ALI M. ALLAHDINA, MD, KATHERINE G. CHEN, MS, JASON A. ALVAREZ, MD,WAI T. WONG, MD, PHD, EMILY Y. CHEW, MD, CATHERINE A. CUKRAS, MD, PHD

Key words: hydroxychloroquine, retina, toxicity.

ABSTRACT
Purpose:To characterize functional and structural changes in hydroxychloroquine (HCQ) retinal toxicity after drug cessation.
Methods: Twenty-two patients (91% female; mean age 58.7 ± 11.4 years; mean duration of HCQ treatment 161.1 ± 90 months; mean dose 5.9 ± 1.9 mg/kg) with detected HCQ retinop- athy were monitored for 6 months to 82 months after HCQ cessation with multimodal imaging including spectral domain optical coherence tomography and fundus autofluorescence imaging at 488 nm (standard) and 787 nm (near-infrared autofluorescence). Tests of visual function including visual acuity, Humphrey visual ield testing, and multifocal electroretinography (mfERG) were performed. Study eyes were categorized into four separate severity stages by qualitative grading of spectral domain optical coherence tomography macular scans taken at the time of HCQ cessation. Changes in outcome measures between drug cessation and last follow-up visit were computed and compared between eyes of different severity stages.Results: Study eyes (n = 44) were categorized based on optical coherence tomography criteria infant immunization into: Stage 1 (subtle changes conined to parafoveal region; n = 14), Stage 2 (clear localized changes in parafovea; n = 17), Stage 3 (extensive parafoveal changes; n = 7), and Stage 4 (foveal involvement, n = 6). Visual acuity measurements across follow-up were stable in Stage 1 and Stage 2 eyes but decreased signiicantly in Stage 3 and 4 eyes. Humphrey visual ield measures were also stable in stages 1 and 2 but deteriorated in Stage 3 eyes. mfERG testing demonstrated signiicant improvement in the R1/R2 ratio after HCQ cessation in Stage 1 eyes (mean change = 20.86 ± 0.79, P = 0.03) but did not change signiicantly in eyes of higher stages.

Decreases in macular thickness in =1 of 9 Early Treatment Diabetic Retinopathy Study subields on spectral domain optical coherence tomography were found in eyes of all stages, with Stage 2 eyes demonstrating thinning in most subields (eight of nine subields). In eyes with a measurable central foveal ellipsoid zone band island (9 of 17 Stage 2 eyes and 7 of 7 Stage 3 eyes), progressive decrease in the foveal ellipsoid zone band length was observed in 6 of 9 (67%) Stage 2 eyes and 6 of 7 (86%) Stage 3 eyes. Changes indicative of progressing reti- nopathy were detected in 17% of Stage 1 eyes, 46% of Stage 2 eyes, and 43% of Stage 3 eyes on standard fundus autofluorescence imaging, and in 17% of Stage 1 eyes, 38% of Stage 2 eyes, and 14% of Stage 3 eyes on near-infrared autofluorescence imaging.Conclusion: Eyes with detected HCQ retinopathy do not demonstrate general stability in retinal structure and function after HCQ cessation but instead demonstrate a range of changes during follow-up whose magnitudes correlate with retinopathy severity at the time of cessation. After cessation, eyes with only subtle and localized retinopathy were mostly stable and may show some functional improvement, whereas more severely affected eyes continued to progress. These indings provide evidence that early detection and prompt cessation in HCQ retinopathy may be needed to arrest retinopathy progression and to optimize long-term outcomes.

Hydroxychloroquine (HCQ) is one of the mainstays of treatment for rheumatologic and inflammatory disorders but its use is limited by the potential risk to the retina.1,2 Although the exact mechanism of retinop- athy is still unclear,3–5 previous studies have shown thatthe major risk factors predisposing patients to the devel- opment of HCQ retinopathy includes a daily dose greater than 5.0 mg/kg, a duration of treatment longer than 5 years, suboptimal renal function, concomitant tamoxifen use, and preexisting macular disease.6 Although the incidence of HCQ retinopathy was previously thought to be low (<1%),7 recent studies have shown that the incidence rate is signiicantly higher in some populations, especially inpatients with longer durations of HCQ use.8 In 2016, the American Academy of Ophthalmology revised their recommen- dations for primary screening of HCQ retinopathy to include spectral domain optical coherence tomography (SD-OCT) testing in addition to Humphrey visual ield (HVF) testing. Spectral domain OCT imaging is an objective test that has been demonstrated to be both sensitive and speciic for HCQ toxicity. Furthermore, inspection of the SD-OCT images allows for the detec- tion of early pathologic changes well before the clin- ically recognizable bull ’s eye appearance of the fundus can be seen—a stage which is now considered late and severe, and in which there is a marked loss of visual function.

Hydroxychloroquine retinopathy is generally thought to be irreversible and has the potential to progress even after cessation of the drug.9 Early stud- ies showed that especially in patients with advanced- stage HCQ retinopathy, characterized by a severe ring scotoma on HVF or a bull ’s eye lesion on fundus examination, there is continued progression in drug toxicity, which can continue for many years after ces- sation of the drug.10,11 For patients with only mild HCQretinopathy, these early studies suggest that there is no progression of retinopathy after the drug is stop- ped.10,11 However, the early studies measured toxicity progression with subjective visual ield analysis and ophthalmoscopic examinations—a relatively insensitive modality in detecting HCQ-related changes in the macula. Additional knowledge about which eyes may continue to progress after drug cessation is needed not only for better patient counseling, but also for understanding the mechanism of these changes. More recent observations have directed attention to additional retina indings, which may be correlated with prognosis and progression. Marmor and Hu9 demonstrated that the presence of retinal pigment epi- thelium (RPE) damage was a predictor of retinopathy progression in an observational study on 11 cases of HCQ retinopathy, and Mititelu et al12 observed in an independent case series that the preservation of certain outer retinal structures, such as the external limiting membrane (ELM) on SD-OCT, carried a positive prognostic value.This study examined the changes in eyes with a range of HCQ retinopathy severity that occurred at the time of drug cessation using multimodal imaging and functional testing. We examined the long-term outcomes in eyes with HCQ toxicity and correlate the observed structural and functional changes with the severity of retinopathy.

This longitudinal observational study was conducted at the eye clinic at the National Eye Institute, National Institutes of Health, Bethesda, MD. Inclusion criteria for the study included 1) a current or previous history of HCQ treatment for a total duration exceeding 5 years; 2) a diagnosis of HCQ toxicity based on the totality of evidence on clinical, imaging, and functional studies; 3) cessation of HCQ at the time of enrollment; and 4) at least one follow-up visit with appropriate functional testing and multimodal imaging. Patients were excluded if they had any other maculopathies or retinal disorders (e.g., diabetic retinopathy, retinal vein occlusion, age-related macular degeneration, or Stargardt disease). Information on patient characteristics, including demographics, medical history, body weight and height, duration and cumulative dose of HCQ therapy, and diagnostic indications for HCQ treatment were obtained by medical history evaluation. The study protocol and informed consent forms were approved by the institutional review board of the National Institutes of Health, and the study was registered on clinicaltrials.gov (identiier NCT01145196). All study participants provided signed informed consent. The study protocol adhered to the tenets of the Declaration of Hel- sinki and complied with the Health Insurance Portability and Accountability Act.All participants underwent a comprehensive ocular examination at baseline and each follow-up study visit, including best-corrected visual acuity (BCVA) testing using the Early Treatment Diabetic Retinopathy Study (ETDRS) protocol, slit-lamp examination, and dilated fundus examination. In addition, all patients under- went multifocal electroretinography (mfERG) testing, HVF testing, and retinal imaging including SD-OCT,and fundus autofluorescence (FAF) imaging at 488 nm (standard) and 787 nm (near-infrared autofluorescence [NIA]). Testing was performed in both eyes of all participants.Perimetric assessment was performed using a stan- dard 10-2 HVF Analyzer (Carl Zeiss Meditec, Inc, Dublin, CA) with a white test spot. The HVF mean deviation (MD) values, representing the overall mean departure of sensitivity at speciic retinal points compared with age-corrected normal values, were obtained from the visual ield output.

mfERG testing was performed according to the International Society for Clinical Electrophysiology of Vision guidelines13 based on the 61-hexagon stimulus pattern of the VERIS Clinic system (Electro-Diagnostic Imaging, Inc, Redwood, CA). The 61 hexagon responses were grouped into 5 concentric rings (R1–R5). The aver- age amplitude was assessed for each ring outside the R1 hexagon. The average response densities (nanovolts per degrees squared) within concentric rings from the center (Ring 1) to the periphery (Ring 5) were generated by the mfERG VERIS software. The ring ratios of the mfERG were deined as ratios of the central hexagon amplitude (R1) to each of the peripheral ring amplitudes (R2–R5). These ratios were calculated for all tested eyes.

Spectral Domain Optical Coherence Tomography Imaging and Analysis
Foveal-centered SD-OCT volumes were obtained for both eyes from each participant on the Cirrus-HD system (Carl Zeiss Meditec, Inc) using the macular cube 512 · 128 scan pattern. The macular thickness map was divided into 9 subields based on the ETDRS grading grid. Mean retinal thicknesses in each of the 9 subields were generated by the manufacturer’s soft- ware version 6.5.0.772 (Carl Zeiss Meditec, Inc).
The SD-OCT images were also acquired in parallel at every visit using the Heidelberg Spectralis HRA and OCT system (Spectralis; Heidelberg Engineering, Heidelberg, Germany). Horizontal 30° images through the fovea with 100 averaged scans were graded man- ually for the presence or absence of anatomical dis- ruptions in the perifoveal ellipsoid zone (EZ) and ELM. In cases where the quality of the Spectralis im- ages was insuficient to visualize this region clearly (1 of 22 participants), corresponding Cirrus HD-OCT im- ages were graded in their place. Two independent readers (A.M.A. and K.G.C.) performed the grading in a masked fashion, with any discordant grades resolved by consensus after joint review with the third reader (C.A.C.). For cases in which EZ and/or ELM disruption on either side of the fovea was present, the length of the disruption was quantiied using ImageJ software version 1.80_72 (U.S. National Institutes of Health; available at http://imagejnihgov/ij/; accessed November 19, 2017). For cases in which ELM and or EZ disruptions were present on both sides of the fovea, the full width of the intact central ELM and/or EZ was measured (see Figure 1, Supplemental Dig- ital Content 1, http://links.lww.com/IAE/A949).
Fundus Autofluorescence Imaging

Standard FAF images (excitation, 488 nm; emis- sion, >500 nm) and NIA images (excitation, 787 nm; emission, >800 nm) were obtained with the Spectralis scanning laser ophthalmoscope (Heidelberg Engineer- ing). Images were presented to graders masked to par- ticipant identiication, and a consensus step grade (deined as the grade demonstrating agreement between =2 of 4 graders [A.M.A., K.G.C., J.A.A., and C.A.C.]) was obtained for each eye on eachimag- ing modality. Grading was based on a previously pub- lished scale by Marmor14 (see Figure 2, Supplemental Digital Content 2, http://links.lww. com/IAE/A950). A score of 0 to 4 was given for each image based on the following criteria: 0 = normal, 1 = patchy damage (1–2 quadrants), 2 = bull ’s eye damage (>2 quadrants), 3 = bull ’s eye damage involving fovea or RPE, and 4 = diffuse posterior pole damage. For images in which there was discrepancy between graders, a consensus grade was agreed upon after joint review of the masked images.Definition of Hydroxychloroquine Severity StagesFor data analysis, the study cohort was divided into four groups based on qualitative grading of a one-line horizontal SD-OCT scan through the fovea at the time of drug cessation. We used previously established OCT criteria15 as a framework to deine and to desig- nate severity stages 1 through 4. Stage 1 eyes had subtle, but not deinite disruption of foveal or parafo- veal EZ. Stage 2 eyes had deinite buy E7080 localized changes in the parafovea on one or both sides of the fovea, a return of the EZ in the retina anterior to the disruption in the parafoveal region, and was without foveal EZ disrup- tion. Stage 3 eyes had extensive parafoveal EZ disrup- tion but with a retained intact EZ in the fovea. Stage 4 eyes had EZ disruption involving the fovea. An exam- ple of SD-OCT images from each severity stage is shown in Figure 1.

Fig.1. Examples of SD-OCT
severity stages in patients with HCQ toxicity and longitudinal change. A diagnosis of HCQ toxicity based on the totality of evidence on clinical, imaging, and functional studies.The severity stage of study eye was then determined by assessing outer retinal structures on a foveal-centered horizontal OCT scan. Stage 1 eye shows subtle but not deinite disruption of parafoveal EZ. Stage 2 eye with deinite parafoveal EZ disruption (on both sides of the fovea = extensive), a return of the EZ in the retina anterior to the disruption in the parafoveal region, and without foveal EZ disruption. Stage 3 eye demon- strates extensive parafoveal EZ disruption but retains an intact EZ in the fovea. Stage 4 eye with EZ disruption involving the fovea.Outcome measures for functional testing (BCVA, HVF, and mfERG) and multimodal imaging (SD- OCT, FAF, and NIA) were examined at each patient’s initial (time of drug cessation) and follow-up visit. The change over time (most recent visit vs. baseline visit) was compared and analyzed across severity stages.All data were analyzed using nonparametric statistics computed with GraphPad Prism 7.0 for Windows (GraphPad Software, La Jolla, CA). Continuous vari- ables were analyzed with the Mann-Whitney test and the Kruskal-Wallis test was used for analysis of two or more discontinuous variables. For all tests, P < 0.05 was considered statistically signiicant.

Results
Forty-four eyes of 22 patients (91% female; 2 female patients of Asian descent) with HCQ toxicity were followed by repeat examinations over 6 months to 82 months in this observational study. The mean age of the patients at time of HCQ cessation was 58.7 years (age range, 34–78), and the mean treatment duration before cessation was 161.1 months (range, 35–375 months). The mean daily dose by body weight was 5.9 mg/kg (range, 1.2–8.3 mg/kg). Using our OCT-deined severity stages, 14 eyes were designated as Stage 1, 17 eyes as Stage 2, 7 eyes as Stage 3, and 6 eyes as Stage 4. Table 1 details the demographic char- acteristics of patients in this study, by severity stage at the time of drug cessation (baseline).At baseline, the mean visual acuity of all eyes was 78 ± 13 letters (Snellen equivalent = 20/32, logMAR 0.20). Stages 1 to 3 had similar visual acuities: Stage 1, 86 ± 5.6 letters (Snellen equivalent = 20/20, log- MAR 0.00); Stage 2, 81 ± 6.5 letters (Snellen equiv- alent = 20/25,logMAR 0.10); Stage 3, 81 ± 3.9 letters, Snellen equivalent = 20/25 (logMAR 0.10). However, Stage 4 eyes had signiicantly worse visual acuity with mean of 50 ± 10 letters (Snellen equivalent = 20/100, logMAR 0.70) (Figure 2A). The change in BCVA, representing the most recent BCVA minus initial BCVA, was calculated for all study eyes and seemed to remain stable over time for stages 1 to 3 eyes, but declined in Stage 4 eyes (mean Δ BCVA: Stage 1, 21.4 ± 6.6 letters; Stage 2, 21.1 ± 5.5 letters; Stage 3, 23.3 ± 4.5 letters; Stage 4, 228 ± 29 letters) (Fig- ure 2B). Two Stage 4 eyes were unable to read the chart at their most recent visits (BCVA = 0 letters).Eyes were also examined with static perimetry testing at last follow-up and at the time of drug cessation. Four Stage 4 eyes were unable to complete the baseline HVF testing, and 2 additional Stage 4 eyes were unable to complete testing at follow-up visits and thus were not included in this analysis. Of the remaining 38 eyes, HVF testing had a negative mean MD (27.1 ± 5.4 dB) at the time of drug cessa- tion demonstrating lower than normal sensitivity and central visual ield loss. Eyes with increasing HCQ severity stages had worse baseline visual ield mean MD (Stage 1, n = 14, 23.2 ± 2.1 dB; Stage 2, n = 17, 28.1 ± 5.2 dB; Stage 3, n = 5, 212 ± 6.2 dB; Stage 4, n = 2, 212 ± 3.1) (Figure 2C). On follow-up exami- nation, Stage 3 eyes showed decreases in mean MD over time (ΔMD = 22.5 ± 2.5 dB) indicating progres- sive ield loss, whereas stages 1 and 2 eyes were rel- atively stable (mean ΔMD: Stage 1, 0.82 ± 1.9 dB; Stage 2, 0.23 ± 1.5 dB) (Figure 2D). An example of the change in visual ield maps of a Stage 1 and Stage 3 eye is shown in Figure 3, A and B, respectively.

Functional testing assessed with mfERG was also performed at the time of cessation and at the most recent follow-up in 24 eyes (Stage 1, n = 7; Stage 2, n = 12; Stage 3, n = 5), and the change in R1 and change in the ratio of R1/R2 were calculated.16 Four Stage 1 eyes and 5 Stage 2 eyes showed functional improvement on their most recent follow-up and no longer met the criteria for toxicity in the R1 or R1/R2 ratio16 (Figure 4, A and B). One eye in Stage 3 showed improvement on follow-up examination whereas a majority in this cohort (80%) showed a decline in R1 amplitude and R1/R2 ratio. Stage 4 eyes were not followed longitudinally with mfERG due to an inability to measure responses reliably, given the worsening vision and noncentral ixation in these patients.Coherence Tomography We examined quantitative measures of retinal thickness as measured by OCT over time in eyes with toxicity. We found that when compared to normative values,17,18 eyes with a greater severity of HCQ toxicity at baseline also had increased retinal thinning in particular ETDRS subields. Retinal thin- ning was most apparent in the center and inner infe- rior subields (Table 2). On follow-up examination, a statistically signiicant change was observed in all 4 severity stages, with Stage 2 eyes demonstrating change in 8 of the 9 subields. Although Stage 3 and Stage 4 demonstrated changes in retinal thickness in fewer subields, progressive loss (retinal thinning) occurred in the center subield of these more severe stages (Table 2).We reviewed autofluorescent images for all pa- tients and evaluated the presence of RPE damage,19 deined as areas of hypoautofluorescence on either FAF or NIA imaging, at the time of HCQ cessation and at follow-up to investigate the relationship between RPE integrity and progression of retinal changes.

Using a reference FAF and NIA grading scale (see Figure 2, Supplemental Digital Content 2, http:// links.lww.com/IAE/A950), we assessed images of 38 eyes at the time of HCQ cessation and at the most recent follow-up visit. We found that eyes with increasing severity grades of HCQ toxicity (as deter- mined by our previously deined OCT grading scale) had a higher score on both the FAF grading scale and NIA grading scale (Figure 2, E and G). A comparison of grading of en face images produced by the two imaging modalities on the same eye depicted in a grid indicates that more severe indings are seen on NIA than on FAF (see Figure 2, Supplemental Digital Content 2, http://links.lww.com/IAE/A950).
At the time of HCQ cessation, 20 eyes with RPE damage were identiied by inspection of autofluores- cence images: 2 Stage 1 eyes, 8 Stage 2 eyes, 4 Stage 3 eyes, and 6 Stage 4 eyes. All 20 eyes demonstrated progression of RPE damage on follow-up autofluor- escence imaging and all but 2 eyes demonstrated

Fig.2.Baseline and changes of visual acuity, visual ield, and autofluorescence grading. A. Best-corrected visual acuity at the time of drug cessation (baseline) compared across all four severity stages. The values are represented in total letters and the Snellen equivalent is noted in parentheses.B.The change in the BCVA, representing the most recent BCVA minus the initial BCVA, is compared across all four severity groups. C. The result of the HVF as mean deviation at the time of cessation are compared across Stages 1 to 3. Stage 4 was not included in the analysis as they could not ixate at the fovea. D. The change in MD, representing the most recent HVF MD minus the initial MD,is compared across severity Stages 1 to 3. E. The results of the FAF image score at the time of cessation are compared across all four severity stages. The mean FAF score is associated with increasing severity stage. F. The change in the Anti-hepatocarcinoma effect FAF image score, represent- ing the most recent score minus the initial score, is compared across all four severity groups. G. The results of the NIA image score at the time of cessation are compared across all four severity stages. The mean NIA score is associated with increasing severity stage. H. The change in the NIA image score, represent- ing the most recent score minus the initial score, is compared across all four severity groupS progression (i.e., worsening) of retinal function. Eyes in Stage 2 and Stage 3 categories demonstrated the most change in FAF scores (54% and 43% of eyes had an increase in FAF score at most recent follow-up visit, respectively) (Figure 2F). One Stage 1 eye dem- onstrated improvement on FAF score at the most recent follow-up visit (FAF = 0) compared with base- line (FAF = 1).

Fig. 3.Longitudinal changes in a Stage 1 eye and a Stage 3 eye on functional tests and multimodal imaging. A. 10-2 HVF of a Stage 1 eye (patient 015) at the time of drug cessation (left) and 55 months later at the mostrecent visit (right). B. 10-2 HVF of a Stage 3 eye (patient 013) at the time of drug cessation (left)and 62 months later at the most recent visit (right). The initial HVF shows decreased sensitivity at certain retinal points. On follow-up, the HVF results show progression of visual ield loss. C. Spectral domain optical coherence tomography horizontal scan of a Stage 1 eye (patient 015) at the time of drug cessation (top) and 55 months later at the mostrecent visit (bottom). D. Spectral domain optical coherence tomography horizontal scan of a Stage 3 eye (patient 013) at the time of drug cessation (top) and 62 months later at the most recent visit (bottom). The horizontal scan at follow-up shows further collapse of outer retinal structures including a decrease in the size of the EZ island in the central macula. E. En face FAF image of a Stage 1 eye (patient 015) at the time of drug cessation (left) and 55 months later at the most recent visit (right). The FAF score remains stable over time. F. En face FAF image of a Stage 3 eye (patient 013) at the time of drug cessation (left) and 62 months later at the most recent visit (right). Hypofluorescent areas in the peripheral retina demonstrate areas of RPE damage. G. En face NIA image of a Stage 1 eye (patient 015) at the time of drug cessation (left) and 55 months later at the mostrecent visit (right). The NIA score remains stable over time. H. En face NIA image of a Stage 3 eye (patient 013) at the time of drug cessation (left) and 62 months later at the mostrecent visit (right). The NIA score does not change as the initial visit was scored at the maximum score.

Fig.4. Multifocal electroreti-
nography at the time of drug cessation and change at the most recent visit. A. Stage 1 eyes’ mfERG test results with R1 amplitude and R1/R2 ratio at the time of drug cessation (closed circles) and the most recent fol- low-up (open circles). R1 data below the published threshold values (*R1 line at 21.6 nano- volts/degree)2 represent lower limit of normal for ages 50 years and older16 and R1/R2 ratio data above published threshold value16 (2.6 across ages)are contained within the shaded region of the graph. Data points in the nonshaded region of the graph represent mfERG values deemed in the normal range. Four eyes demonstrated a return to mfERG values in the normal range on follow-up visit. B. Stage 2 eyes ’ mfERG test results with R1 amplitudes and R1/R2 ratios at the time of drug cessation (closed circles) and the mostrecent follow-up (open circles). Five eyes demonstrated a return to mfERG values in the normal range on follow-up visit. Stage 3 and Stage 4 eyes are not included in this igure; not all Stage 3 and 4 eyes were able to complete themfERG test on follow-up visits.Of those eyes that did not have evidence of hypoautofluorescence abnormalities on FAF imaging at the time of HCQ cessation, 5 eyes (2 Stage 2 and 3 Stage 3) demonstrated progression with RPE damage present on follow-up images.Using NIA imaging, we observed the largest change in NIA scores in Stage 2 eyes (Figure 2H). As with FAF grading, we also encountered limitations of the NIA grading scale because most Stage 3 and Stage 4 eyes were at the scale ceiling at the time of HCQ cessation (NIA score = 4 in 86% of Stage 3 eyes and 67% of Stage 4 eyes, Figure 2G). An example of the changes observed in a Stage 1 and Stage 3 eye can be seen in Figure 3, E–H.

Horizontal SD-OCT images through the fovea of 42 eyes were examined at the time of cessation to quantify EZ disruption in the parafovea and the EZ length overlying the fovea in cases where the disrup- tion was present on both sides of the fovea. The measurements were also performed on the mostrecent visit to quantify the amount of change over time.Examination of Stage 1 eyes (n = 14) at baseline demonstrated an intact EZ throughout the 30° scan, which remained present for all Stage 1 eyes on the most recent scan. The reflectivity of the EZ line on follow-up images of 2 of the 14 eyes showed qualita- tive improvement (brighter, unequivocally intact). In addition, 2 eyes that had focal disruption to the ELM in the parafoveal region at baseline demonstrated an intact ELM across the entire horizontal scan on follow- up visit. Examples of a Stage 1 eye with stable EZ integrity on follow-up can be seen in Figures 1 and 3C.Examination of Stage 2 eyes (n = 15) at the time of cessation revealed 2 phenotypic subgroups based on the extent of the EZ loss. Eyes that had focal disruption of the EZ in the parafoveal region in either the nasal side or the temporal side, but not both, were labeled as the focal subgroup (n = 6). Eyes that had EZ disruption on both sides of the parafoveal region were labeled as the exten- sive subgroup (n = 9). We found that eyes with focal EZ disruption at baseline showed statistically signiicant improvement in the measured gap, the length of EZ disruption, on follow-up imaging (baseline mean EZ gap = 730 ± 399 μm; mean ΔEZ gap = 2369 ± 326 μm, P = 0.03) (Figure 5, A and B). Eyes with extensive EZ disruption demonstrated a range of outcomes on the mostrecent visit. Of the nine eyes with extensive gaps, two eyes showed an improvement (decrease) of the EZ gap on both sides, one eye showed stable EZ gap on both sides, three eyes showed worsening (increase) of the EZ gaps, and three eyes showed one side with stable EZ gap while the other side had worsening (increased) EZ gap (baseline mean EZ gap = 1,449 ± 1,072 μm; follow-up mean ΔEZ gap = 261 ± 504 μm) (Figure 5, A and B). In addition, we measured the length of the intact EZ over the fovea (EZ island) for eyes with extensive EZ disrup- tion and compared the baseline to the mostrecent visit. On follow-up visits, most eyes showed a worsening or shortening of the EZ length over the fovea; six eyes showed a decrease in EZ island, one eye showed a stable EZ island, and two eyes had improvement or lengthening of the EZ island (Figure 5, C and D).

Examination of Stage 3 eyes (n = 7) at baseline and follow-up revealed a statistically signiicant decrease in the foveal EZ island length over time (baseline mean foveal EZ length = 2,564 ± 1,241 μm; mean Δ foveal EZ length = 2297 ± 286 μm, P = 0.03) (Figure 5, C and D). An example of the shortening of the EZ island can be seen in Figure 3D. Examina- tion of Stage 4 eyes (n = 6) showed an absence of foveal EZ with further collapse of the outer nuclear layer in the fovealand parafoveal regions over time for all eyes in this group.We evaluated the features of the RPE and ELM in the context of the OCT severity–based grading scale (Figure 1). Almost all eyes with Stage 1 severity were without RPE damage or ELM disruption. All Stage 4 eyes and all but one Stage 3 eye had RPE damage and ELM disruption (one Stage 3 eye had ELM disruption but was without RPE damage). Stage 2 eyes showed the most heterogeneity, with about one-third (6/17) showing no evidence of RPE damage or ELM disruption, about one-third (7/17) having both RPE damage and ELM disruption, 2/17 having evidence of RPE damage without ELM dis- ruption, and 2/17 having evidence of ELM disrup- tion without RPE damage.Functionally, most eyes classiied as Stage 1 at the time of cessation remained stable, with some eyes even showing objective improvement on imaging and visual function. Interestingly, the 2 Stage 1 eyes that showed mild progression on autofluorescence and mild progression of functional loss revealed that they had an interruption to the ELM on baseline OCT imaging at the time of drug cessation. Eyes classiied as Stage 3 and Stage 4 with EZ and ELM loss as well as RPE damage all demonstrated progression of retinopathy on follow-up visits. Most Stage 2 eyes progressed on structural and functional measures. However, 2 of the Stage 2 eyes did not progress. A closer examination of the baseline horizontal foveal-OCT scan with attention to the ELM reveals that the ELM was intact over the entire horizontal foveal-OCT scan.

Discussion
In this study,we characterize the structural and functional changes that occur in HCQretinopathy after cessation of the drug in a larger cohort of patients

Fig. 5. A and B. Ellipsoid zone
disruption measurements in Stage 2 eyes at the time of drug cessation and change at the most recent visit. A. Ellipsoid zone gap measurements at the time of drug cessation for Stage 2 eyes with focal disruption (EZ gap on either nasal or temporal parafo- vea, but not both) and Stage 2 eyes with extensive disruption (EZ gap on both nasal and temporal parafovea). Measure- ment of both nasal and temporal disruption,if present, is included in this plot. B. The change in the EZ disruption, representing EZ disruption measurement at the most recent visit minus initial visit, is com- pared for the focal cohort and extensive cohort of Stage 2 eyes.C and D. Ellipsoid zone island measurements of Stage 2 and Stage 3 eyes at the time of drug cessation and change at the most recent visit. C. Ellipsoid zone island measurements at the time of drug cessation in Stage 2 eyes, with extensive EZ disrup- tion, and Stage 3 eyes. D. The results of change in intact EZ band for Stage 2 eyes, with extensive disruption, and Stage 3 eyes using multimodal imaging including SD-OCT, FAF, and tests of visual function including visual acuity, visual ields, and mfERG. Consistent with previous studies,12,15,20–22 we found that these eyes are not static but demonstrate a range of changes on follow- up testing. The eyes with the least severe retinopathy (Stage 1 and the focal subgroup of Stage 2) demon- strate a low likelihood of progression of HCQ retinop- athy and even display evidence of functional and structural improvement. However, eyes with more severe retinopathy (Stage 3 and Stage 4) demonstrate progressive deterioration of outer retinal structures and a decline in function even years after cessation of the drug.

Although we generally observe that eyes with more severe damage progress in their pathology more than those that have less severe damage at cessation, the prognosis of eyes in a given severity group cannot be entirely predicted by the extent of EZ damage alone. In addition to EZ damage, information on the health of the RPE as gained from autofluorescence imaging provides important prognostic information. The pres- ence of RPE damage correlates with eyes that will progress in their pathology, and we ind that in nearly all cases of RPE damage (18/20), we observe sub- sequent progression on follow-up. However, even in eyes that do not demonstrate observable damage to the RPE at the time of drug cessation, progression of EZ loss can be observed, although less commonly. Further analysis of these eyes with earlier stages of toxicity at the time of cessation reveals that the ELM provides additional information that seems to correlate with progression. The results of our study also support previous observations12,20 that preservation of the ELM carries a positive prognostic value. An intact ELM at baseline may even be associated with improvement in measures of retinal structure and func- tion possibly due to photoreceptor outer segment regeneration after cessation of the drug.Additional modalities provide added speciicity to the indings and diagnosis and also aid in the understanding of the pathophysiology of this process. We can extrapolate on the cross-sectional studies to develop a hypothesis of the nature of the pathologic insult incurred by HCQ toxicity. In the earliest observable (least severe) signs of toxicity, elements of the outer retina—within the outer nuclear layer— can be seen to demonstrate abnormalities sometimes even before there are signs of frank EZ loss, ELM disruption, or RPE abnormalities.15,21,23 In slightly more severe pathology, the EZ is observed to be dis- rupted.9,15,21,23 Later, in more severe cases, the ELM becomes disrupted and RPE demonstrates abnormality on autofluorescence imaging.
Previous reports have described some evidence of improvement in early toxicity after cessation, with reports of OCT appearing to demonstrate some reconstitution of the EZ.9,12 Some reports indcorrob- oration of functional improvement with positive changes on HVF testing.12,20 Whether these positive changes are variations in measurements or true evidence of biologic improvement will require further study; however, the corroborating evidence of multi- modal assessments make variability a less likely expla- nation. Our data also suggest evidence of structural improvement bolstered by changes on the OCT that were quantiiable on measured reflectivity bands. Evi- dence of functional improvement came from mfERG ring ratio analyses in some cases. Because ring ratios could potentially obscure a reduction in R1 reflected as a normalization of the ring ratio, these results must be carefully interpreted.

In our longitudinal follow-up, we ind that in both cases of ELM disruption and RPE abnormalities, retinal pathology progresses, as some biologic tipping point is reached. We observe that once cellular damage reaches a certain point, the retina cannot withstand the insult and continues to deteriorate even when the drug is removed. This point of tissue tipping appears to be when the ELM is disrupted, as observed on inspection of SD-OCT, and certainly once the RPE appears altered on autofluorescence imaging. Before ELM is disrupted, the retina can seemingly stabilize itself when the drug is stopped. However, when the ELM is disrupted, the retina can no longer maintain itself and damage progresses as in a retinal degeneration. The ELM represents the interaction between Müller cells and photoreceptors and this interaction is believed to be vital in maintaining support for rods and cones.24 In many degenerative diseases of the macula, the ELM can persist even with EZ disrup- tion.24 Recent studies on OCT analysis of macular ELM demonstrate that although abnormal imaging characteristics are observed, presence of the ELM sug- gests the presence of functioning photoreceptors.25,26 In addition, the ELM may be an indicator of reclaim- able photoreceptors in areas of relatively uncollapsed retinal structure.26 For patients with HCQretinopathy, an intact ELM at the time of drug cessation may be associated with the potential for improvement in meas- ures of retinal structure and function due to photore- ceptor outer segment regeneration after cessation of the drug.The indings in our study provide evidence for the clinical importance of early detection and prompt cessation to optimize long-term outcomes. Eyes with more severe damage at the time of cessation not only have worse visual function when identiied, but are at high risk of continued progression and worsening even years after taking HCQ. Although there is no one best test for HCQ screening, the utilization of information gained with SD-OCT, which is now a ubiquitous instrument in ophthalmologist’s ofice, is central to the evaluation for toxicity. Categorization of the severity of toxicity based on OCT indings may also be useful in guiding prognosis and patient expectations.