Safety analysis of repeated low-level red light therapy on cone mosaic

For the complete article and additional details, click the link below:

https://www.sciencedirect.com/science/article/pii/S1572100026001407?via%3Dihub#bib0001

1. Introduction

Myopia is a worldwide health issue which was estimated to affect half of the world’s population by 2050 [1]. Once myopia progressed into high myopia, sight-threatening complications such as myopic macular degeneration and retinal detachment, can be at a higher risk [2,3].

Repeated low-level red-light (RLRL) therapy has gained significant attention around the world. RLRL therapy is typically administered using desktop or laptop devices for controlled exposure; whereas most RLRL therapies are currently done with laptop devices, some are done with red-light-emitting spectacles. Children are instructed to look into this device for 3 min, twice a day [4]. By irradiating the retina, it is thought to increase the choroidal thickness and slow axial elongation. Several clinical studies have shown the myopia control effect [[4], [5], [6], [7]]. It is observed that the RLRL therapy could thicken the choroid across the macular region, potentially improving the choroidal blood flow and controlling myopia progression [8,9].

However, there were concerns over the safety of RLRL therapy. As defined by the International Electrotechnical Commission standard 60825–1:2014.2, Class 1 lasers are low-powered devices that are considered safe from potential hazards. Although certain RLRL therapy devices are classified as Class I lasers, prolonged exposure (i.e., extended viewing time) or larger-than-average pupil size can still lead to photochemical or thermal damage to the retina [10]. One case of foveal structural damage and reversible vision loss was reported after 5 months of RLRL therapy. Photoreceptor and retinal pigment epithelium damage was noticed in the optical coherence tomography (OCT), including disruption of the foveal ellipsoid zone and discontinuity of the interdigitation zone [11]. These and similar risks have also been highlighted in additional reports. The relative reflectance of the ellipsoid zone (rEZR), photoreceptor outer segment (rPOSR) increased after RLRL therapy [12]. Reduced cone density in the paracentral fovea and other subtle retinal abnormalities were also reported [13].

While RLRL therapy presents a promising avenue for myopia control, it is imperative to thoroughly assess its safety profile. Currently, clinical assessments of the safety of RLRL therapy primarily involve measurements of best-corrected visual acuity (BCVA), fundoscopy, OCT, and afterimage duration [14,15]. These methodologies are primarily focused on functional assessments or designed to evaluate the overall retinal status and structural integrity, yet lack the capability for cellular-level analysis. Adaptive optics (AO) is a technology that permits observation of optical aberrations in real-time, enabling high-resolution imaging of the retina [16]. This approach allows for the direct observation of individual cone photoreceptors with resolutions of approximately 2 to 3 μm, providing valuable insights into retinal health and the potential impacts of RLRL therapy[17].

The human macular photoreceptor mosaic is characterized by a precise yet complex spatial distribution of three cone subtypes[18]. Long-wavelength (L, red-sensitive) and medium-wavelength (M, green-sensitive) cones are densely packed, typically in a roughly 2:1 ratio, though this ratio exhibits significant inter-individual variability [19]. In contrast, short-wavelength (S, blue-sensitive) cones constitute only 5–10% of the total cone population and are generally absent from the central ∼0.35 mm of the fovea, which is also essentially rod-free. While adaptive optics scanning laser ophthalmoscopy (AOSLO) allows for the direct visualization of individual photoreceptors at cellular resolution, it does not inherently distinguish between these spectral subtypes. Given that RLRL therapy utilizes 650 nm light—a wavelength to which l-cones are most sensitive—it is theoretically possible that such treatment could influence the individual spatial distribution or density of these photoreceptor populations. Therefore, evaluating the collective structural stability of the cone mosaic is essential for assessing the safety profile of RLRL therapy.

To make a more detailed evaluation at the cellular spatial distribution level of the safety of RLRL therapy, we aim to use adaptive optics scanning laser ophthalmoscopy (AO-SLO) to analyze the cone photoreceptors at a cellular level.

2. Materials and methods

2.1. Subjects

We conducted a cross-sectional study at the Ophthalmology Department of the First Affiliated Hospital of Chongqing Medical University. This study was conducted in adherence to the principles of the Declaration of Helsinki, and written informed consent was obtained from parents or legal guardians, and written assent was provided by all participating minors. Institutional Review Board/Ethics Committee approval was obtained (No. 202417601). Participants were consecutively recruited from the pediatric myopia clinic of this single tertiary referral center. As a tertiary referral center, this setting may select for children with more severe refractive errors and higher socioeconomic status compared to community-based populations, due to referral patterns for specialized myopia management and the associated costs and accessibility of advanced treatments such as RLRL therapy. All epidemiological inferences and analyses should therefore be interpreted within the context of this selection bias.

Each measurement was performed by a single experienced ophthalmologist, who repeated the examination twice in each subject to verify the method’s repeatability.

2.2. Clinical assessment

Each subject underwent a comprehensive ophthalmologic examination during daytime clinic hours (typically 09:00–17:00), including cycloplegic refraction, BCVA, axial length (IOLMaster-700, Carl Zeiss Meditec, Dublin, CA, USA), intra-ocular pressure, swept-source optical coherence tomography (SS-OCT) (TowardPi Medical Technology, Beijing, China) and AO-SLO imaging. OCT included radial scan of the macular region and the structure of the retina was assessed for safety evaluation. Cycloplegia was achieved using Compound Tropicamide eye drops (Zhuobian, 0.5% tropicamide and 0.5% phenylephrine, Xingqi Pharmaceutical Co., Ltd., Shenyang, China). Three drops were instilled in each eye, 5 min apart, with cycloplegia confirmed after 30 min by assessing pupil dilation and absence of light reflex.

2.3. Group classification criteria

Subjects were divided into three groups: Non-Myopic group, Myopia-Untreated group and Myopia-RLRL group. No subgroup analyses (e.g., age/sex stratification within cohorts) were conducted. The diagnosis of myopia was based by a spherical equivalent refraction (SER) after cycloplegia ≤−0.50D Children with a cycloplegic SER >−0.50D were classified as the Non-Myopic group. Children who received the RLRL therapy >8 months were classified as Myopia-RLRL group. The 8-month duration was selected based on a combination of clinical feasibility, prior literature, and the need to balance sufficient exposure time with practical study constraints. Cao et al. found significant changes in axial length and spherical equivalent refraction in myopic children after 6–12 months of RLRL therapy [6]. We chose 8 months as an intermediate duration to assess retinal safety while ensuring adequate exposure to detect potential changes in cone photoreceptor metrics and microvascular parameters using AO-SLO. This cross-sectional analysis assessed outcomes at a single index visit (AO-SLO imaging date). Exposure status (RLRL therapy) was defined based on treatment history prior to the index visit. For analyses of duration effects (e.g., >8 months therapy), follow-up started at therapy initiation, and imaging windows were standardized relative to this start date. Other inclusion criteria include a BCVA of ≥16/20, a cylinder correction within 2D Participants were excluded from our present study if they had amblyopia, high myopia (SER<−6.00D) or other ocular disease (such as congenital cataract, glaucoma, corneal scar…), history of any ocular surgery, and any systemic diseases or conditions that may affect visual function and development.

Children in the Myopia-RLRL group received RLRL treatment twice daily, with each session lasting 3 min and at least 4 h between treatment, 5 days per week (daily). The RLRL intervention was provided by a desktop device (Eyerising, Suzhou Xuanjia Optoelectronics Technology, Kunshan city, China). This technology incorporates semiconductor laser diodes that emit low-intensity, single-wavelength 650 (10)-nm red-light laser beam at an illuminance level of approximately 1600 lx through the pupil to the fundus. The laser beam was parallel to ensure uniform macular irradiance and avoid any focal concentration of energy. The light power of this RLRL device entering a 4-mm pupil (the maximum pupil size under the condition of bright-light exposure over 10 s) is 0.29 mW. These laser power levels are classified as Class 1. Guardians/parents of the children received training to oversee the intervention program at home. To ensure accurate tracking and recording of treatment history, the device was integrated with an automated diary function and connected to the internet. In addition, two investigators were tasked with managing intervention compliance. To enhance treatment compliance, parents or legal guardians received weekly reminders about the intervention program in an attempt to maintain consistency and effectiveness in the intervention process.

2.4. Retinal imaging

AO images were captured using a commercialized AO-SLO system (Mona II, Robotrak Technologies, Nanjing, China). This AO-SLO system utilizes an 840 nm light source with a full-width half-maximum (FWHM) of 40 nm.

After administering tropicamide to dilate the pupil, the subject was instructed to fixate on a green target controlled by the operator to capture retinal images corresponding to the field of view on the retina of 2.4° × 2.4° (approximately 700 × 700 μm). To better estimate cone morphology, the retina was divided into five regions of interest (ROI): four ROIs at 1° eccentricity (≈0.30 mm) from fovea: nasal, temporal, superior, inferior, and one ROI at 0.5° eccentrical superior-temporal (≈0.15 mm) from fovea. (Fig. 1)

For quantitative cell analysis, the software uses artificial intelligence-based algorithm for automatic photoreceptor segmentation and generates statistical descriptors of photoreceptor morphology properties including density, spacing, dispersion and regularity. Cone density refers to the number of cones per square millimeter, Cone cell spacing is mean distance to nearest neighboring cones. Cone cell dispersion is the ratio of the standard deviation of distances between a cell and its nearest neighbor to the average of these distances, quantifying the extent of cell spread or clustering. Cone cell regularity is the proportion of cells with a specific number of nearest neighbors within a set distance, indicating how uniformly cells are distributed. A regularity value close to 6 typically signifies a more uniform cell arrangement. In blood vessel mode, the vessel morphology parameters including lumen diameter, vessel diameter and wall-to-lumen ratio are manually measured, while blood flow speed and flow rate are automatically measured and calculated using line-scan technique.

2.5. Endpoint definition

Primary endpoint: Cone density in the central 2.4° × 2.4° macular area.
Secondary endpoints: Cone spacing (μm), regularity (%), dispersion (%); retinal microvasculature parameters (lumen diameter, blood flow velocity (mm/s).

2.6. Statistical analysis

Statistical analysis was performed using SPSS statistics (version 29. IBM Corp., Chicago, IL). Continuous variables were reported as mean and standard deviation (SD). Normality of the data was checked using the Shapiro-Wilk test. Due to the non-normal distribution of variables in this study, those variables were compared by means of Mann-Whitney U test and Kruskal-Wallis test. Spearman correlation analysis was employed to evaluate the associations among cone morphology, SER and AL. The power analysis was conducted using our preliminary study data demonstrating mean cone densities of 49,053 ± 7834 cones/mm² (non-myopic group), 42,179 ± 3844 cones/mm² (myopia-RLRL group), and 40,319 ± 5219 cones/mm² (myopia-untreated group). Based on these parameters, a minimum of 14 eyes per group was required to achieve 80% statistical power (β=0.20) at α=0.05 significance level for detecting intergroup differences via Kruskal-Wallis test in PASS 15.0.5 software. This calculation accounted for potential non-normal distribution of photoreceptor metrics and multiple comparison adjustments.

To account for the paired nature of the eye data, generalized estimating equations (GEEs) were used with an exchangeable working correlation structure, clustering observations by Subject ID, while adjusting for age and gender as covariates. All primary analyses report effect size indices with 95% confidence intervals (CIs).

For parametric models (GEE/ multivariate linear regression), β coefficients (mean differences) are presented.

3. Results

Initially, 68 eyes of 34 subjects who met the study eligibility criteria were enrolled. 2 eyes were subsequently excluded because of poor-quality AO-SLO images, one eye each from Myopia-Untreated group and Non-Myopic group. The primary and secondary outcomes had <5% missing values. The blood flow analysis was conducted under the prerequisite that the sampling confidence interval (CI) was greater than 0.7.

3.1. Demographic and clinical characteristics

Demographic data for the Myopia-RLRL, Myopia-Untreated group and Non-Myopic group were presented in Table 1. The average axial length (AL) was 24.67±0.47 mm in the Myopia-RLRL group, 24.57±1.12 mm in the Myopia-Untreated group, and 22.60±1.07 mm in the Non-Myopic group. The average spherical equivalent refraction (SER) was −2.21±1.53 D in the Myopia-RLRL group, −2.26±1.52 D in the Myopia-Untreated group, and 1.29±1.46 D in the Non-Myopic group. There was no significant difference in the AL and SER between the Myopia-RLRL group and Myopia-Untreated group (Mann-Whitney U, p = 0.863, 0.849).

Table 1. Demographic characteristics and quantitative analysis of cone cell imaging in subjects from three groups.

Empty Cell	Myopia-RLRL (n = 20)	Myopia-Untreated (n = 32)	Non-Myopic (n = 14)	RLRL vs. Untreated	RLRL vs. Non-Myopic	Untreated vs. Non-Myopic	Referent	Non-Myopic		RLRL
Empty Cell	Myopia-RLRL (n = 20)	Myopia-Untreated (n = 32)	Non-Myopic (n = 14)	Model 1			Model 2
Empty Cell	Empty Cell	Empty Cell	Empty Cell	P value			Empty Cell	P value	β (95% confidence intervals)	P value	β (95% confidence intervals)
Age	10.90±1.86	12.00±3.90	9.29±5.55	0.656	0.011	0.002	Untreated	0.358	−0.358 (−1.121, 0.405)	0.683	−0.416 (−2.412, 1.580)
Gender, M/F	12/8	17/15	5/9	0.627	0.163	0.277	Untreated	0.502	−0.001 (−0.003, 0.001)	0.331	0.464 (−0.479, 1.396)
Axial length, mm	24.67±0.47	24.57±1.12	22.60±1.07	0.691	<0.001	<0.001	Untreated	<0.001	−0.161 (−0.164, −0.159)	0.027	0.624 (0.073, 1.175)
SER, D	−2.21±1.53	−2.26±1.52	1.29±1.46	0.985	<0.001	<0.001	Untreated	0.058	1.421 (−0.050, 2.893)	0.386	−0.522 (−1.704, 0.659)
Density (/mm²)	41,274.00± 6227.62	40,202.21± 7364.91	45,648.94± 5092.01	0.627	0.058	0.013	Untreated	0.017	5234.437 (936.174, 9532.701)	0.758	667.39 (−3570.307, 4905.088)
Spacing (μm)	3.95±0.12	3.97±0.30	3.75±0.19	0.998	0.010	0.005	Untreated	0.017	−0.215 (−0.391, −0.038)	0.803	−0.016 (−0.139, 0.107)
Dispersion (%)	16.47±2.43	17.36±2.14	17.80±2.25	0.030	0.024	0.605	Untreated	0.886	−0.127 (−1.865, 1.611)	0.251	−0.91 (−2.466, 0.645)
Regularity (%)	95.47±1.55	95.14±1.23	95.65±1.56	0.243	0.969	0.319	Untreated	0.802	0.116 (−0.794, 1.027)	0.712	0.171 (−0.736, 1.078)

Model 1: Non-parametric analysis of Kruskal-Wallis test, Model 2: Generalized estimating equations (GEEs).

3.2. Qualitative and quantitative assessment with AO images

In the 5 ROIs, there was no significant difference of dispersion among the three groups. Non-myopic group had higher cone density than Myopia-Untreated group in ROI 2 (p = 0.025, β=5709.10 /mm2) (superior 0.3 mm from the fovea). (Supplementary Table 3) The spacing was increased in ROI 1 and ROI 2 in Myopia-Untreated group than Non-Myopic group (p = 0.006, β=−0.31 μm, p = 0.014, β=−0.284 μm, respectively). The cone regularity in ROI 2 was reduced in the Myopia-RLRL group than Myopia-Untreated group (p = 0.012, β=−1.385%)(Supplementary Table 3). The Kruskal-Wallis test results were consistent with the above findings. (Supplementary Figure 1)

Within the central 2.4° × 2.4° area, the Non-Myopic group exhibited higher cone density and smaller cone spacing (Fig. 2A-B). The Myopia-Untreated group showed significantly reduced density compared to the Non-Myopic group (Fig. 2A), while both Myopia-Untreated and Myopia-RLRL myopia groups displayed enlarged spacing relative to Non-Myopic controls (Fig. 2B). No differences in regularity were observed among the three groups (Fig. 2D). These findings also remained consistent in the GEE model (Table 1). Lower cell dispersion was observed in the Myopia-RLRL group compared to both the Myopia-Untreated and Non-Myopic groups. (p = 0.030, Kruskal-Wallis test). Following adjustment for gender and age using Generalized Estimating Equations (GEEs), dispersion in the Myopia-RLRL group showed a non-significant trend toward reduction compared to the Myopia-Untreated group (p = 0.251, β = −0.91%). (Fig. 2C, Table 1).

By comparing the horizontal and vertical meridian, no significant difference among the three groups was found in cone cell density and dispersion (Supplementary Figure 2A and D). However, the regularity was significantly lower in the horizontal meridian than the vertical meridian in the Myopia-Untreated and Non-Myopic groups (Supplementary Figure 2C). The spacing was significantly larger in the vertical meridian compared to the horizontal meridian across all three groups (Supplementary Figure 2B).

Some atypical findings were observed in 8 eyes. Two luminous, circular spots were observed at parafoveal locations in right eye from subject no 10 from the Myopia-Untreated group (Supplementary Figure 3. A1–2). This subject has denied any previous trauma or exposure to laser. However, discontinuity of the interdigitation zone was observed in the corresponding area (Supplementary Figure 3 A-3). Silk membrane-like materials at the macula were observed in right eye from subject no 6 and no 7 from Myopia-RLRL group and Myopia-Untreated subjects no.23 and no.26 (Supplementary Figure 3 B1–4 and Movie 2). After careful examination of fundus, we assume the membrane-like material is likely the posterior vitreous limiting membrane. Clumps of hyperreflective material were observed in the parafoveal area from the right eye from subject no 20 from Myopia-RLRL group, but no abnormalities were detected in OCT images (Supplementary Figure 3 C1–2).

3.3. Correlation analysis

The cone cell density was positively correlated with the SER in the central 2.4 × 2.4° area (p<0.001, r = 0.540), as well as the 5 ROIs (p<0.01, r = 0.330∼0.451).

The cone cell spacing was negatively correlated with the SER in the central 2.4 × 2.4° area (p<0.001, r=−0.611), as well as the 5 ROIs (p<0.001~p = 0.028, r=−0.271~0.452). The cone cell dispersion was not significantly correlated with SER, age or AL in the central 2.4 × 2.4° area. In the 5 ROIs, no significant correlation was found between the cone cell dispersion and SE and AL. Age was negatively correlated with cone cell regularity (p = 0.002, r=−0.366). AL was also negatively correlated with cone cell regularity (p = 0.014, r=−0.300). (Fig. 3)

A multivariate linear regression analysis was conducted regarding the correlations between cone cell density, axial length, gender and RLRL therapy use. Those results indicate a significant negative correlation between AL and cone cell density in the central 2.4 × 2.4°area (p < 0.001), with a variance inflation factor (VIF) <2, suggesting no significant multicollinearity (Supplementary Table 2). Regarding RLRL treatment duration, we found no significant correlations with cone cell density in the central 2.4 × 2.4°area or the five ROIs (p > 0.05).

3.4. Quantitative assessment of retinal arteriole and venule

The retinal arteriolar morphological analysis showed that there was no significant difference in the vessel diameter (μm), lumen diameter (μm), wall diameter (μm), wall-to-lumen ratio, wall cross-section area (μm[2]), blood flow velocity (mm/s) between the Myopia-RLRL and Myopia-Untreated using Model 1 (Supplementary Table 1 and Moive 1). Following adjustment using GEEs, lumen diameter in the Myopia-RLRL group showed a significant reduction compared to the Myopia-Untreated group (p = 0.020, β = −13.696 μm). However, none of the other parameters showed significant differences, consistent with the Kruskal–Wallis test.

4. Discussion

There were many discussions and concerns over the safety of RLRL therapy. In our study, we offered a novel non-invasive way of evaluating the macula at a near histological level. The qualitative and quantitative analysis suggested that in myopic children, the cone cell had a lower density and higher spacing. The Myopia-RLRL group exhibited lower cone dispersion compared to both Non-Myopic and Myopia-Untreated groups. We found a high correlation of cone cell density, spacing and regularity with age and SER. The vessel parameters were not significantly different among the three groups.

The absence of significant differences in cone density and dispersion among the three groups within the predefined ROIs indicates that RLRL therapy may not exert a substantial impact on photoreceptor cell population. However, the observed reduction in cone regularity within ROI 2 following RLRL therapy raises the possibility that there could be localized structural alterations. Reduced regularity in cone photoreceptor arrangement was reported in laser-induced retinopathy [17] and retinopathy of prematurity [20]. While this finding does not directly indicate harm, it underscores the need for long-term follow-up to monitor potential subclinical alterations. Subtle irregularities could reflect transient metabolic stress during photobiomodulation (PBM) or early signs of photoreceptor maladaptation. For instance, in patients with retinitis pigmentosa (RP) who exhibited normal BCVA, noticeable spacing irregularities was reported [21]. In the context of RLRL therapy, the observed localized decline in cone regularity in ROI 2 still warrants close attention.

While one AO-based study of RLRL has recently emerged and sparked vigorous debate, data comparisons reveal discrepancies. Wang et al. reported a cone density of 28,700 ± 360 cones/mm² at 0.3 mm nasal eccentricity in non-RLRL myopes[13]. This contrasts with Curcio et al.'s histological benchmark of 53,500 cones/mm² for the same location[22]. The difference may relate to cell counting methodology: Wang's group used manual Image J counting on montage images, whereas our study employed an AO-SLO integrated AI algorithm for automated photoreceptor segmentation. Our result at the same location (non-RLRL myopic group: 44,221.06±8684.48 cones/mm²) aligns more closely with histology, suggesting potentially greater accuracy (Supplementary Table 3). It is worth noting that the potential impact of methodology and device variability on such comparisons was a key point of discussion in our formal comment on Wang's study [23]. Furthermore, Wang's RLRL cohort used four different devices, introducing potential device-related bias, which our study minimized by using a single standardized device type.

In the central 2.4° × 2.4° area analysis, the myopic group exhibited a statistically significant reduction in cone cell density and an increase in inter-cone spacing. This finding is consistent with correlation analyses revealing strong associations between cone density/spacing and both axial length (AL) and spherical equivalent refractive error (SE), which is also consistent with previous studies showing that cone density was decreased in myopia and that the cone density was highly correlated with AL [[24], [25], [26]]. It suggests that myopia progression may induce measurable alterations in macular cone mosaic due to axial elongation. Importantly, both the extent (degree) of myopia and its duration significantly may influence cone photoreceptor density. Although direct longitudinal data on myopia duration are limited, cross-sectional studies show that greater cumulative axial elongation (reflecting longer disease duration) amplifies the reduction in cone packing density [27,28]. Progressive axial elongation with longer duration of myopia exposure mechanically stretches the retina. This stretching increases inter-cone spacing and reduces retinal thickness, thereby lowering packing density. Specifically, as the eye elongates, tangential tensile forces may lead to spatial rearrangement of photoreceptors and potential functional consequences. What seemed to be in conflict with our results of analysis of ROIs is that, in the central 2.4° × 2.4° area, no significant differences in cone regularity were observed among the three groups. This finding may reflect limitations in spatial resolution when assessing the foveal centralis, where densely packed cone photoreceptors challenge the precision of AO-SLO. Specifically, the foveal cone mosaic—characterized by peak cell density (∼199,000 cones/mm²) [22] and minimal inter-cone spacing (∼2.5 μm)—approaches the diffraction limit of conventional AO-SLO systems (∼3 μm lateral resolution) [29]. Consequently, subtle irregularities in cone arrangement, particularly within the rod-free foveola, may be obscured by instrument noise or motion artifacts.

The observed reduction in cone cell dispersion within the central 2.4° × 2.4° retinal area in the Myopia-RLRL group compared to both Non-Myopic and Myopia-Untreated groups requires cautious interpretation due to methodological constraints. GEE analysis adjusted for sex and age indicated a non-significant trend toward reduced cell dispersion in the Myopia-RLRL group. Cone cell dispersion quantifies spatial heterogeneity in photoreceptor distribution, reflecting the degree of cellular clustering. Whether RLRL therapy directly influences cone cell patterning remains unresolved and warrants further investigation. Prior studies hypothesize that RLRL exerts its effects through photobiomodulation. PBM mechanisms are reported to involve mitochondrial activation, enhanced antioxidative defense, and inflammatory reduction [[30], [31], [32]]. By targeting cytochrome C oxidase enzyme in mitochondria through light absorption, PBM activates molecules such as nitric oxide, ATP, calcium ions and reactive oxygen species (ROS) [33]. It is reported to inhibit hypoxia-inducible factor 1α and promote collagen synthesis [34]. Our previous study found higher cone cell dispersion in high myopia patients [35]. Parallels with diabetic retinopathy studies suggest that decreased cellular dispersion correlated with milder diabetic retinopathy severity and better best-corrected visual acuity [36], suggesting tighter cellular organization may reflect retinal homeostasis. Nevertheless, the photobiomodulation effect is a preliminary speculation based on the current study’s findings. The precise biomechanical mechanisms underlying this phenomenon in myopia management require further investigation.

The hypothesized mechanism for RLRL therapy in myopia control has centered on its potential to enhance retinal and choroidal microcirculation [14,37]. However, our vascular analysis revealed no significant differences in arteriolar/venular parameters—including vessel diameter, lumen diameter, wall-to-lumen ratio, or blood flow velocity—among the Myopia-RLRL group, Myopia-Untreated, and Non-Myopic controls. Previous OCTA studies with a resolution of 20 μm could only measure choroidal and retinal blood flow density. Studies by others have shown that RLRL therapy in myopic children increases choroidal thickness and vascular density, but does not alter retinal blood flow density or vascular integrity [[38], [39], [40]]. The AO-SLO used in our study, with its ∼3 μm resolution, enables clearer visualization of retinal microvasculature and calculates blood flow velocity by tracking erythrocyte movement. However, our findings suggest that retinal small vessels may exhibit strong adaptability to RLRL-induced hemodynamic modulation, consistent with a recent study on the dose-dependent effects of RLRL therapy, which reported a decline in enhanced microcirculatory parameters after cessation of RLRL treatment [41]. This plasticity may explain the apparent dichotomy between short-term vascular modulation and long-term structural preservation observed across studies. Future investigations employing multimodal imaging—combining AO-SLO with OCT angiography and widefield fluorescein angiography—could better delineate RLRL's effects on the retinal-choroidal microcirculatory unit. Additionally, functional assessments such as retinal oximetry or metabolic imaging may complement structural analyses to fully characterize therapeutic mechanisms.

The limitation of our study is the relatively small sample size and the single-center nature. Selection bias may arise from recruiting participants solely through a tertiary academic center, potentially limiting representativeness to children accessible to specialized care. Secondly, we could not accomplish the arteriolar assessment for all of the subjects because of the participants’ short fixation time. However, rigorous quality control was maintained by excluding vascular datasets with confidence indices <0.7. Thirdly, due to the cross-sectional nature of this study, the associations observed cannot be interpreted as causal relationships. Longitudinal studies are required to establish causality. To have a more comprehensive understanding of the changes after RLRL therapy, future studies are essential to investigate the functional parameters like microperimetry and electroretinography (ERG). It is also important to note that 8 out of 10 participants (80%) in the Myopia-RLRL group were still actively receiving therapy at the time of AOSLO imaging, with only two subjects having recently discontinued. RLRL therapy is known for a significant modest rebound effect after stopping the therapy[42]. Had measurements been performed a few months later during a potential rebound phase, rapid axial elongation could have further stretched the retina, leading to decreased cone density and increased inter-cone spacing. Despite rigorous exclusion of comorbidities and adjustment for confounders (age and sex), unmeasured factors (e.g., environmental light exposure, genetic susceptibility) may persist. Residual confounding from subtle baseline imbalances cannot be fully ruled out.

In conclusion, we found that age and axial elongation during myopia progression were significantly related to the structural changes of macular cone cells. Eight-month RLRL exposure did not cause significant cone cell density decrease or evident macular abnormality. RLRL therapy might lower the cone cell dispersion in certain regions of the macula. Diameters of the retinal arterioles and venules, as well as flow velocity measured by AOSLO, remained unchanged after RLRL treatment compared with both myopia-untreated and emmetropic controls. While lower cone dispersion was observed in RLRL-treated eyes, the cross-sectional design precludes definitive conclusions about causality. Longitudinal multicenter studies are needed to determine whether this association reflects a therapeutic effect of RLRL. This investigation pioneers the integration of adaptive optics (AO) imaging into RLRL safety protocols, offering unprecedented cellular-level resolution for evaluating retinal integrity. It should be noted that our observations represent structural changes in cone photoreceptor arrangement; functional correlates—such as those assessed by microperimetry or electroretinography—were not evaluated in this study and remain to be investigated in future work. Future research should expand cohorts to include diverse populations and standardize imaging protocols to minimize confounders. Parallel functional assessments, such as microperimetry or ERG, could further clarify whether structural irregularities correlate with diminished photoreceptor sensitivity.