Macular Retinal Ganglion Cell–Inner Plexiform Layer Thickness in Patients on Hydroxychloroquine Therapy

Abstract

Purpose.: We evaluated macular ganglion cell-inner plexiform layer (GC-IPL) thickness using spectral-domain optical coherence tomography (SD-OCT) in patients with chronic exposure to hydroxychloroquine (HCQ).

Methods.: This study included 130 subjects, who were divided into three groups: Group 1A, 55 patients with HCQ use ≥5 years; Group 1B, 46 patients with HCQ use <5 years; and Group 2, 29 normal controls. In all patients with exposure to HCQ, fundus examination, automated threshold perimetry, fundus autofluorescence photography, SD-OCT, and GC-IPL thickness measurement using the Cirrus HD-OCT ganglion cell analysis algorithm were performed. Average and minimum macular GC-IPL thickness were compared between subjects groups, and correlations between GC-IPL thickness and duration or total dose of HCQ use were analyzed.

Results.: Among the 101 patients of Group 1, six patients who showed clinically evident HCQ retinopathy also showed markedly thin macular GC-IPL. In addition, weak but significant negative correlations were observed between the average and minimum GC-IPL thickness of Group 1 patients and cumulative dose of HCQ, even when analyzing without the six patients with HCQ retinopathy. However, when analyzing after exclusion of patients with high cumulative doses (>1000 g), significant correlations were not observed.

Conclusions.: This study revealed that macular GC-IPL thickness did not show definite correlations with HCQ use. However, some patients, especially with HCQ retinopathy or high cumulative doses, showed thin GC-IPL.

Introduction

Hydroxychloroquine (HCQ), a chloroquine analogue and originally an antimalarial drug, is used widely in the treatment of various rheumatologic diseases, such as systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA). Retinal toxicity from HCQ is a well-known side effect and is characterized classically by bilateral bull's eye maculopathy sparing the foveal center.¹ Patients with bull's eye maculopathy may suffer from irreversible, sometimes progressive, loss of visual acuity, central visual field, and color vision.^1,2 After HCQ has largely replaced chloroquine due to its better tolerability, retinal toxicity from HCQ has a lower incidence, but still remains a major concern because of its potentially irreversible visual loss and the considerable number of patients on HCQ therapy. Indeed, patients with HCQ retinopathy usually do not show significant recovery, and there often is continuing functional loss even after discontinuation of the drug. Therefore, early detection is important to minimize loss of visual function. 

The overall incidence of retinal toxicity from HCQ is low.^2,3 A recent study of 3995 patients who previously used HCQ revealed that the incidence of definite or probable retinal toxicity was 0.65%. However, the risk of toxicity was markedly increased after 5 to 7 years of use (or 1000 g of cumulative dose) and exceeds 1%.⁴ Moreover, recent advances in objective imaging modalities, including spectral-domain optical coherence tomography (SD-OCT) and fundus autofluorescence (FAF) can detect subtle changes of macula, which may reveal more patients with HCQ retinopathy.⁵ 

Previous histopathological studies of human retinas with chloroquine retinopathy showed destruction of photoreceptors and neuroretina.^6–8 Rosenthal et al.⁹ reported that chloroquine caused an initial dramatic effect on the retinal ganglion cells (RGCs), with subsequent degeneration of RGCs and photoreceptors in rhesus monkeys. Hallberg et al.¹⁰ showed that chloroquine leads to morphologic and biochemical signs of phospholipidosis in the neuroretina, but does not affect phospholipid metabolism of the RPE of mice. Whether HCQ also impairs the RGCs in patients is not clear. A recent study by Pasadhika et al.¹¹ reported that selective thinning of perifoveal RGC and inner plexiform layers was found in patients with chronic exposure to HCQ in the absence of functional or structural clinical changes involving the photoreceptor or RPE layers. However, this study had a small sample size (n = 8), and the correlation between dose of HCQ and perifoveal thinning was not investigated. 

To evaluate retinal structural changes related to RGCs, measurement of the macular ganglion cell–inner plexiform layer (GC-IPL) thickness can be used. Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA, USA) can automatically segment and measure macular GC-IPL thickness by ganglion cell analysis (GCA) algorithm. The macular GC-IPL thickness by GCA algorithm has been reported to have diagnostic values in glaucoma patients.^12–15 

In the present study, we investigated the association between HCQ use and macular GC-IPL thickness of 101 patients on HCQ therapy by Cirrus HD-OCT GCA algorithm to find whether there were structural changes related to RGCs by chronic exposure to HCQ. 

Patients and Methods

A retrospective review of medical records of 101 patients who had used HCQ for the treatment of rheumatologic disease, and were referred to ophthalmology department for screening of HCQ retinopathy between December 2012 and August 2013, was conducted. This study was approved by the Institutional Review Board of the Samsung Medical Center, and the work was done in accordance with the Declaration of Helsinki. Exclusion criteria were known optic nerve diseases; glaucoma; retinal diseases, including conditions that may affect outer retinal structures on SD-OCT; inflammatory eye diseases; and media opacity that interfere clear fundus or OCT examination. 

According to the revised American Academy of Ophthalmology (AAO) guidelines regarding HCQ screening, all patients underwent screening examinations, including dilated fundus examination, visual field testing in the form of static, automated threshold perimetry (10-2 Humphrey Field Analyzer, Model 750I; Humphrey Instruments, Inc., San Leandro, CA, USA), FAF, and SD-OCT (Spectralis HRA+OCT; Heidelberg Engineering, Heidelberg, Germany).⁵ The HCQ retinopathy was categorized as early, moderate, or severe according to the prior criteria.^16,17 In addition, the Cirrus HD-OCT GCA algorithm was used to detect the macular GC-IPL, and to measure the thickness of the overall average and minimum GC-IPL within a 6 × 62 mm elliptical annulus area centered on the fovea. The GCA algorithm was described previously in detail.^15,18 Eyes with segmentation error, which was defined as disruption of the detected border and/or border wandering, were excluded from the study. The GCA also was performed for normal control subjects who had no ocular diseases, including retinal diseases, glaucoma, and optic neuropathies. 

Because duration of HCQ use exceeding five years is a known risk factor for HCQ retinopathy, the 130 subjects were divided into three groups: Group 1A, 55 patients with HCQ use ≥ 5 years; Group 1B, 46 patients with HCQ use < 5 years; and Group 2, 29 normal controls.⁴ In addition, other risk factors for HCQ retinopathy, including daily dose and kidney or liver dysfunction, also were documented.^4,5 

Only the right eye of each participant was included for analysis. Average and minimum GC-IPL thickness of Group 1A and 1B patients were compared to those of Group 2 subjects. Average and minimum GC-IPL thickness of Group 1A patients were compared to those of Group 1B patients. In addition, Pearson correlation analysis was conducted between GC-IPL thickness and duration or total dose of HCQ use. The analysis was repeated after excluding patients with HCQ retinopathy and patients with high cumulative doses (>1000 g). All statistical analyses of the data were done using PASW Statistics 18 (SPSS, Inc., Chicago, IL, USA) and the significance level was set at a P < 0.05. 

Results

Table 1 shows the demographic and HCQ-related clinical characteristics of all 130 patients. The mean age of 101 patients in Group 1 was 41.3 years, and 89 patients (88.1%) were female. Mean duration of HCQ use in Group 1 was 67.1 months and mean total exposure of HCQ was 557.9 g. The mean daily dose was 292.5 g and 24 (23.8%) patients had taken mean 400 mg or more of daily dose of HCQ. Mean daily dose per kilogram of ideal body weight was 5.7 g and 36 (35.6%) patients had used 6.5 mg or more per kilogram of ideal body weight of daily dose of HCQ (Table 1). Group 2 subjects comprised 29 normal controls, whose mean age was not statistically different from Group 1 patients (P = 0.337, independent t-test). 

Table 1

 

Demographic and Clinical Characteristics of the Patients

Parameters	Group 1A	Group 1B	Group 1	Group 2
	HCQ Use ≥ 5 y	HCQ Use < 5 y		Normal Controls
Number of subjects	55	46	101	29
Age, y, mean ± SD	44.1 ± 12.0	37.9 ± 13.6	41.3 ± 13.1	43.8 ± 10.8
Female, n (%)	46 (83.6)	43 (93.5)	89 (88.1)	20 (69.0)
Body weight, kg, mean ± SD	56.5 ± 7.5	57.2 ± 9.5	56.8 ± 8.6	NA
BMI, kg/m2, mean ± SD	22.3 ± 2.5	22.6 ± 3.2	22.4 ± 2.9	NA
Duration of use, mo, mean ± SD (range)	97.1 ± 25.9	31.2 ± 16.9	67.1 ± 39.7	NA
	(62–174)	(2–58)	(2–174)
Cumulative dose, g, mean ± SD	778.6 ± 318.7	294.1 ± 173.4	557.9 ± 356.6	NA
Daily dose, mg, mean ± SD	268.9 ± 78.5	320.8 ± 85.9	292.5 ± 85.6	NA
Daily dose per ideal body weight, mg/kg, mean ± SD	5.2 ± 1.8	6.3 ± 1.8	5.7 ± 1.9	NA
Number of patients whose daily dose ≥400 mg, n (%)	6 (10.9)	18 (23.1)	24 (23.8)	NA
Number of patients whose daily dose ≥6.5 mg/kg ideal body weight, n (%)	14 (25.5)	23 (50.0)	37 (36.6)	NA
Diagnosis of patients,* n (%)
Systemic lupus erythematosus	43 (78.2)	35 (76.1)	78 (77.2)	NA
Sjögren syndrome	4 (7.3)	8 (17.4)	12 (11.9)	NA
Rheumatoid arthritis	4 (7.3)	6 (13.0)	10 (9.9)	NA
Others	7 (12.7)†	1 (2.2)‡	8 (7.9)	NA
Systemic disease, n (%)
Kidney dysfunction	21 (38.2)	8 (17.4)	29 (28.7)	NA
Liver dysfunction	9 (16.4)	0 (0.0)	9 (8.9)	NA

 BMI, body mass index; NA, not applicable.

*  Seven patients were diagnosed with more than one disease.

†  Five patients were diagnosed with antiphospholipid antibody syndrome, three patients with mixed connective tissue disease. One of them had both diseases.

‡  One patient was diagnosed with dermatomyositis and systemic lupus erythematosus.

After screening examinations for HCQ retinopathy, six (5.94%) patients were diagnosed with definite or probable HCQ retinopathy. Of the six patients, three showed severe HCQ retinopathy, one patient showed moderate retinopathy, and two patients showed early retinopathy. Five of the six patients were asymptomatic and the best corrected visual acuity (BCVA) was 20/25 or more in all eyes. One patient with severe retinopathy showed decreased central vision with BCVA of 20/200 in the right eye and 20/50 in the left eye. The six patients had used HCQ for a mean duration of 116.2 months and the mean cumulative dose was 958.0 g. Macular GC-IPL thickness of all these six patients were markedly decreased (Table 2). The representative case is shown in Figure 1. 

Figure 1

 

Photographs show HCQ retinopathy in the right (A) and left (B) eyes of a 31-year-old female who was diagnosed with SLE and had used HCQ for 119 months. A cumulative dose of HCQ was 786 g and the daily dose per ideal body weight was 4.4 mg/kg. Her BCVA was 20/20 in both eyes. Color fundus photographs and FAF show a bull's eye pattern maculopathy and SD-OCT demonstrates loss of parafoveal photoreceptor layer in both eyes. Humphrey 10-2 visual field testing shows paracentral scotoma in both eyes. (C) Ganglion cell analysis of Cirrus HD-OCT reveals markedly reduced average and minimum macular GC-IPL thickness in both eyes.

Table 2

 

Macular GC-IPL Thickness of Patients

Parameters	Group 1, n = 101	Group 1A HCQ Use ≥ 5 y, n = 55				Group 1B HCQ Use < 5 y, n = 46	Group 2 Normal Controls, n = 29
		All Patients	Patients With HCQ Use < 1000 g, n = 47	Patients Without HCQ Retinopathy, n = 49	Patients With HCQ Retinopathy, n = 6
GC-IPL thickness
Average, μm	80.1 ± 11.2	77.9 ± 14.2	79.2 ± 12.4	81.7 ± 6.0	46.8 ± 23.3	83.1 ± 4.7	83.7 ± 4.3
P value I*		0.013	0.049	0.206	<0.001		0.607
P value II†	0.113	0.007	0.027	0.124	0.012	0.607
Minimum, μm	76.5 ± 14.2	73.4 ± 17.7	75.0 ± 15.9	78.3 ± 7.8	32.8 ± 24.5	80.3 ± 6.9	80.6 ± 5.8
P value I*		0.010	0.041	0.207	0.005		0.816
P value II†	0.113	0.007	0.031	0.176	0.005	0.816

*  The average GC-IPL thickness was compared to that of Group 1B (independent t-test).

†  The average GC-IPL thickness was compared to that of Group 2 (independent t-test).

Average and minimum macular GC-IPL thicknesses of Group 1 patients were 80.1 ± 11.2 and 76.5 ± 14.2 μm, respectively, which were not significantly different from those of Group 2 (Table 2). Average and minimum macular GC-IPL thickness of Group 1A patients were significantly lower than those of Groups 1B and 2, respectively (Table 2). However, the average and minimum macular GC-IPL thicknesses of Group 1B patients were not significantly different from those of Group 2. When excluding patients with HCQ use more than 1000 g in Group 1A, the average and minimum macular GC-IPL thickness also were significantly lower than those of Groups 1B and 2, respectively (Table 2). 

Correlation analysis showed that average and minimum macular GC-IPL thicknesses were significantly correlated with duration and cumulative dose of HCQ in Group 1 (Table 3; Fig. 2). To exclude the outlier effect of six patients with clinical HCQ retinopathy, who all showed markedly thin macular GC-IPL, we analyzed the correlation between the average and minimum GC-IPL thicknesses of Group 1 patients without HCQ retinopathy (n = 95) and duration and cumulative dose of HCQ. The results showed significant negative correlations between the average and minimum GC-IPL thicknesses of Group 1 patients without HCQ retinopathy (n = 95) and cumulative dose of HCQ (Table 3; Fig. 2). However, the duration of exposure to HCQ was not significantly correlated with GC-IPL thickness of these 95 patients (Table 3). After excluding the patients with HCQ retinopathy and HCQ use more than 1000 g in Group 1 (n = 88), there was no significant correlation between the average and minimum GC-IPL thicknesses and cumulative dose of HCQ (Fig. 2). The representative case of thin macular GC-IPL thickness without clinically evident HCQ retinopathy in Group 1 is shown in Figure 3. 

Figure 2

 

Graphs show linear regression between (A) average and (B) minimum macular GC-IPL thickness and cumulative dose of HCQ use in whole subjects on HCQ therapy (n = 101), and between (C) average and (D) minimum macular GC-IPL thickness and cumulative dose of HCQ use in patients without HCQ retinopathy (n = 95). Red dots indicate six patients with HCQ retinopathy. Linear regression between (E) average and (F) minimum macular GC-IPL thickness and cumulative dose of HCQ use in patients without HCQ retinopathy and with HCQ use <1000 g (n = 88) demonstrates no significant correlation.

Figure 3

 

Photographs show screening examinations of right (A) and left (B) eyes of a 27-year-old female who was diagnosed with SLE. She had chronic exposure to total 903 g of HCQ for 123 months. Her daily dose per ideal body weight was 5.0 mg/kg. The BCVA was 20/20 in both eyes. Color fundus photographs, FAF, and Humphrey 10-2 visual field testing show no abnormality. (C) Ganglion cell analysis of Cirrus HD-OCT reveals thinning of macular GC-IPL in both eyes.

Table 3

 

Correlation Analysis Between Macular GC-IPL Thickness and Duration or Total Dose of HCQ Use in Patients on HCQ Therapy

	Average GC-IPL Thickness, μm		Minimum GC-IPL Thickness, μm
	P Value*	Pearson Coefficient	P Value*	Pearson Coefficient
Group 1, n = 101
Duration of HCQ use	0.001	−0.337	<0.001	−0.347
Cumulative HCQ dose	<0.001	−0.430	<0.001	−0.419
Group 1 without HCQ retinopathy, n = 95
Duration of HCQ use	0.125	−0.159	0.151	−0.148
Cumulative HCQ dose	0.033	−0.219	0.025	−0.229
Group 1 without HCQ retinopathy or HCQ use > 1000 g, n = 88
Duration of HCQ use	0.636	−0.051	0.714	−0.040
Cumulative HCQ dose	0.475	−0.077	0.457	−0.080

*  Pearson's correlation test.

Discussion

In the present study, we investigated whether there were changes in macular RGC layers in patients who were on chronic HCQ therapy. As a surrogate biomarker for changes in RGC, measurement of macular GC-IPL thickness using the Cirrus HD-OCT GCA algorithm was used. This study revealed that macular GC-IPL thickness showed a weak but significant negative correlation with cumulative dose of HCQ. However, when analyzing after exclusion of the patients with high cumulative doses (>1000 g), significant correlations were not observed. These findings suggest that although some patients, especially with HCQ retinopathy or high cumulative doses, showed thin GC-IPL, thinning of GC-IPL was not a general phenomenon in patients with exposure to HCQ. 

This study investigated the possible role of measuring GC-IPL thickness as an objective ancillary testing to detect HCQ retinopathy earlier. Although thin GC-IPL was observed especially in patients with HCQ retinopathy or high cumulative doses, some patients with low cumulative doses also showed thin GC-IPL. This suggested that thinning of GC-IPL is not likely to be a sensitive diagnostic sign. However, thinning of GC-IPL was evident in all six patients with HCQ retinopathy. Therefore, longitudinal study on long-term users is warranted to determine whether thinning of GC-IPL precedes HCQ retinopathy. 

The HCQ retinopathy is uncommon; however, its worldwide use in a large number of patients with rheumatologic diseases arouses interest in finding a potential screening procedure. The revised AAO recommendations suggest that annual screening should be performed on all patients within first year of use and annually after 5 years of HCQ use.⁵ Wolfe and Marmor⁴ reported that the prevalence was dependent on the duration of HCQ use and the risk of toxicity was higher than previously believed. Before visible fundus changes, subtle functional loss, such as paracentral scotoma demonstrated by automated threshold perimetry and localized paracentral depression in multifocal electroretinogram may manifest an early retinal toxicity. Objective early anatomic changes, such as parafoveal thinning of retinal layers or loss of photoreceptor inner/outer segment junction line in SD-OCT and increased FAF, could be found before visual field loss actually occurs.⁵ Because clearance of HCQ from the body can take many months after therapy is stopped, there may be a continuing functional loss even after cessation of HCQ: Earlier detection of HCQ toxicity is extremely important.⁵ 

The mechanism of chloroquine and HCQ retinopathy is not well-understood. The binding capacity of pigmented tissues, such as choroid, RPE, ciliary body, and iris for chloroquine, is high.^19–21 The HCQ is known to bind to melanin in RPE and this binding may contribute to or prolong toxic effects.⁴ However, a study of Rhesus monkeys by Rosenthal et al.⁹ revealed that chloroquine also accumulates in the neurosensory retina (including RGCs) and affects the photoreceptors before the RPE shows any sign of pathology. The first pathological changes in the retina occurred within 1 week of the onset of chloroquine administration in RGCs, with the photoreceptors affected shortly thereafter. Also, chloroquine caused degeneration of RGCs and photoreceptors over several years.⁹ Other animal studies have found that the initial reaction of the retina to chloroquine is the formation of membranous cytoplasmic bodies in RGCs.^22–24 Therefore, HCQ also may cause degeneration of RGCs in patients, which may result in thinning of RGC layer. With the current SD-OCT system, segmentation of the RGC layer only is not possible. Thus, in this study, we measured GC-IPL thickness with automated detection and segmentation algorithm. However, outer retinal damage or thinning can interfere the automatic segmentation of retinal layers by the Cirrus HD-OCT GCA algorithm. In eyes with parafoveal outer retinal damage or RPE damage, it might be a confounding factor for detecting the early changes of GC-IPL.²⁵ Therefore, we excluded patients with retinal diseases that may affect outer retinal structures on SD-OCT and also examined whether there were segmentation failures in all cases. 

Several studies have reported localized thinning of the parafoveal inner retina on SD-OCT in patients with chronic exposure to HCQ. Pasadhika and Fishman²⁶ found that patients with abnormal fundus showed thinning of inner, outer, and full-thickness retina, and patients with chronic exposure to HCQ without fundus changes showed significant thinning of the inner retina only. Another study by Pasadhika et al.¹¹ on patients who were visually normal and without fundus abnormality or any defect in Humphrey 10-2 visual field testing reported that selective thinning of the GC-IPL was observed only in the parafoveal area. However, this study enrolled only eight patients and eight controls. In addition, the dose–response relationship between the two was not investigated. On the contrary, our study included a large number of patients with varying exposure to HCQ and correlation between the dose and GC-IPL thickness was investigated. On the contrary to previous studies, there were no definite correlations between macular GC-IPL thickness and HCQ use. 

In this study, patients with HCQ retinopathy showed markedly thin GC-IPL, as well as parafoveal photoreceptor layer thinning on SD-OCT. In human diseases, like retinitis pigmentosa, and animal models of photoreceptor degeneration, photoreceptor loss may impair inner retinal cells, including RGCs in the long term.^27–30 Therefore, the thinning of macular GC-IPL does not necessarily mean RGC degeneration is caused by HCQ. Photoreceptor loss followed by consecutive transneuronal degeneration of RGCs also may make thinning of GC-IPL. However, in retinitis pigmentosa, where primary degeneration occurs in photoreceptors, inner retinal structure is preserved for long time.^28,31 Vámos et al.²⁸ reported that, in retinitis pigmentosa, the pericentral GC-IPL on OCT are preserved in eyes in which the underlying photoreceptors already have been reduced, but GC-IPL also shows thinning with disease progression thereafter. Thus, although concomitant secondary degeneration by loss of underlying photoreceptors also is possible, severe thinning of GC-IPL in patients with HCQ retinopathy could be caused by chronic exposure to HCQ. 

In conclusion, this study revealed that macular GC-IPL thickness did not show definite correlations with cumulative dose of HCQ. However, some patients, especially with HCQ retinopathy or high cumulative doses, showed thin GC-IPL. 

Acknowledgments

Supported by Grant HI13C1826 from the Korean Health Technology R&D Project, Ministry of Health & Welfare, Republic of Korea, and by Samsung Medical Center Grant #SMR1120521. 

Disclosure: M.G. Lee, None; S.J. Kim, None; D.-I. Ham, None; S.W. Kang, None; C. Kee, None; J. Lee, None; H.-S. Cha, None; E.-M. Koh, None