June 2022
Volume 63, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2022
Do Normative Percentiles of Retinal Nerve Fiber Layer Thickness (RNFLT) Improve Prediction of Glaucomatous Visual Field Loss?
Author Affiliations & Notes
  • Rishabh Singh
    Schepens Eye Research Institute of Massachusetts Eye and Ear, Boston, Massachusetts, United States
    Boston University School of Medicine, Boston, Massachusetts, United States
  • Yangjiani Li
    Schepens Eye Research Institute of Massachusetts Eye and Ear, Boston, Massachusetts, United States
  • Mohammad Eslami
    Schepens Eye Research Institute of Massachusetts Eye and Ear, Boston, Massachusetts, United States
  • Saber Kazeminasab
    Schepens Eye Research Institute of Massachusetts Eye and Ear, Boston, Massachusetts, United States
  • Nazlee Zebardast
    Massachusetts Eye and Ear Department of Ophthalmology, Boston, Massachusetts, United States
  • Mengyu Wang
    Schepens Eye Research Institute of Massachusetts Eye and Ear, Boston, Massachusetts, United States
  • Tobias Elze
    Schepens Eye Research Institute of Massachusetts Eye and Ear, Boston, Massachusetts, United States
  • Footnotes
    Commercial Relationships   Rishabh Singh None; Yangjiani Li None; Mohammad Eslami Genentech Inc, Code F (Financial Support); Saber Kazeminasab None; Nazlee Zebardast None; Mengyu Wang Genentech Inc, Code F (Financial Support); Tobias Elze Genentech Inc, Code F (Financial Support)
  • Footnotes
    Support  NIH R01 EY030575 , P30 EY003790 , BrightFocus Foundation
Investigative Ophthalmology & Visual Science June 2022, Vol.63, 633 – A0373. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Rishabh Singh, Yangjiani Li, Mohammad Eslami, Saber Kazeminasab, Nazlee Zebardast, Mengyu Wang, Tobias Elze; Do Normative Percentiles of Retinal Nerve Fiber Layer Thickness (RNFLT) Improve Prediction of Glaucomatous Visual Field Loss?. Invest. Ophthalmol. Vis. Sci. 2022;63(7):633 – A0373.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Optical coherence tomography (OCT) manufacturers add color-coded normative information to circumpapillary RNFLT printouts, to aid in the clinical diagnosis of glaucoma (Fig 1A). In this retrospective study, we investigate whether normative percentiles of RNFLT (pRNFLT) from the Spectralis OCT platform (2009 European-descent norms) enable better prediction of glaucomatous visual field (VF) loss than absolute RNFLT.

Methods : A longitudinal database of Mass Eye & Ear Infirmary patients seen in glaucoma clinic from 2009-2020 was extracted. From a source of 158,484 Humphrey 24-2 VFs and 30,452 OCT scans, reliable OCT-VF pairs acquired within 30 days of each other were selected (Fig 1B). Spectralis normative distributions were extracted from color-coded machine printouts (Fig 1A). Supervised machine learning models predicting mean deviation (MD) and respective glaucoma staging were applied (Fig 1B). Regional structure-function associations were explored with univariate regression, correlating RNFLT or pRNFLT by scan sector with MD. Then, multivariate classification methods were applied to predict VF loss using age, scan radius, and 32-sector RNFLT or pRNFLT. R2, ROC-AUC, and accuracy scores of models were estimated using cross-validation (CV) techniques.

Results : 3021 OCT-VF pairs from 1427 patients met our reliability criteria. Extracted Spectralis norms were found to be normally distributed across the 768-point scan circle and independent of patient-specific parameters. Regional analysis showed significant decrease in R2 from pRNFLT input compared to RNFLT in inferotemporal sectors, across all tested regressors (T-tests, p=0.00-0.02). Other regions showed no significant differences (Example in Fig 2A). In multivariate classification, there were also no statistically significant differences in CV accuracy of any model between RNFLT and pRNFLT inputs (T-tests, p>0.05). This is confirmed by the ROC-AUC analysis on the testing set for each model (Fig 2B), with mean score change 0.00±0.02.

Conclusions : Our results challenge the assumption that normative percentiles from OCT machines improve prediction of VF loss. RNFLT alone shows strong prediction of glaucomatous VF loss, with no models showing significant improvement from the manufacturer norms. Greater patient stratification based on demographic factors may produce more predictive norms.

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.

×