Abstract
Purpose :
Measurements of quantitative retinal vascular parameters (RVP) and their association to systemic factors and disease is an ongoing area of interest. Results may depend upon the centering of analyzed fundus images which often differs between studies. To assess whether this contributes to discrepancies in results, we aim to evaluate the effect of image centering on values of RVP.
Methods :
In a cross-sectional study of patients with center-involving diabetic macular edema (DME), two 45° macula- and disc-centered fundus images on the same eye were captured (Topcon 3D OCT-2000 Spectral domain OCT, Topcon, Tokyo, Japan). Patients with active proliferative diabetic retinopathy or retinal photocoagulation within four months were excluded. Central retinal artery (CRAE) and vein equivalent (CRVE) and arteriole-to-venule ratio(AVR) were measured (VAMPIRE 3.1, Universities of Dundee and Edinburgh, UK) and marked d (disc) and m (macular) according to centering.
Patient characteristics and RVP are reported as percent, mean±SD or median(IQR) as appropriate. Evaluation of correspondence between RVP was done by single-rater, consistency, two-way mixed intraclass correlation coefficients (ICC) with 95% CI. We used paired t-test and linear regression for evaluation of systematic and proportional bias, respectively.
Results :
Forty-one eyes of 31 patients were included. Patients characteristics were: 71.0% males, age 63±12 years, Best Corrected Visual Acuity 75(10) Early Treatment Diabetic Retinopathy Study letters, HbA1c 63±20 mmol/mol and central retinal thickness 360(95) µm.
RVP values were: CRAEd 143(25)µm and CRAEm 131(31)µm , CRVEd 213(17)µm and CRVEm 204(26)µm , AVRd 0.67±0.09 and AVRm 0.68±0.14. We report moderate correlation of CRAE and CRVE (ICC=0.59(0.34;0.76) and ICC=0.55(0.29;0.73)) and poor correlation of AVR (ICC=0.32(0.02;0.57)) between images. We found systematic bias of CRAE and CRVE (p<0.001), but no systematic bias of AVR (p=0.68). Proportional bias was present in AVR (p=0.003), but not in CRAE and CRVE (p=0.49 and p=0.41).
Conclusions :
Agreement between RVP in macula- and disc-centered fundus images in patients with DME was moderate to poor. This can result in diverging values of RVP, and standardization of image centering is needed for comparable results. Further studies are needed, assessing the implication of image centering on RVP association to systemic factors and disease.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.