Similar criteria were derived for N of the ten pre-defined clusters and for MD. Time to detect change was compared between criteria using stratified Cox proportional hazards survival models.

Results : The median time to detect subsequently-confirmed change was shortest when N was around 10% of the available clusters, for a wide range of widths θ°, as shown in the Table. Using all eyes, overlapping clusters detected change significantly sooner than MD (p=0.01 for θ=90°, all others p<0.01); but similarly to using pre-defined clusters, as seen on the survival curves. However, restricting analyses to the 240 eyes that had abnormal visual fields on their last visit, overlapping clusters detected change sooner than either MD (p=0.08 for θ=90°, all other θ p<0.05) or pre-defined clusters (all p<0.05).

Conclusions : Subsequently-confirmed change is detected sooner using several overlapping clusters of visual field locations than when using either MD or pre-defined non-overlapping clusters, for the same specificity.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

 

×
×

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.

×