Purchase this article with an account.
Giovanni Montesano, Luca Mario Rossetti, Paolo Fogagnolo, Francesco Oddone, Paolo Lanzetta, Andrea Perdicchi, Chris A Johnson, David F Garway-Heath, David P Crabb; Effects of fundus tracking on structure-function relationship in glaucoma. Invest. Ophthalmol. Vis. Sci. 2019;60(9):3922. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To investigate if fundus perimetry can improve the diagnostic precision of structure-function (SF) metrics in glaucoma.
We analysed data from 1009 eyes of 244 healthy subjects and 329 with glaucoma from the validation study of the Compass fundus perimeter (CMP, CenterVue). All subjects were tested with a 24-2 grid with CMP and with the Humphrey Field Analyzer (HFA, Zeiss Meditec) and underwent an OCT scan of the peripapillary Retinal Nerve Fibre Layer (RNFL). SF correlation was assessed between the mean deviation (MD) of six visual field clusters and the corresponding RNFL sector thickness using a multivariate mixed effect model.For discrimination analysis, we used global MD, average RNFL thickness and a combined SF Index (SFI) obtained as the predicted logit from a logistic regression with the average RNFL thickness and the MD as predictors and the binary classification (Healthy/Glaucoma) as the response. A fixed effect accounted for systematic differences between the two SD-OCT devices used in different centres, the Spectralis (Heidelberg Engineering) and the RTVue XR Avanti (Optovue). We balanced the Healthy:Glaucoma proportions to be 1:1 with each OCT device by random constrained sampling, leaving a total of 786 eyes. We calculated partial Receiver Operating Characteristic (pROC) curves (minimum specificity at 75%, Figure 2) and Areas Under the Curve (pAUCs). Confidence intervals (CIs) and p-values were computed via paired bootstrap.
The R2 of the SF model was the same (0.47) for HFA and for CMP (Figure 1). The pAUC was greater with MD from CMP than HFA (p < 0.001). There was no statistically significant difference between the CMP SFI and the HFA SFI (p = 0.075). The pAUC was significantly higher with both the CMP SFI (p = 0.002) and HFA SFI (p = 0.017) compared to the RNFL alone and MD alone (p < 0.001).
Metrics from CMP and HFA have similar SF relationship. Functional data from perimetry can be integrated with structural information to significantly increase discrimination ability of structure and function alone. No evidence could be found that fixation stabilization significantly increased diagnostic precision.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
Figure 1. SF relationship with the CMP (left) and the HFA (right)
Figure 2. pROC curves and relative CIs. HFA Glaucoma Hemifield Test (GHT) is reported for reference
This PDF is available to Subscribers Only