Cone Density Distribution and Related Factors in Patients Receiving Hydroxychloroquine Treatment

Abstract

Purpose: Hydroxychloroquine is an effective treatment for rheumatic diseases; however, retinal damage is a possible side effect. We aimed to identify the retinal area and related risk factors associated with cone density reduction caused by hydroxychloroquine.

Methods: We recorded the retinal images of patients with rheumatic diseases taking hydroxychloroquine (n = 44) and compared them with images of healthy controls (n = 107). Cone density was obtained in vertical and horizontal axes. Regions of decreased cone density and associations between age, rheumatic disease type, dosage for ideal body weight, and cone density were evaluated.

Results: Cone densities were significantly lower in hydroxychloroquine-treated patients than in sex- and age-matched controls in the vertical axis (P < 0.001), with no significant difference in the horizontal axis (P = 0.120); in healthy elderly than in healthy young people in the horizontal axis (P < 0.001), with no significant difference in the vertical axis (P = 0.100); in hydroxychloroquine-treated elderly than in hydroxychloroquine-treated young patients in both axes (both P < 0.05); among patients with different rheumatic disease types, with no significant difference in the vertical axis (P = 0.294). The daily dose was negatively correlated with cone density in the vertical axis and inferior quadrant.

Conclusions: Hydroxychloroquine reduces retinal cone cell density in the vertical axis. Cone density loss in the horizontal axis increases with age; further, hydroxychloroquine dosage is negatively correlated with cone density in the vertical axis and inferior quadrant. Early screening of hydroxychloroquine-related retinal injury should consider changes in cone density in the vertical axis.

Hydroxychloroquine is widely used in the treatment of rheumatic diseases, but can cause retinal damage as a secondary effect.¹ The prevalence of retinal damage in patients treated with hydroxychloroquine for more than 5 years is 7.5%,² increasing to 20% after 20 years of treatment, and occasionally leading to blindness in severe cases.³ The mechanism underlying this retinal damage remains unclear. However, there is evidence to suggest that hydroxychloroquine may bind with melanin in the retinal pigment epithelial cells, destroying the foveal cones and the outer nuclear layer, and subsequently leading to bullseye maculopathy.⁴ 

Considering the irreversibility of hydroxychloroquine-induced retinal damage, early detection and subsequent discontinuation of hydroxychloroquine treatment before the occurrence of structural retinal pigment epithelial damage are critical. The American Ophthalmological Association guidelines recommend screening for hydroxychloroquine-induced retinal damage,⁵ for which automatic threshold field and SD-OCT is the primary tool.^6,7 Other available objective screening tools include multifocal electroretinography, but none of these tools has been established as the gold standard.⁸ 

Adaptive optics (AO) imaging technology, which eliminates the disadvantage of the histological evaluation of the retinal cells and allows noninvasive visual evaluation of retinal cone cells, has been widely used in the study of diabetic retinopathy,⁹ AMD,¹⁰ glaucoma¹¹ and CNGA3-associated achromatopsia.¹² The cumulative dose of hydroxychloroquine is negatively correlated with cone density in patients without clinical evidence of macular disease, as previously reported.¹³Additionally, the cone density of patients treated with high dosages of hydroxychloroquine is significantly lower than that of patients treated with low dosages.¹⁴ However, a previous study of patients treated with hydroxychloroquine found that the cone density did not significantly decrease with an increase in dose.¹⁵ These studies used AO to evaluate cone density in patients with hydroxychloroquine-induced damage. However, the results to date remain contradictory, perhaps because only the impact of drug dosage on the overall cone cell density was analyzed in these previous studies. Moreover, the distributions of cone cell density in varying locations within the retina,¹⁶ as well as the impact of age, sex, and disease type on cone cells,¹⁷ were not considered in past research. Age, disease type, and hydroxychloroquine dosage can all affect cone cell density at different locations. 

This study aimed to compare the cone densities between patients treated with hydroxychloroquine and age- and sex-matched healthy controls, and to identify cone density reduction in the vertical and horizontal axes. Moreover, we explored the relationship between age, rheumatic disease type, hydroxychloroquine dose, and cone density. 

Methods

Study Participants

Patients with systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), Sjögren's syndrome, systemic sclerosis, and undifferentiated connective tissue disease who received hydroxychloroquine treatment at Hainan Hospital of Chinese PLA General Hospital between May 2021 and June 2021 were recruited for this study. Healthy controls who did not receive any medical treatment were also recruited. Data on medical history, including patient characteristics, demographics, disease duration, weight, height, duration of hydroxychloroquine treatment, cumulative dose, and average daily dose for ideal weight, were obtained. We only enrolled patients who received hydroxychloroquine treatment and met either the American College of Rheumatology 2010 RA classification criteria,¹⁸ the 1997 SLE classification criteria,¹⁹ the 2012 Sjögren's syndrome classification criteria,²⁰ and/or the 2013 American College of Rheumatology/European League Against Rheumatism classification criteria for systemic sclerosis.²¹ Patients with a history of eye surgery, serious eye disease, or a systemic or eye disease that may cause retinopathy, such as glaucoma, cataract, AMD, diabetes, or other inherited rod–cone dystrophies were excluded. 

The study protocol complied with the ethical standards of the Declaration of Helsinki, and was approved by the Ethics Committee of the Hainan Branch of People's Liberation Army General Hospital (approval number: 301HNFY36). Informed consent was obtained from all study participants. 

AO Imaging

AO images were obtained from a commercially available rtx1 AO device (Imagine Eyes, Orsay, France). The rtx1 is based on an incoherent flood lighting design using the principle of near-infrared (850 nm) reflection imaging. The imaging depth of the device can be adjusted to 0 −80 µm, and the clearest image was obtained by spherical equivalent correction (−12 D to +10 D); most individuals achieve the clearest image at a focal depth of approximately +50 µm. After pupil dilation with tropicamide, the patient was asked to immobilize their eye on a yellow target controlled by the operator to obtain retinal images at a 3° eccentricity in the center and along the four quadrants (nasal, temporal, superior, and inferior) (Fig. 1A). The final high-resolution AO image was obtained at a 4° × 4° field of view (1.2 × 1.2 mm on the retina). 

Figure 1.

 

Adaptive optics cone images of the right eye of a patient receiving hydroxychloroquine treatment (A). AO imaging is used to obtain raw AO images of 4° × 4° size, centered at 3° of eccentricity from the fovea, revealing the superior and inferior quadrants along the vertical axis and the nasal and temporal quadrants along the horizontal axis (B). The measurement of cone metrics is performed within a region of interest (white frame) corresponding to a square of 0.21° × 0.21°, and the resulting images of the cone mosaic are shown in (C).

Imaging Processing

The local density was analyzed in a 0.21° × 0.21° square region of interest, which was evenly distributed across the nine measurable regions of quadrants (Fig. 1B). The measurements used for analysis were obtained by taking the average of the values obtained for the nine squares within each image. The cone could be automatically identified by detecting the central coordinates of small dots with a brightness higher than that in the surrounding background (Fig. 1C). If the region of interest fell on a shaded area, it was moved slightly toward the measurable area. Images with low quality were excluded from the analysis to ensure reliable results. The method used by Debellemanière et al¹³ was used for counting cones, because it is has been shown to have good repeatability. 

Statistical Analysis

SPSS version 22.0 (IBM Corp., Armonk, NY) was used for statistical analysis. A single-sample Kolmogorov–Smirnov test was used to assess the normal distribution of the overall cone density data. Results are expressed as the mean ± standard deviation. Independent sample t tests were used for comparison in the four retinal quadrants between hydroxychloroquine-treated patients versus healthy controls. One-way ANOVA was used for comparisons between the different rheumatic disease types, and the Bonferroni method was used to adjust for multiple comparisons. Linear regression analysis was applied to analyze the relationship between dose for ideal body weight and cone density. Statistical significance was set at a P value of ≤0.05. 

Results

Participants

A total of 44 patients with rheumatic diseases treated with hydroxychloroquine and 107 healthy controls were enrolled. The mean age of the patients with rheumatic diseases and healthy controls did not differ significantly (42.6 ± 13.9 vs. 39.7 ± 17.0 years; P = 0.321); however, the hydroxychloroquine cohort had significantly more women than the control cohort (88.6% vs. 26.2%; P < 0.001). To avoid bias caused by sex and age differences, 29 patients with rheumatic disease treated with hydroxychloroquine were 1:1 matched for sex and age with 29 healthy controls, yielding 29 pairs successfully matched for sex (82.8% vs. 82.8%) and age (44.5 ± 15.0 vs. 44.3 ± 15.0). 

Overall, 14 patients had RA, 18 patients had SLE, 6 patients had Sjögren's syndrome, 4 patients had undifferentiated connective tissue disease, and 2 patients had systemic sclerosis. The hydroxychloroquine dose for ideal body weight was 6.5 ± 1.5 mg/kg/day, the total cumulative hydroxychloroquine dose was 368.3 ± 267.2 g, the duration was 77.4 ± 53.6 months, and the hydroxychloroquine exposure time was 36.5 ± 26.5 months. 

Cone Density Reduction Between Hydroxychloroquine-treated and Control Groups

Single-sample Kolmogorov–Smirnov testing validated that the four-quadrant cone density data followed a normal distribution, with a P value of 0.2. The mean cone density in the four quadrants of the hydroxychloroquine group (17,962 ± 3,080) was significantly lower than that of the healthy control group (20,212 ± 3,263; P = 0.009) (Table 1). The densities of the cone cells in the vertical axis, superior quadrant, and inferior quadrant were significantly lower in the hydroxychloroquine group than in the healthy control group (all P <0.001) (Fig. 2). No significant difference was noted in cone density in the horizontal axis (P = 0.120), temporal quadrant (P = 0.092), or nasal quadrant (P = 0.179) between the hydroxychloroquine and healthy control groups. 

Table 1.

 

Mean Cone Density of the Participants

 

Figure 2.

 

AO cone images of health control and patient receiving hydroxychloroquine. Visualization of cone mosaic in four quadrants from the healthy control group (A, B, C, and D) and from the hydroxychloroquine group (E, F, G, and H).

Cone Density and Age

A total of 95 young people (<45 years) were included in the sample, accounting for 62.9% of all study participants, including 71 in the healthy control group and 24 in the hydroxychloroquine group (Table 2). The average cone density in the four quadrants was significantly lower in elderly participants (≥45 years) than in young people (control, P = 0.008; hydroxychloroquine, P = 0.004). 

Table 2.

 

Cone Density in the Different Age Groups

 

In the control group, the cone density in the horizontal axis was significantly lower in the elderly group than that in the young group (P < 0.001), with no significant decrease in cone density in the vertical axis (P = 0.100). Meanwhile, the cone densities in both the vertical and horizontal axes were significantly lower in elderly people in the hydroxychloroquine treatment group than in young people in the hydroxychloroquine treatment group (both P < 0.05). 

Cone Density Among Different Rheumatic Diseases

No significant difference was noted in the mean cone density of the four quadrants (P = 0.054), or in the cone density in the vertical axis (P = 0.294). However, cone density in the horizontal axis of patients with RA was significantly lower than that of patients with SLE (P = 0.005; Fig. 3A), and the age of patients with RA was significantly higher than that of patients with SLE (P < 0.001) (Fig. 3B). 

Figure 3.

 

The cone density in patients treated with hydroxychloroquine according to rheumatic disease type. (A) Cone density in the four quadrants, along the vertical and horizontal axes. (B) Age of the three groups according to rheumatic disease type. CTD, connective tissue disease.

Dose and Cone Density

According to the calculation of the daily dose relative to ideal body weight, 25 cases (56.8%) had a daily dose relative to ideal body weight of >6.5 mg/kg/day. Linear regression analysis revealed that the daily dose relative to ideal body weight was negatively correlated with the mean four-quadrant cone density (r² = 0.11, P = 0.028; adjusted age and gender: r² = 0.27, P = 0.025), vertical cone density (r² = 0.10, P = 0.034; adjusted age and gender: r² = 0.20, P = 0.038) (Fig. 4A), and inferior quadrant density (r² = 0.10, P = 0.037; adjusted age and gender: r² = 0.27, P = 0.038) (Fig. 4B). However, no correlation was noted with cone densities in the superior quadrant (Fig. 4C), horizontal axis (Fig. 4D), nasal quadrant (Fig. 4E), or temporal quadrant (Fig. 4F). 

Figure 4.

 

Relationship between cone density in different distributions and daily hydroxychloroquine dose for ideal body weight. (A) Vertical. (B) Inferior. (C) Superior. (D) Horizontal. (E) Nasal. (F) Temporal.

Discussion

To the best of our knowledge, this study is the first to use an rtx1 AO fundus camera to observe differences in cone density between patients with rheumatic diseases treated with hydroxychloroquine and age- and sex-matched healthy controls. Notably, we identified significant reductions in cone density in the vertical axis, but not the horizontal axis, in hydroxychloroquine-treated patients compared with healthy controls. The cone density between the horizontal and vertical axes exhibited differences of 14%, 20%, 16%, 11%, and 8% at 2°, 3°, 4°, 5°, and 6°, respectively, indicating a higher density along the horizontal meridian.¹⁷ One explanation suggests that this difference may have evolved for a specific reason.²² Indeed, the horizontal meridian is believed to have experienced stronger selective pressures throughout evolution owing to a variety of factors. For example, detecting objects in the horizontal plane (e.g., predators, prey) and navigating the environment horizontally (e.g., across landscapes) may have contributed to the optimization of the visual system along the horizontal axis, including a higher cone density. Another explanation proposes that, during reading, the horizontal retina becomes more excited than the vertical retina, therefore requiring a more pixelated view.²³ Reading predominantly involves processing information along the horizontal axis, moving from left to right, as written languages follow this direction. The visual demands of reading, such as the recognition and differentiation of the fine details of letters and words, require higher resolution and pixelation in the horizontal direction. However, the reason why hydroxychloroquine more strongly affects the cones in the superior/inferior retina is not entirely clear. The differential effects of hydroxychloroquine on retinal cone cell density in the vertical and horizontal axes can be explained by both evolutionary adaptations and functional variations. Adaptations driven by evolution may have led to the development of protective mechanisms along the horizontal axis to mitigate potential harm from increased excitation. Functional variations between the two axes, such as differences in cellular composition and metabolic processes, may also have contributed to this differential susceptibility. However, it is important to note that these explanations are hypothetical, and further studies are required to fully understand the precise mechanisms underlying this phenomenon. 

In the present study, we also analyzed the influence of age on cone density, and found that the mean cone density was lower in elderly people than in young people, regardless of hydroxychloroquine use, which is consistent with the findings in the literature that advancing age is an important factor leading to the decrease of cone density in healthy individuals.^17,24 Upon further analysis of healthy controls, we found that the cone density decreased significantly in the elderly in the horizontal axis, but not the vertical axis, suggesting that age primarily affects cone density in the horizontal axis. We further found that the cone densities in both the vertical and horizontal axes decreased significantly in elderly persons treated with hydroxychloroquine. Interestingly, the cone density reduction in the horizontal axis may be related to age, whereas that in the vertical axis may be related to hydroxychloroquine use. This result further confirms our previous finding that hydroxychloroquine affects the density of the vertical axis cones. The reason for this selective decrease in cone density along the horizontal axis with aging may be related to differences in mechanical stress. The mechanical properties of the retina are strongly influenced by blood vessel orientation, with vessel distribution being a key factor in determining the mechanical environment of the cells.²⁵ The horizontal meridian experiences greater mechanical stress and strain than the vertical meridian. As the vertical transition strain decreases and the horizontal transition strain increases with age,²⁶ it is unclear whether this change in mechanical stress would affect the density of cone cells; therefore, further research is needed. 

As yet, there is no consensus on the difference between rheumatic disease types and retinopathy in patients treated with hydroxychloroquine. The detection methods are mostly multifocal electroretinography, fundus autofluorescence, and OCT, without AO detection data. Retinopathy caused by hydroxychloroquine is not related to SLE or RA,²⁷ and patients with RA are more prone to hydroxychloroquine-induced retinal toxicity than patients with SLE.²⁸ According to our analysis, cone density was not significantly different in the four quadrants and the vertical axis among the groups with different rheumatic diseases (RA, SLE, and other connective tissue diseases); however, cone density in the horizontal axis was lower in the RA group than in the SLE group, which may be due to the age difference between the groups, as the patients with RA were significantly older than those with SLE. This finding is consistent with our finding that age predominantly affects cone density on the horizontal axis. Unfortunately, because of the small sample size and failure to match for age in the analysis, more convincing evidence could not be obtained. 

To date, no studies have yet analyzed the linear relationship between daily dose and cone density, despite the fact that the daily dose is an important risk factor for hydroxychloroquine-related retinal injury and is the only controllable factor.²⁹ Indeed, studies using other detection methods have found that the main factors leading to retinal damage caused by hydroxychloroquine include daily dosage.^30,31 One study comparing cone density between high- and low-dose groups found that the density of the cone cells was significantly decreased in the high-dose group.¹⁵ We further analyzed the linear relationship between daily dose and cone density, and found that the daily dose relative to ideal body weight was weakly negatively correlated with cone density in the vertical axis, particularly in the inferior quadrant, but not in the horizontal axis. Again, this result supports our finding that hydroxychloroquine affects cone density in the vertical axis from another perspective. This finding suggests that we should focus on the effect of hydroxychloroquine on the density of vertical cone cells, particularly in the inferior quadrant. 

Current guidelines recommend performing ophthalmic testing for patients whose daily hydroxychloroquine dosage at the start of the medication is more than 5 mg/kg/day.²⁸ A retrospective study of 675 patients using hydroxychloroquine found that 56% of women and 46% of men received higher than the recommended dose of hydroxychloroquine.³² In a cohort study of 3,995 patients, 53.6% took hydroxychloroquine doses of more than 6.5 mg/kg/day.³³ Both studies indicate that the use of hydroxychloroquine exceeding the recommended daily dose is common in daily clinical practice. In our analysis, 25 patients (56.8%) had a daily dose of >6.5 mg/kg/day. Thus, our results are consistent with previous findings, and this outcome may be related to the widespread use of 200-mg tablets, because doses of 200 and 400 mg/day seem to be reasonable for most patients.³⁴ 

This study has several limitations. First, because this was a retrospective, single-center study, and not all patients who were administered hydroxychloroquine were included, selection bias may have occurred. Further, the purpose of this study was to detect retinal lesions in the early stages of the disease, so the mean exposure time was short (mean, 36.5 months), which decreases the chance of detecting hydroxychloroquine effects. In the future, we will recruit a cohort with longer exposure to increase the chances of discerning a drug's effect on cone density. Choosing the region of interest is crucial for the accuracy and repeatability of the image. However, owing to technical limitations, the focal spot center cannot be accurately measured, resulting in an inability to accurately locate the region of interest. Moreover, we strove to align the AO montage with the anatomical landmarks of the central fovea as much as possible, which is crucial for the eccentricity of cone mosaic analysis, and to ensure accurate and repeatable localization of the region of interest.³⁵ Despite these limitations, our study is important, as the specific location of retinal lesions was unclear before our results. 

Overall, we found that hydroxychloroquine decreased the density of the retinal cone cells, with damage occurring predominantly in the vertical axis, of patients with rheumatic diseases compared with age- and sex-matched healthy controls. Cone density loss in the horizontal axis increased with age, and hydroxychloroquine dosage was negatively correlated with cone density in the vertical axis and inferior quadrant. Thus, we suggest that attention should be paid to the changes in vertical cone density, particularly in the inferior quadrant during the early screening of hydroxychloroquine-related retinal injury. 

Acknowledgments

The authors are grateful to Houbin Huang for their help regarding the acquisition of the AO-RTX-DR data. 

Supported by the Medical Science and Technology Innovation Project of Sanya (grant number 2019YW06 to XL), the Scientific Research Project of Health Industry in Hainan Province (grant number 22A200352 to XL), and the Tutorial System of Excellent Medical Undergraduate (grant number LTMCMTS202117 to JT). 

Author Contributions: All authors contributed to the study's conception and design. All authors have read and approved the final version of the manuscript. Data collection was performed by SM, HL, ZZ, and JT. Data analysis was performed by JT and XL. The initial draft of the manuscript was written by JT, HL, SM, and XL. Authors XL, SM, and BH reviewed the successive versions and participated in the study design. 

Conflict of Interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. 

Data Availability Statement: All datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. 

Disclosure: J. Tang, None; H. Liu, None; S. Mo, None; Z. Zhu, None; H. Huang, None; X. Liu, None