Abstract
Purpose :
To characterize the diagnostic accuracy and interobserver variability of macular spectral-domain optical coherence tomography (OCT) in patients presenting to an ophthalmic teaching hospital with suspected macular disease among eight retinal specialists.
Methods :
Four ophthalmologists and four optometrists analyzed 997 macular OCTs acquired with the 3D OCT - 2000 (Topcon, Japan) from a randomly selected retrospective cohort of 1000 patients presenting with a suspected macular problem to Moorfields Eye Hospital between June 2012 and January 2017. Scans were analysed by each specialist on two occasions in a random order at least one week apart - the first review consisted of the OCT alone while the second incorporated additional clinical information, such as age, visual acuity, fundus photo and a brief clinical narrative. Clinicians were asked to give a diagnosis of choroidal neovascularization (CNV), non-CNV related macular retinal edema (MRE), other pathology or normal. The gold standard diagnosis was the final diagnosis at a specialist retinal clinic following ancillary investigations, such as fluorescein angiography. Interobserver variability was assessed with the Fleiss kappa statistic. Diagnostic accuracy outcome measures were sensitivity, specificity, positive predictive value and negative predictive value. Pearson correlation coefficient was used to examine the association between years of experience and diagnostic accuracy.
Results :
Interobserver variability was excellent among both optometrists (κ 0.739) and ophthalmologists (κ 0.827) improving when OCT was combined with clinical information (optometrists: κ 0.812, ophthalmologists: κ 0.878). Diagnostic accuracy was generally high, particularly when combined with additional clinical information. For the sight-threatening pathology of CNV, sensitivity and specificity were respectively 89% and 99% for ophthalmologists and 86% and 97% for optometrists. A strong correlation was noted between years of experience and diagnostic accuracy (r = 0.86, p < 0.01).
Conclusions :
OCT is highly sensitive and specific in the diagnosis of macular disease, especially when combined with additional demographic and clinical details. Diagnostic accuracy appears to improve proportionally with years of retinal experience. There is strong agreement among experts in qualitative OCT interpretation across a range of macular pathology.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.