Hydroxychloroquine (HCQ) is a widely used medication for the treatment of rheumatologic and inflammatory diseases, such as systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA). However, the use of HCQ is limited by the potential risk of retinopathy [1]. HCQ use might also result in severe visual loss that can aggravate even after cessation [2]. This aggravation might be severity-dependent, as a recent study reported that early retinopathy exhibited limited progression after drug cessation [3]. The incidence of HCQ retinopathy is reported in different studies to range from 1.6%-13.8% [4-6] and is associated with several risk factors, including high daily doses of the drug, pre-existing retinal disease, and long-term use [7,8]. Therefore, regular eye examination and careful monitoring are recommended for patients taking HCQ to minimize the risk of developing retinopathy [1].

In recent years, there has been a growing interest in identifying early signs of HCQ retinopathy, as early detection is crucial for the preservation of vision. In 2016, the American Academy of Ophthalmology revised the recommendations for screening chloroquine and HCQ retinopathy, which include fundus autofluorescence, optical coherence tomography (OCT), and visual field examination evaluations [1].

While it is still unclear which technique provides the best ability for early diagnosis of HCQ retinopathy, some studies have suggested that early Humphrey visual field changes may be an early sign of HCQ retinopathy, prior to the detection of spectral domain-OCT (SD-OCT) changes [9-11]. Other studies have reported early HCQ retinopathy changes in OCT, including a decrease in retinal thickness, attenuation of the ellipsoid zone, or loss of the interdigitation zone, without Humphrey visual field changes [12,13]. Melles and Marmor [14] recently reported that a sequential plot of retinal thickness might show a rapid decrease following an inflection before HCQ retinopathy, providing objective evidence of toxicity. Additionally, one study suggested that choroidal involvement may also be a feature of HCQ retinopathy [15].

Given the lack of consensus on the effective early detection techniques of HCQ retinopathy, our study aimed to investigate and compare early changes in choroidal thickness and retinal thickness, particularly outer nuclear layer (ONL) thickness, in patients using HCQ.

This study is a retrospective review of patients records on HCQ-related retinopathy screening at the Asan Medical Center in Seoul, Korea, between 2016-2021 and follow-up for at least 12 months. The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board (IRB) and Ethics Committee of the Asan Medical Center (IRB approval number: 2021-0279). Exclusion criteria were as follows: history of ocular trauma or intraocular surgery other than cataract extraction, other coexisting retinal/choroidal diseases such as epiretinal membrane and other retinal choroidal vascular diseases, high myopia (refractive errors > -6 diopters), and poor quality of OCT images. Ultimately, we included 67 patients diagnosed with RA and SLE who were taking HCQ medication. We grouped patients with and without HCQ retinopathy on the basis of findings from visual field, OCT, and autofluorescence evaluations. HCQ retinopathy was defined as significant disruption of the outer retina in pericentral, parafoveal, or foveal lesions on OCT and/or hyperfluorescence or hypofluorescence on autofluorescence along with a corresponding visual field defect [1]. Patients with no signs of HCQ retinopathy were divided into two groups: 49 patients in the HCQ-continued group (group A) and 9 patients in the HCQ-discontinued group (group B). The baseline OCT of group A was based on the earliest OCT available during the period of HCQ medication. For group B, the baseline OCT was established based on the earliest OCT obtained after discontinuation of the HCQ. The latter group included individuals who had ceased HCQ for reasons other than retinal problems, such as liver enzyme elevation (n = 3), blood abnormalities (n = 1), and patient refusal (n = 5). Nine patients showing HCQ retinal toxicity were classified as group C. The baseline OCT for group C was based on the closest OCT performed at the time of diagnosis of HCQ-related retinopathy and after the discontinuation of HCQ. Among these nine patients, eight patients showed pericentral pattern, while one patient showed parafoveal pattern. Both eyes of the patient classified as groups A, B, and C were included in the analysis, and the eyes that met the exclusion criteria were excluded from the analysis. If both eyes did not meet the exclusion criteria, one eye was randomly selected for inclusion. A “normal” group comprised individuals without rheumatologic diseases who were not taking HCQ medication. We randomly selected 35 normal eyes of 35 patients who underwent OCT due to contralateral eye disease, particularly of the epiretinal membrane.

We collected data on age, sex, weight, spherical equivalent, time and cumulative dose, dose per weight of HCQ, and time interval between the first and last OCT. Tomographic evaluation was performed with SD-OCT (Heidelberg Engineering). The retinal layers were automatically segmented; this was automatically reported by the software. All data were reviewed for centration and accurate segmentation of each layer. Thickness maps were obtained in the nine lesions, according to the Early Treatment of Diabetic Retinopathy Study (ETDRS): central lesion with diameter of 1 mm, surrounding two concentric rings of 3 mm and 6 mm diameter with four lesions (upper, temporal, inferior, and nasal), as shown in the lower part of Fig. 1. During choroidal thickness (ChT) measurement, the location of the scans was confirmed with infrared reflectance images of the retina. The choroid was defined as the layer between the outer border of the Bruch’s membrane and the inner border of the choroid-scleral interface, as shown in the upper part of Fig. 1. The ChTs were measured using a manual caliper at the subfovea and 1.5 mm nasal and temporal to the fovea in enhanced depth imaging (EDI)-OCT. The ChTs in EDI-OCT were measured by two independent investigators (C.H.M, J.J). Average values of the measured ChTs were used for analysis.

Fig. 1. Example of patients from groups A, B, and C. (A) An example of a patient from group A, a 47-year-old female with systemic lupus erythematosus (SLE). The patient had taken a total dose of 1,796.2 mg of hydroxychloroquine (HCQ; daily dose of 200 mg), with a body weight of 57.35 kg. Significant thinning of choroidal thickness (ChT) in all measured areas and thinning of the outer nuclear layer (ONL) in the superior inner, superior outer, temporal outer, and nasal outer areas can be observed (the darker red area represents the ONL thinning area). (B) An example of a patient from group B, a 50-year-old female with SLE. The patient had taken a total dose of 1,350.7 mg of HCQ (daily dose of 100 mg) and had been on medication for 6 years before discontinuing due to decreased renal function. The patient showed recovery of ChT, but ONL thickness continued to decrease. (C) An example of a patient from group C, a 51-year-old female with SLE. The patient had taken a total dose of 1,470.4 mg of HCQ (daily dose of 200 mg) and had been on medication for 15 years before discontinuing due to HCQ retinopathy. Thinning of both ChT and ONL thickness at baseline can be observed.

Statistical analysis was performed using IBM SPSS ver. 23.0 (IBM Corp.). Wilcoxon signed rank test and paired t-tests were used to analyze the difference between the baseline and last OCT parameters in each group. Between-group analysis was performed using the Mann-Whitney test and t-test. Linear regression analysis was used to confirm the association between HCQ parameters in the HCQ-continued group (group A).

Demographic and clinical characteristics

Table 1 presents the demographic and clinical characteristics of the included patients, which compares these aspects between the groups. There were no significant differences between the groups in mean age, body weight, rates of diabetes, hypertension, and follow-up periods. Groups with HCQ medication history (groups A, B, and C) had no statistical difference in daily doses per body weight.

Table 1

Demographics and characteristics between the groups

Variable	No Retinopathy, HCQ Continued (A) (n = 49)	No Retinopathy, HCQ Discontinued (B) (n = 9)	HCQ Retinopathy (C) (n = 9)	Normal (n = 35)	Multiple Comparison (p-value)
Sex, female	46 (94)	9 (100)	9 (100)	30 (86)	NS
Age (years)	47.3 ± 1.8	49.2 ± 5.1	52.0 ± 3.1	51.9 ± 1.8	NS
DM	4 (8)	2 (22)	1 (11)	4 (11)	NS
HTN	9 (18)	3 (33)	0 (0)	10 (29)	NS
Body weight (kg)	56.56 ± 1.22	56.35 ± 2.88	55.29 ± 1.52	59.21 ± 1.65	NS
Follow-up period (months)	34.85 ± 2.77	29.27 ± 6.99	40.99 ± 10.65	35.94 ± 1.95	NS
Dose/kg	4.12 ± 0.27	4.80 ± 0.84	4.97 ± 0.72	-	NS
HCQ duration (months)	178.55 ± 13.60	89.23 ± 24.77	138.40 ± 22.23	-	0.008*,†

Values are presented as number (%) or mean ± standard deviation. For the variables “dose/kg” and “HCQ duration (months)”, there are no values for the normal group.

HCQ = hydroxychloroquine; NS = no significant in between-group analysis; DM = diabetes mellitus; HTN = hypertension.

*Significant p-value: p ≤ 0.05, Mann-Whitney test; ^†A vs. B.

ChT and ONL thickness at baseline

At baseline, the duration of HCQ medication was variable. While the duration in groups A and B showed a significant difference (p = 0.008), with group A having a longer duration, there were no significant differences observed between groups A and C, nor between groups B and C (Table 1).

The average ChT at baseline in the eyes with HCQ retinopathy (group C, n = 9) had significant thinning compared to group A (n = 49) and the normal group (n = 35). Baseline ONL thickness of the inner and outer ETDRS ring areas was also significantly decreased in eyes with HCQ retinopathy, compared to patients with continuous HCQ treatment without retinopathy and normal controls (Table 2).

Table 2

Mean spherical equivalent baseline and differences of average ChT and average ONL in the eyes of the groups

Variable	No Retinopathy, HCQ Continued (A) (n = 49)	No Retinopathy, HCQ Discontinued (B) (n = 9)	HCQ Retinopathy (C) (n = 9)	Normal (n = 35)	Multiple Comparison (p-value)
Spherical equivalent	-1.07 ± 0.16	-0.68 ± 0.44	-1.54 ± 0.36	-1.05 ± 0.35	NS
Baseline Avg ChT	227.15 ± 11.46	212.33 ± 19.01	171.04 ± 27.35	238.41 ± 5.81	C vs. normal: 0.009*
Last Avg ChT	205.77 ± 8.82	225.04 ± 19.67	167.74 ± 27.40	234.79 ± 12.70	B vs. C: 0.047*
					C vs. normal: 0.009*
ΔAvg ChT	-21.38 ± 4.33	12.70 ± 5.08	-3.29 ± 3.97	-3.61 ± 3.15	A vs. B: < 0.001*
					A vs. C: 0.025*
					A vs. normal: 0.003*
					B vs. C: 0.015*
					B vs. normal: 0.018*
Baseline Avg ONL (ETDRS inner)	66.49 ± 1.43	63.22 ± 2.54	54.22 ± 3.58	66.87 ± 1.34	A vs. C: 0.004*
					C vs. normal: 0.002*
Last Avg ONL (ETDRS inner)	65.48 ± 1.52	62.34 ± 2.15	50.61 ± 3.47	66.96 ± 1.45	A vs. C: 0.001*
					B vs. C: 0.015*
					C vs. normal: < 0.001*
Baseline Avg ONL (ETDRS outer)	54.08 ± 1.26	50.81 ± 2.52	33.31 ± 2.61	53.41 ± 1.02	A vs. C: <0.001*
					B vs. C: 0.001*
					C vs. normal: <0.001*
Last Avg ONL (ETDRS outer)	52.56 ± 1.36	49.28 ± 2.26	31.33 ± 2.57	53.26 ± 0.99	A vs. C: <0.001*
					B vs. C: 0.002*
					C vs. normal: <0.001*
ΔAvg ONL ETDRS inner	-1.01 ± 0.65	-0.86 ± 1.06	-3.61 ± 1.85	0.09 ± 0.73	NS
ΔAvg ONL ETDRS outer	-1.52 ± 0.51	-1.53 ± 0.54	-1.97 ± 1.16	-0.14 ± 0.36	A vs. normal: 0.045*
					B vs. normal: 0.032*

Values are presented as mean ± standard deviatioin.

ChT = choroidal thickness; ONL = outer nuclear layer; HCQ = hydroxychloroquine; NS = no significant in between-group analysis;

Avg = average; ETDRS = Early Treatment of Diabetic Retinopathy Study.

*Significant p-value: p ≤ 0.05, Mann-Whitney test (A vs. B, A vs. C, B vs. C, B vs. normal, C vs. normal), t-test (A vs. normal).

Changes in ChT and ONL thickness during follow-up

During the follow-up period, eyes with continuous HCQ treatment without retinopathy (group A) showed a ChT thinning (-21.38 ± 4.33 μm, p < 0.001). Interestingly, eyes of patients who discontinued HCQ without retinopathy (group B) showed an increase in ChT (12.70 ± 5.08 μm, p = 0.038), suggesting a partial recovery from choroidal thinning after HCQ cessation. Normal eyes showed a mean decrease of 3.62 μm with no significance, indicating minimal variation during the follow-up period. The eyes with HCQ retinopathy also seemed to have a small decrease in average ChT with no significance, indicating limited change, although the possibility of pre-existing thinning cannot be ruled out (Supplementary Table 1).

Fig. 2 illustrates changes in ChT at each location from baseline to the last examination. Group A (n = 49) showed progressive ChT thinning at all measured locations, while the other groups did not exhibit progressive thinning in ChT. Notably, g roup B (n = 9) demonstrated a tendency toward recovery in that the average ChT significantly increased compared to the baseline.

Fig. 2. ChT at each location at baseline and last examination. (A) Avg ChT shows significant thinning. (B) Avg ChT shows a tendency toward recovery. (C) There is no significant change in ChT. (D) There is no significant change in ChT. Error bars indicate standard deviation. HCQ = hydroxychloroquine; ChT = choroidal thickness; SubF = subfovea; Avg = average. *Statistically significant p-values.

At baseline and last follow-up, the ONL of the inner ETDRS r ing area was thinner in g roup C (n = 9) when compared to those in group A (n = 49) and the normal group (n = 35), while at baseline and last follow-up, the ONL of the outer ETDRS ring area was thinner in group C (n = 9) compared to the other groups (Table 2).

During the follow-up period, ONL thickness continued to decrease in all HCQ groups regardless of HCQ intake. Again, normal eyes did not show any significant change in ONL thickness (Supplementary Table 1).

Fig. 3 illustrates changes in ONL thickness at each location f rom b aseline t o t he l ast e xamination. Significant thinning in the ONL thickness of the outer ETDRS circle, corresponding to the perifoveal lesion, was observed in both groups A and B. The baseline outer ETDRS circle ONL thickness of group C (n = 9) was also thinner compared to those of group A (n = 49) and the normal group (n = 35).

Fig. 3. ONL thickness at each location at baseline and last examination. (A) ONL thickness shows significant thinning in the inner nasal and all outer areas except the outer superior. (B) Avg outer area of ONL thickness shows significant thinning. (C) There is no significant change in each area, while the average inner area shows significant thinning. (D) There is no significant change in ONL thickness. Error bars indicate standard deviation. HCQ = hydroxychloroquine; ONL = outer nuclear layer; F = fovea; SI = superior inner; TI = temporal inner; II = inferior inner; NI = nasal inner; SO = superior outer; TO = temporal outer; IO = inferior outer; NO = nasal outer; Avg = average. *Statistically significant p-values.

For more numerical details corresponding to Figs. 2, 3, please refer to Supplementary Table 1.

Relationship between ONL thickness change and ChT in group A

Among eyes with no HCQ retinopathy with intake of HCQ (group A, n = 49), univariate linear regression was used to analyze the relationship between ONL thickness change and clinical characteristics and ChT change. The results showed a significant positive association between the average outer ONL thickness and average ChT changes (β coefficient = 0.038, p = 0.028). Specifically, a s t he C hT d ecreases, the thickness of the outer ONL tends to decrease as well. However, no significant association was found between the change in the inner ONL thickness and other parameters. Multiple linear regression analysis showed that the average outer ONL thickness change was also positively associated w ith t he average ChT change (β coefficient = 0.033, p = 0.042), a nd negatively associated w ith d aily dose per kg (β coefficient = -0.629, p = 0.030) and year (categorical parameter, ≥ 5:1 and < 5:0; β coefficient = -3.904, p = 0.017). Thus, a decrease in ChT is linked with a decrease in outer ONL thickness, while a higher dose of HCQ and a longer exposure duration tend to decrease outer ONL thickness. Regarding the change in average inner ONL thickness, only year showed a significant negative association (categorical parameter, ≥ 5:1 and < 5:0; β coefficient = -4.351, p = 0.048). These results are presented in Tables 3, 4.

Table 3

Linear regression between average ONL thickness change of outer ETDRS ring and clinical characteristics and average ChT change in group A

On ΔOuter ONL Thickness Independent Variable	Univariate			Multivariate
	Coefficient	Standard Error	p-value	Coefficient	Standard Error	p-value
ΔAvg ChT	0.038	0.017	0.028*	0.033	0.016	0.042*
Age	-0.004	0.044	0.935
DM	0.017	1.866	0.993
HTN	1.188	1.311	0.370
Daily dose/kg	-0.462	0.294	0.123	-0.629	0.280	0.030*
Year (≥ 5) (≥ 5:1; < 5:0)	-3.207	1.619	0.054	-3.904	1.574	0.017*
Spherical equivalent	-0.326	0.350	0.356
Constant					2.094	0.041*

ONL = outer nuclear layer; ETDRS = Early Treatment of Diabetic Retinopathy Study; ChT = choroidal thickness; Avg = average; DM = diabetes mellitus; HTN = hypertension.

*significant p-value: p ≤ 0.05.

Table 4

Linear regression between average ONL thickness change of inner ETDRS ring and clinical characteristics and average ChT change in group A

On ΔInner ONL Thickness Independent Variable	Univariate			Multivariate
	Coefficient	Standard Error	p-value	Coefficient	Standard Error	p-value
ΔAvg ChT	0.026	0.021	0.224	0.022	0.021	0.311
Age	-0.006	0.056	0.921
DM	0.347	2.337	0.882
HTN	0.158	1.642	0.360
Daily dose/kg	-0.229	0.376	0.546	-0.428	0.380	0.266
Year (≥ 5) (≥ 5:1; < 5:0)	-3.923	2.033	0.060	-4.351	2.132	0.048*
Spherical equivalent	-0.337	0.440	0.448
Constant					2.836	0.076

ONL = outer nuclear layer; ETDRS = Early Treatment of Diabetic Retinopathy Study; ChT = choroidal thickness; Avg = average; DM = diabetes mellitus; HTN = hypertension.

*Significant p-value: p ≤ 0.05.

In the current study, we investigated whether thinning of the choroidal layer or ONL could be early indicators of HCQ toxicity before developing HCQ retinopathy. First, we found that the choroidal layer and the outer ETDRS ONL were significantly thinner in patients with HCQ retinopathy compared to normal patients. To check the early HCQ effect among patients with no retinopathy, we compared eyes with continuous HCQ intake (group A) to those who discontinued HCQ (group B). Group A showed a significant decrease in both ChT and outer ETDRS ONL thickness during the follow-up, whereas group B showed a decrease in only the outer ETDRS ONL thickness but an increase in ChT during the follow-up periods. The extent of recovery in ChT was prominent at the subfoveal and temporal areas among topographical locations but was significant only in the temporal area. In contrast, a partial recovery in ChT was not observed after discontinuation of HCQ among patients with HCQ retinopathy.

The exact mechanisms of HCQ retinopathy are not completely understood. There are debates on HCQ effects on the choroid and retina.

Concerning the choroid, choroidal thinning in HCQ retinopathy has been documented in some reports [15-17], and some have proposed that HCQ interaction with melanin is related to toxicity [18]. HCQ has an ionic binding affinity for melanin, causing it to accumulate in melanin-rich structures, such as the choroid and retinal pigment epithelium (RPE) [19]. The concentration of HCQ in animal models has shown that HCQ concentrate in the uvea is 9-32 times that of the retina [19,20]. While choroidal melanin is continuously synthesized throughout an individual’s life, melanin in the RPE is synthesized only during a short period of fetal and perinatal life, and its production decreases significantly thereafter [21-23]. This might explain the progressive choroidal thinning in HCQ-continuing patients (group A), in contrast to the recovery of choroidal thinning in patients who stopped taking HCQ (group B).

Regarding the retina, inner retinal damage associated with HCQ use has been reported in some studies, while others have reported outer retinal damage only [24-27]. The outer retinal layers, consisting of the RPE and photoreceptor layers, are nourished by the choriocapillaris [28]. It is reasonable to consider the possibility that the choroid plays a role in the development of HCQ retinopathy. The results from our study showed that ONL thinning persisted even after the patients had discontinued HCQ (group B); this finding is consistent with the finding from another study conducted by de Sisternes et al. [29] However, in our study, a significant ONL thinning was in the perifoveal lesion, which corresponds to the outer circle of ETDRS, while de Sisternes et al. [29] found ONL thinning in a parafoveal lesion. This difference in findings may be attributed to ethnic reasons, as Asian patients may exhibit a more pericentral pattern [30,31]. In this study, 16 eyes showed a pericentral pattern, and 1 eye showed a parafoveal pattern. We observed a thinning of ONL in both the inner and outer circles of the ETDRS, but it was more pronounced in the outer circle.

The current recommendations for screening HCQ retinopathy aim to identify definitive signs of toxicity at an early stage in order to prevent vision loss, rather than discontinuing the drug in response to borderline abnormalities [1]. However, if patients taking HCQ develop retinal pigmentary or outer retinal changes visible on OCT images, this may indicate that irreversible damage has already occurred in the outer retinal layers. Therefore, early detection of HCQ retinopathy is crucial in minimizing functional deficits. Our study demonstrated progressive ChT thinning in patients who were taking HCQ prior to developing HCQ retinopathy. However, in patients who discontinued HCQ, we observed that this thinning stabilized or partially recovered, suggesting that discontinuing HCQ before the onset of retinopathy may be helpful for recovery. In contrast, for patients with established HCQ retinopathy, the ChT was already thin at baseline, and there was no significant recovery observed. Based on these findings, it is possible that there is a point of no return after which damage to the choroid becomes irreversible. Alternatively, considering that most patients with HCQ retinopathy have ceased HCQ usage for a considerable period, the changes in ChT might have already been completed. Further cohort studies are needed to determine the point of early HCQ retinopathy.

Results of the univariate and multiple linear regression analyses showed a significant association between ChT and ONL thickness changes in group A. In particular, ONL thickness change at the outer ETDRS circle was significantly associated with ChT change and HCQ retinopathy risk factors, such as dose per kg and duration of HCQ use. This finding suggests that monitoring changes in ChT and ONL thickness in the outer ETDRS area may be a useful approach for early detection of HCQ retinopathy. Considering the larger magnitude of thickness change in ChT, monitoring may be more straightforward in ChT.

There are several limitations to this retrospective study that should be considered. One potential limitation is the possibility of selection bias in patient selection, as only patients who received screening at a specific medical center during a specific time period were included. Another limitation of this study is that the decision to discontinue HCQ medication was made for reasons other than retinal problems, which could introduce confounding factors.

Furthermore, due to the rarity of patients who discontinued HCQ for reasons other than retinal problems and were followed up for long periods, additional patients are needed to confirm the findings of this study. Due to the rarity of patients who discontinued HCQ, a limited number of individuals were included in the normative group of this study in order to equalize sample sizes, which can also lead to bias.

Additionally, in another study targeting SLE patients, it was mentioned that patients using anticoagulants or those with lupus nephritis may exhibit thinner ChT [32]. Therefore, in this study targeting SLE or RA patients, the possibility of bias occurring due to the disease activity of SLE cannot be excluded.

Although there were no significant differences in age among the groups in this study, and the follow-up period also did not show significant differences, it should be acknowledged that age-related choroidal thinning has been observed in a number of studies, implying that age differences among the groups could be a potential confounding factor [33,34].

Finally, since this study was conducted with a single ethnic group in Korea, the results may not be generalizable to other populations. Nonetheless, the present study highlights the importance of regular ophthalmologic examinations for patients on HCQ therapy to prevent irreversible damage. The findings of this study could contribute to the development of better screening and monitoring strategies for patients taking HCQ.

In conclusion, the findings of this study suggest that monitoring changes in ChT and ONL thickness may be useful for detecting HCQ retinopathy before its onset. This could lead to earlier dose adjustment or shorter follow-up to maintain better visual outcomes for patients undergoing long-term HCQ therapy. However, these results should be validated by performing larger studies with more patients to confirm the findings from this study and establish an early diagnostic method for HCQ retinopathy.

Portions of this paper were presented at the Association for Research in Vision and Ophthalmology (ARVO) 2022 Annual Meeting as poster presentations with interim findings. The poster abstracts were published in Investigative Ophthalmology and Visual Science, June 2022, Vol 63.

The authors declare no conflicts of interest relevant to this article.

Conception (C.H.M., J.Y.L., Y.H.Y.); Design (C.H.M., J.Y.L., Y.H.Y.); Data acquisition (C.H.M., J.J.); Analysis (C.H.M., J.J.); Writing (C.H.M.), Review (C.H.M., J.Y.L., Y.H.Y.); Final approval of the article (All authors)