April 2014
Volume 55, Issue 13
Free
ARVO Annual Meeting Abstract  |   April 2014
More accurate modelling of visual field progression in glaucoma: ANSWERS
Author Affiliations & Notes
  • Haogang Zhu
    School of Health Sciences, City University London, London, United Kingdom
    Institute of Ophthalmology, University College London, London, United Kingdom
  • David Paul Crabb
    School of Health Sciences, City University London, London, United Kingdom
  • David F Garway-Heath
    Institute of Ophthalmology, University College London, London, United Kingdom
  • Footnotes
    Commercial Relationships Haogang Zhu, None; David Crabb, Allergan (F), Allergan (R), Merck (F), Merck (R), Pfizer (F); David Garway-Heath, Alcon (C), Alcon (R), Alimera (C), Allergan (C), Allergan (F), Allergan (R), Bausch & Lomb (C), Bausch & Lomb (R), Carl Zeiss Meditec (F), Forsight (C), Heidelberg Engineering (F), OptoVue (F), Pfizer (F), Quark (C), Sensimed (C), Teva Pharmaceutica (C)
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science April 2014, Vol.55, 5628. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Haogang Zhu, David Paul Crabb, David F Garway-Heath, ; More accurate modelling of visual field progression in glaucoma: ANSWERS. Invest. Ophthalmol. Vis. Sci. 2014;55(13):5628.

      Download citation file:


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

      ×
  • Supplements
Abstract
 
Purpose
 

Analysis with Non-Stationary Weibull Error Regression and Spatial Enhancement (ANSWERS; Zhu et al, PLoS ONE 2013) estimates the rate of change at locations in the visual field (VF) series, taking into account increasing measurement variability during disease progression and the spatial correlation among test locations. The rate of change may be used to predict future VF sensitivity. The assumption is that a more accurate model of progression would result in more accurate prediction of the future VF. This study evaluates the prediction accuracy of ANSWERS in comparison with conventional ordinary least squares linear regression (OLSLR).

 
Methods
 

The comparison was made in VF series from the UK glaucoma treatment study (UKGTS; Garway-Heath et al, Ophthalmology 2013), in which patients were followed up for 2 years. The 24-2 VFs were acquired with Humphrey Field Analyser (Carl Zeiss Meditec, CA) using SITA standard. Only the VFs with false positive response <15% were used, resulting in 659 eligible eyes of 437 patients. ANSWERS and OLSLR were used to predict the VF at next visit using subseries that are within 6, 12, 18 or 24 months from the baseline visit. The prediction accuracy was summarised by the mean absolute difference (MAD) between the predicted and measured VFs at 52 test locations. The association between the difference in prediction accuracy between OLSLR and ANSWERS (MAD difference) and the amount of change from baseline (average difference between VF being predicted and baseline) was also investigated.

 
Results
 

There are 6, 8, 12 and 14 VFs respectively in the subseries that are within 6, 12, 18, 24 months from the baseline visit. With all lengths of subseries, ANSWERS produced significantly (<0.01%, Wilcoxon signed rank test) more accurate prediction with lower MAD (Table). The relative accuracy of ANSWERS over OLSLR is greater in shorter series. ANSWERS produced more accurate predictions across the range of change from baseline (Figure).

 
Conclusions
 

ANSWERS predicts future loss of visual field better than OLSLR across a wide range of series lengths, but especially with short series. The improvement is also evident regardless of the amount of progression.

 
 
Table. Prediction MAD of OLSLR and ANSWERS.
 
Table. Prediction MAD of OLSLR and ANSWERS.
 
 
Figure. MAD difference between OLSLR and ANSWERS stratified by difference between VF being predicted and baseline. Blue lines is linear regression. Positive value on y-axis indicates better prediction of ANSWERS.
 
Figure. MAD difference between OLSLR and ANSWERS stratified by difference between VF being predicted and baseline. Blue lines is linear regression. Positive value on y-axis indicates better prediction of ANSWERS.
 
Keywords: 642 perimetry • 550 imaging/image analysis: clinical • 758 visual fields  
×
×

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.

×