High-density OCT Analysis Detects Photoreceptor Thinning Prior to Hydroxychloroquine Retinal Toxicity Diagnosis

Abstract

Purpose : Diagnosis of Hydroxychloroquine (HCQ) retinal toxicity currently relies upon qualitative assessment of optical coherence tomography (OCT) outer retinal thinning. This study uses quantitative, high-density OCT cluster analysis to examine whether photoreceptor changes occur prior to HCQ retinal toxicity diagnosis.

Methods : Sixty-seven participants using HCQ and 134 participants not using HCQ were included in the study, propensity-score matched by age, sex, and refraction. All participants had normal macular health. One eligible eye per participant was randomly selected for study. Two graders blinded to participants’ group status manually segmented the outer plexiform and outer nuclear layers’ OCT borders, including Henle’s fibre layer (OPL/HFL/ONL). High-density gridwise thicknesses of the OPL/HFL/ONL were custom extracted then clustered between groups to reveal topographic differences.

Results : Comparing participants using HCQ versus not using HCQ, there was one cluster (C₋₁) of significantly thinned OPL/HFL/ONL (−1.6 [−2.76, −0.44] µm, P < 0.01) with small effect size (−0.25; Figure). The cluster of thinned OPL/HFL/ONL was primarily localised to the nasal quadrant (81.89% nasal versus 1% temporal quadrant area, P < 0.0001), and equally distributed in the superior and inferior quadrants (17.11% versus 17.56% respectively, P = 0.87).

Conclusions : Photoreceptor thinning prior to formal diagnosis of HCQ toxicity can be detected using quantitative, high-density OCT cluster analysis. This suggests current qualitative assessment of photoreceptor integrity in clinical practice may be inadequate for detection of HCQ retinal toxicity, and further implementation of quantitative measures could assist early detection.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

Figure. OPL/HFL/ONL topography between HCQ-usage versus non-HCQ-usage groups. Cluster differences (µm) presented graphically (left) and topographically (right). Z-score below x-axis. Significance: ** P < 0.01. Clusters labelled by rank/magnitude of difference from zero, i.e., C₀ = no significant difference from zero. Image in right eye format and scale at bottom right.