To investigate the application of ultra-widefield fundus autofluorescence (UWF-FAF) imaging in evaluating hydroxychloroquine (HCQ) retinopathy and to report peripheral autofluorescence findings in Asian patients with this condition.
Retrospective case series.
Setting: institutional. Patient population: 58 eyes of 29 patients with HCQ retinopathy. Observation procedures: UWF-FAF imaging was performed, and the images were compared to conventional FAF images obtained using a confocal digital ophthalmoscope. The sensitivities of detecting retinopathy using the 2 modalities were compared. Peripheral autofluorescence findings in the eyes with HCQ retinopathy were assessed, and their association with the Humphrey visual field test results obtained using the 30-2 and full-field 120 (FF-120) protocols was analyzed. Main outcome measurements were abnormal FAF findings.
In 41 of 58 eyes (70.7%) with HCQ retinopathy, abnormal FAF findings were noted in the retinal periphery outside the field of view of conventional FAF as hypoautofluorescent (23 eyes, 39.7%) and hyperautofluorescent (38 eyes, 65.5%) lesions. In 5 eyes (8.6%), differences were revealed between conventional FAF and UWF-FAF in detecting retinopathy. Most of the eyes with severe retinopathy showed the most extensive hypoautofluorescence in the nasal peripheral retina. The areas with abnormal FAF findings were significantly correlated with the number of unseen spots on FF-120 results and mean deviation and pattern standard deviation of the 30-2 test results (all P < .001).
Peripheral autofluorescence findings varied in eyes with HCQ retinopathy according to the severity of the retinopathy. The retinal findings with UWF-FAF were functionally correlated to visual field results. UWF-FAF may be useful for evaluating HCQ retinopathy, particularly in Asian patients.
Hydroxychloroquine (HCQ) is a 4-aminoquinoline derivative used for the treatment of rheumatologic and dermatologic disorders such as systemic lupus erythematosus and rheumatoid arthritis. HCQ retinopathy refers to retinal toxicity characterized by photoreceptor and/or retinal pigment epithelium damage, which usually occurs in long-term users of HCQ who have often received excessive daily doses. , HCQ is known to be relatively safer than chloroquine, but the advanced, severe stage of HCQ retinopathy can lead to blindness. Its progressive, irreversible nature suggests that patients taking HCQ should be appropriately screened for early detection. ,
Retinal toxicity caused by HCQ varies according to the involved retinal area. Eyes with HCQ retinopathy can present with parafoveal, pericentral, or mixed pattern of retinopathy. Interestingly, specific patterns of retinopathy are observed in some ethnic groups (e.g., Asian patients with the pericentral pattern). Accordingly, the American Academy of Ophthalmology (AAO) guidelines recommend the use of wide-field optical coherence tomography (OCT) and fundus autofluorescence (FAF) imaging for the detection of pericentral retinopathy particularly in Asian patients taking HCQ. However, as current screening modalities have limitations in covering the whole retinal area, the extent of HCQ retinopathy and how peripheral findings affect the visual field of the patients remain unclear. Thus, retinal damage occurring in the retinal periphery should be investigated in eyes with HCQ retinopathy; however, peripheral findings of HCQ retinopathy have not been evaluated extensively.
There are several types of commercially available FAF imaging systems, including the fundus camera, confocal scanning laser ophthalmoscope, and ultra-widefield FAF (UWF-FAF). UWF-FAF is an imaging modality that can cover 200° or 82% of the retina within a single image. , In contrast to conventional FAF, which only covers a limited area (30°–50°), UWF-FAF could potentially visualize damage in the entire retina in eyes with HCQ retinopathy. This study evaluated the usefulness of the imaging modality by comparing it with conventional FAF. Furthermore, the study sought to evaluate retinal damage in the retinal periphery and to present the patterns of retinopathy in the whole retina of eyes with HCQ retinopathy.
The patients included in this study were recruited from the authors’ cohort with HCQ retinopathy, which consisted of 42 patients with diagnoses of HCQ retinal toxicity, at Hanyang University Hospital from January 2012 to January 2019. Among the patients who underwent both conventional scanning laser ophthalmoscope FAF (40°) and UWF-FAF (200°) at the same visit, cases were excluded with (1) a history of ocular trauma or intraocular surgery other than cataract extraction or other coexistent macular or retinal diseases such as age-related macular degeneration and retinal/choroidal vascular diseases; (2) a history of the use of other drugs that could induce retinal toxicity (e.g., tamoxifen); or (3) a family history of inherited retinal diseases (e.g., retinitis pigmentosa). Finally, 58 eyes from 29 patients with HCQ retinopathy were included in this study.
The diagnosis of HCQ retinopathy was based on the recent guidelines of the AAO. Specifically, at least 1 objective test abnormality (i.e., OCT, FAF, or multifocal electroretinogram) confirming the subjective test abnormality (i.e., automated visual field examination) was used for the diagnosis of HCQ retinopathy. , In all the included patients, characteristic photoreceptor defects (disruption or loss of the ellipsoid zone and/or external limiting membrane lines) with or without retinal pigment epithelium (RPE) involvement were confirmed on OCT B-scans. Visual field abnormality corresponding to photoreceptor loss identified on OCT was also confirmed by using at least 1 Humphrey visual field (HVF) protocol, either 10-2, 30-2, or full-field 120.
All patients received a comprehensive ophthalmic examination, including slit-lamp examination, noncontact tonometry, and indirect fundoscopy. Swept-source OCT (DRI-OCT; Topcon Inc., Tokyo, Japan) was performed in all patients. FAF was performed using an F-10 confocal scanning laser ophthalmoscope (cSLO; Nidek, Gamagori, Japan) after patients’ eyes were dilated using tropicamide 0.5% and phenylephrine 0.5%. UWF-FAF imaging was performed using an Optos 200 Tx (Optos PLC, Dunfermline, United Kingdom) retinal imaging device, which reaches 200° using a 532-nm excitation. Images were automatically corrected for 3- to 2-dimensional projection errors using the V2 Vantage Pro software provided by the manufacturer. The optical resolutions of the authors’ FAF devices were 16–20 μm and 14–20 μm for standard FAF and UWF-FAF, respectively, indicating similar image resolution for both FAF devices. The HVF examination (Zeiss Meditec., Dublin, California) was performed using the 30-2 protocol in all eyes and full-field 120 (FF-120) protocol in 22 eyes. Interpretation of all images was performed by 2 independent reviewers (S.J.A. and J.J.), who were blinded to the clinical information.
The extent of abnormal autofluorescence, hypo- and hyperautofluorescence, was quantified by measuring the area of hypo- and hyperautofluorescence in the image by using the software provided by the manufacturer, as shown in Figure 1 . The FAF abnormalities were identified by the reviewers, who circumscribed the areas using the software ( Figure 1 ), and the software automatically calculated the areas. The measurements were independently performed by each reviewer, and the average of the results was used for analyses. The areas with abnormal findings in pixels were converted to disc areas (DA) by dividing the areas by the pixel area of the optic disc.
HCQ retinopathy was classified according to severity and pattern of retinopathy, as performed in previous studies. , The eyes with HCQ retinopathy were divided into early (patchy photoreceptor defects without RPE involvement [ Figure 2 ]), moderate (photoreceptor damage constituting a partial [>180°] or full ring [ Supplemental Figure 3 , left panel]), and severe (combined RPE damage [ Figure 1 ]) retinopathy. Depending on the distribution of retinal involvement, those with photoreceptor/RPE disruption in 2°–8° from the fovea and those with damage in the retina (>8°) from the fovea were divided into parafoveal and pericentral pattern groups, respectively, and those with both patterns of retinal involvement were considered to have a mixed pattern.
Descriptive statistics were obtained for demographic data and the details of HCQ dosage. For comparison between conventional FAF and UWF-FAF, the agreement between the 2 reviewers and sensitivity in retinopathy detection between the 2 FAF modalities were compared. Kappa (κ) statistics were used for evaluating the agreement between standard and UWF-FAF, and sensitivity data were compared using the chi-square test. Correlation analyses were performed to evaluate the association of the area of abnormal findings with the results of HVF tests, mean deviation (MD), and pattern standard deviation (PSD) of the 30-2 protocols, and the number of unseen spots (indicating the number of scotomas) on the FF-120 results.
Statistical analyses were performed using SPSS software version 19 (IBM Corp., Armonk, New York). P values <.05 were considered statistically significant. Continuous variables were expressed as mean ± SD.
Demographic data and HCQ dosage information of the patients are presented in Table 1 . Of the 29 patients, 28 were women 55.5 ± 11.2 years old (range, 27–77 years old). The mean duration of HCQ use was 15.5 ± 5.6 years (range, 6–29 years), and the mean daily dose was 259.4 ± 84.2 mg (range, 131.3–400 mg). The cumulative dose was 1,397 ± 442 g, on average, and the mean daily dose-to-real body weight ratio was 5.1 ± 1.9 mg/kg. The number of patients with the ratio >5 mg/kg was 12, which represented 41.4% of the patients. Mean daily dose-to-ideal body weight (IBW) was 5.2 ± 1.8 mg/kg, and the number of patients at this ratio >6.5 was 8 (27.6%). In all patients, except for 1 who continued to use the drug for 6 months after the diagnosis of HCQ retinopathy, HCQ was discontinued after diagnosis.
|Males:females (%)||1:28 (3.4:96.6)|
|Mean ± age, y||55.5 ± 11.2|
|Diagnosis SLE:RA (%)||19:10 (65.5:34.5)|
|Mean ± refractive errors, D||−1.84 ± 2.63|
|Daily dose, mg||259.4 ± 84.2|
|Body-mass index, kg/m 2||21.5 ± 3.4|
|Ideal body weight, kg a||51.2 ± 4.6|
|Daily dose/body weight, mg/kg||5.1 ± 1.9|
|Daily dose/ideal body weight, mg/kg||5.2 ± 1.8|
|Duration of HCQ use, y||15.5 ± 5.6|
|Cumulative dose/body weight, g/kg||26.7 ± 9.3|
|Severity of retinopathy, early:moderate:severe (%)||19:13:26 (32.8:22.4:44.8)|
|Pattern of retinopathy, pericentral:parafoveal:mixed (%)||41:4:13 (70.7:6.9:22.4)|
|Shortest distance from the fovea to the photoreceptor defects on optical coherence tomography, μm (range)||1989 ± 1222 (0–4350)|
Table 1 presents the clinical details of HCQ retinopathy in the included eyes. More specifically, 19 (32.8%), 13 (22.4%), and 26 (44.8%) eyes had early, moderate, and severe retinopathy, respectively, which was determined based on swept source OCT and FAF modalities. Parafoveal, pericentral, and mixed pattern of retinopathy were noted in 4 (6.9%), 41 (70.7%), and 13 (22.4%) eyes, respectively. In most patients, both eyes showed retinopathy that was symmetrical in terms of severity and pattern of retinopathy. However, 5 patients showed asymmetry in pattern or severity: 1 patient had mixed retinopathy in one eye and pericentral in the other eye, and 4 patients had moderate retinopathy in one eye and early or severe retinopathy in the other.
Sensitivity and Subjectivity of UWF-FAF
Comparison with the Standard FAF: Forty-seven and 52 eyes showed hyper- or hypoautofluorescence in the pericentral or parafoveal areas on standard FAF and UWF-FAF, respectively. The sensitivities of standard FAF and UWF-FAF for the detection of abnormalities were 81.0% and 89.7%, respectively. More specifically, the numbers of eyes with hyperautofluorescence and those with hypoautofluorescence and those with both hyper- and hypoautofluorescence on standard FAF were 45 (77.6%), 26 (44.8%), and 24 (41.4%), respectively, whereas those on UWF-FAF were 50 (86.2%), 26 (44.8%), and 24 (41.4%), respectively. The difference in sensitivities between standard FAF and UWF-FAF was not statistically significant (P = 0.189 based on a chi-square test). Specificities of standard FAF and UWF-FAF were 91.7% and 89.1%, respectively, among the 386 eyes without HCQ retinopathy screened for HCQ toxicity by using both standard FAF and UWF-FAF during the same period. Although repeatability for each standard FAF and UWF-FAF could not be assessed, as only one FAF image was obtained per eye at the baseline for each FAF modality, all 26 patients with follow-up (interval <1 year) FAF images showed corresponding hyper- or hypoautofluorescence in both baseline and follow-up images for both FAF modalities. Figure 2 provides photographic examples of standard FAF and UWF-FAF in a representative case of HCQ retinopathy. Both FAF images showed the area with retinal toxicity as hyperautofluorescent patches. However, the more peripherally located, patchy hyperautofluorescence in the inferior retina, which was noticeable in UWF-FAF ( Supplemental Figure 1 , arrowhead), was not evident in standard FAF. Supplemental Figure 1 also depicts a case showing disagreement in retinopathy detection using standard FAF and UWF-FAF, in which the abnormal finding was noticeable only in UWF-FAF.
Among the eyes included in this study, no disagreement was observed between the 2 reviewers in terms of the findings obtained using UWF-FAF, which had a κ value of 1.0. In contrast, in 3 eyes (5.6%), a disagreement between the reviewers, as shown in Supplemental Figure 1B , was observed in terms of abnormal FAF findings obtained using the standard FAF, which had a κ value of 0.824. In those eyes with early retinopathy showing disagreement between the reviewers, the identification of retinal damage was facilitated by UWF-FAF, as the demarcation between hyperautofluorescence, and other retinal areas was more evident in UWF-FAF ( Supplemental Figure 1 ). Accordingly, in the eyes with early pericentral retinopathy, agreement for judging abnormal findings between the two reviewers was slightly better when UWF-FAF was used. However, in all eyes in which both reviewers identified FAF abnormality, the location of abnormal findings was the same between the 2 reviewers.
Abnormal Autofluorescence Findings on the Peripheral Retina
Outside the covered area of the standard FAF, abnormal findings were observed in 41 (70.7%) eyes on UWF-FAF. More specifically, hyper- and hypoautofluorescence were observed in the peripheral retina in 38 (65.5%) and 23 (39.7%) eyes, respectively. Peripheral involvement was observed in 9, 9, and 23 eyes with early, moderate, and severe stages of disease, respectively. Remarkably, such involvement was noted in a majority of the eyes with severe retinopathy (23 of 26, 88.5%). All abnormal findings in the 4 eyes with a parafoveal pattern were within the coverage of the standard FAF. However, owing to the much larger field of view of the retina using UWF-FAF compared to standard FAF, all abnormalities were covered by the former in 28 and 13 eyes with pericentral and mixed retinopathy, respectively.
The extent of retinopathy, represented by the areas of hyper- or hypoautofluorescence in each eye, is summarized in Table 2 . In these patients, the mean areas of hyper- and hypoautofluorescence were 13.0 ± 12.7 and 18.6 ± 32.6 DA, respectively. Good interobserver agreement was obtained between the 2 measurements for hyperautofluorescence (intraclass correlation coefficient [ICC] = 0.803; 95% confidence interval [CI], 0.688–0.878) and hypoautofluorescence (ICC = 0.888; 95% CI, 0.818–0.932). The central areas (within a circular disc centered on the fovea with a radius equivalent to the distance from the optic disc to the foveal center) showing that FAF abnormality had a significant correlation with the peripheral counterparts (r = 0.769; P < .001), indicating that more extensive central toxicity may be accompanied by more widespread peripheral damage. However, the correlations of the average of the abnormal areas in both eyes with cumulative HCQ dose (correlation coefficient [r] = 0.333), duration of HCQ use (r = 0.197), cumulative dose/body weight ratio (r = 0.053), or daily dose/body weight ratio (r = −0.105; all P > 0.05) were not significant.