FormalPara Key Summary Points

Why carry out this study?

Myopia progression is characterized by axial elongation that is considered irreversible.

No human studies to date have suggested any treatment for myopia control could result in axial shortening within a timeframe of months or years.

What was learned from the study?

In this post hoc analysis of a randomized controlled trial, repeated low-level red-light (RLRL) therapy was found to cause substantial reversal in elongation of axial length among nearly one-quarter of the participants.

Introduction

Myopia is considered as a progressive, irreversible condition characterized by an excessive elongation of axial length (AL) [1]. Even in the fifth decades of life, AL elongation persists among some myopic adults [2, 3]. AL elongation is accompanied by a dramatically increased risk of vision-threatening complications such as macular hemorrhage, retinal detachment, and glaucoma [4]. Considering these adversities, preventing AL elongation is a primary focus for myopia control [5]. Current treatments for controlling myopia aim to slow AL elongation, but no treatments can reverse or cure AL elongation within a timeframe of months or years.

Repeated low-level red-light (RLRL) therapy is a novel therapeutic method for controlling myopia, with promising efficacy and safety. It was developed from the principle that adequate exposure to outdoor bright light is a strong protective factor against myopia development [6,7,8]. Our recent randomized controlled trial (RCT) demonstrated exposure to RLRL twice daily reduced myopic progression by 69.4% over a 1-year period [9]. In particular, when compliance of RLRL therapy increased from < 50% to 75%, the efficacy of reduction in AL elongation rose from 44.6% to 76.8%, and efficacy for spherical equivalent changes improved from 41.7% to 87.7% [9].

During the clinical trial, we observed clinically significant AL shortening among participants, which was greater than the level (> 0.05 mm) that could be attributed to measurement error. These findings challenge the view that myopia is an irreversible condition and provide new insights that may underlie the mechanisms of myopia development. In this post hoc analysis, we systematically report the frequency of AL shortening among treatment and control arms, and evaluate changes of AL and choroidal thickness (ChT) among those with significant AL shortening.

Method

Study Design and Participants

The current study is a post hoc analysis of a prospective 1-year RCT of children aged 8–13 years with myopia. Previous reports describe the methodology of the study in greater detail [9]. In brief, a parallel-group, single-masked, randomized controlled study was conducted across five study centers from four tertiary hospitals in China. The study aimed to determine the efficacy and safety of RLRL therapy for myopia control. Children aged 8–13 years with myopia of cycloplegic spherical equivalent refraction (SER) between −1.00 and −5.00 D, astigmatism 2.50 D or less, anisometropia 1.50 D or less, and best-corrected visual acuity (BCVA) 1.0 in either eye were recruited. Children were excluded if they had history or presence of strabismus, binocular visual function abnormalities, other ocular diseases, systemic diseases, or had received other myopia control treatment [including but not limited to atropine therapy, orthokeratology, or Defocus Incorporated Multiple Segments (DIMS) glasses]. A total of 400 random numbers were pregenerated, and all centers adopted the same centralized randomization list using the centralized web-based randomization service [Solomon electronic data capture (EDC) system, Vision Tech Medical Technology, Guangzhou, China]. A total of 264 eligible children were randomly allocated to an intervention group (RLRL group, n = 119) or control group (single vision spectacle, SVS group, n = 145) at the beginning of the study, with follow-up at 1, 3, 6, and 12 months.

The study was approved by the institutional ethics committee of Zhongshan Ophthalmic Centre (2019KYPH093) and adhered to the Declaration of Helsinki for study of human subjects. All participants and their guardians were informed of the study details and signed informed consent. The RCT protocol was registered at ClinicalTrials.gov (NCT04073238).

Intervention

As a routine treatment for myopia, all children including children in RLRL group wore SVS lenses throughout the study and updated their glasses if necessary. In addition to wearing SVS, children in the intervention group received RLRL therapy. The RLRL therapy was provided by a semi-conductor laser diode product (Eyerising, Suzhou Xuanjia Optoelectronics Technology, Jiangsu, China), which has been sold and widely used in China for amblyopia treatment for a decade. The device includes semiconductor laser diodes (LDs) and emits low-level red-light at a wavelength of 650 ± 10 nm. The light power going through a 4 mm pupil is 0.29 mW, which is considered as class 1 under the IEC 60825-1:2014 standard, and safe for direct ocular exposure, indicating there is no retinal thermal hazard. Children in the intervention group logged into the device with their unique accounts and passwords, and initiated the RLRL therapy twice a day, 3 min per session, with a minimum interval of 4 h. Data on the date and time of each login was automatically recorded in the device and transferred to an online database that monitored treatment compliance. The compliance of RLRL treatment was calculated as the percentage of actual number of treatments divided by the number of assigned treatment sessions (two times per day, 5 days per week) for the entire treatment period.

Due to the nature of the intervention in this study, the children and their parents were aware of the allocation. All outcome assessors were masked, including technicians, optometrists, and statisticians.

Measurements

The examination methods of this study are described elsewhere [9]. All variables for each participant were measured by the same device at baseline and follow-up. All participants received slit-lamp examination and were tested for uncorrected visual acuity (UCVA), best corrected visual acuity (BCVA), autorefractor, and ocular biometric parameters. The right eye was examined before the left eye. Measures of AL were conducted on each eye prior to cycloplegia using IOL-Master (Carl Zeiss 500, Meditec, Oberkochen, Germany). At least five measurements were performed with excellent precision (i.e., ≤ 0.05 mm) according to standard procedure, and the average value was used for analysis. Cycloplegia was achieved using two drops of 0.5% tropicamide 5 min apart for three times prior to autorefraction. Pupil light reflex and pupil diameter were checked to confirm full cycloplegia after an additional 15 min. A pupil size > 6 mm and/or an absent light reflex was considered successful dilation. Refraction was measured using Topcon KR-8800 (Topcon, Tokyo, Japan) three times with desired precision (i.e., ≤ 0.25 diopter), and the average of the three readings were used for analysis. Cycloplegic SER was calculated using the sum of the spherical power and half cylindrical power.

Choroidal thickness (ChT) was an optional measure in the RCT and was performed in centers with swept source optical coherence tomography (SS-OCT) equipment available. A total of 162 children in two study centers (Ouzhuang, Zhongshan Ophthalmic Center and Zhujiang New Town, Zhongshan Ophthalmic Center) underwent SS-OCT scanning (RLRL group, n = 72; SVS group, n = 90). After pupil dilatation, the children underwent SS-OCT imaging under standardized mesopic light conditions by the same technician masked to the study protocol, using the SS-OCT device (DRI-OCT Triton, Topcon, Tokyo, Japan) for choroidal imaging. This device uses a laser wavelength of 1050 nm, a scanning speed of 100,000 scans per second, and an axial and transverse resolution of 8 μm and 20 μm, respectively [10]. This study adopted 12.0 mm radial scans protocol (resolution, 1024 × 12) centered on the fovea. Eligible images had > 70 image quality scores (range 0–100) with no segmentation errors or artifacts. All eligible images were then verified by a masked technician. Data from built-in software that automatically segmented the choroidal layer and measured ChT parameters were used. Subfoveal ChT was defined as the distance between outer choroid-scleral margin and retinal pigment epithelium (RPE)–Bruch’s complex under the foveal region.

Outcomes

The primary outcome for this analysis was the frequency of AL shortening, which was categorized into cutoffs of > 0.05 mm, > 0.10 mm, and > 0.20 mm. Secondary outcomes included the change in AL and subfoveal ChT per year among those with AL shortening. Baseline age was categorized into 8–10- and 11–13-year groups. Baseline SER was categorized into mild myopia (−3.0 < SER ≤ −1.0 D) and moderate myopia (−5.0 < SER ≤ −3.0 D).

Statistical Analysis

Data analysis was performed in accordance with the principle of intention-to-treat analysis (ITT) and included subjects who completed randomization. No imputation of missing data was performed. Given the strong correlation between right and left eyes, only data from right eyes were used for analysis. Continuous variables were presented as mean ± standard deviation (SD). Categorical variables were presented as number and percentage. Scatter plots with linear fit described the distribution of AL change per year between treatment groups, and were stratified for age, sex, and baseline SER. The Fisher’s exact test compared group differences in the frequency of AL shortening under the three cutoffs, and were stratified by age, sex, and baseline SER. Logistic regression models investigated the relationship between frequency of AL shortening in the three cutoffs and age, sex, AL at baseline, and SER at baseline. For the subset of children who underwent SS-OCT imaging, logistic regression models investigated the relationship between magnitude of AL shortening and baseline, and changes in ChT. The R2 of the regression quantified the variance in AL shortening, which were explained by changes in subfoveal ChT. We used Stata (StataCorp, Stata Statistical Software: Release 17; College Station, TX, USA) for statistical analyses. A p-value < 0.05 was considered statistically significant.

Results

All 264 randomized children (119 in RLRL group, 145 in SVS group) were included in the present analysis, with 57 males (47.9%) in the RLRL group and 73 males (50.3%) in the SVS group.

Frequency of Axial Length (AL) Shortening

Table 1 presents the frequency of AL shortening at each visit based on cutoffs of > 0.05 mm, > 0.10 mm, and > 0.20 mm. For > 0.05 mm cutoffs, 40.34% children achieved AL shortening at 1 month, followed by 28.57% at 3 months, 21.85% at 6 months, and 12 months in the RLRL group. AL shortening in the SVS cohort was 4.14% at 1 month, 0.69% at 3 months, 0% at 6 months, and 1.38% at 12 months. There were 26/119 (21.85%) in the RLRL group versus 2/145 (1.38%) in the SVS group who had observed AL shortening > 0.05 mm (p < 0.001) at the 12-month follow-up. For cutoffs of > 0.10 mm, AL shortening in RLRL group was recorded in 12.61% of participants at 1 month, 15.17% at 3 months, 10.08% at 6 months, and 15.13% at 12 months, while the AL shortening did not present in SVS group in any visits, except at 1 month (0.69%). For cutoffs of > 0.20 mm, AL shortening was only observed in the RLRL therapy arm but not in the control group, and the AL shortening in participants at 12 months was 7/119 (5.88%) (p < 0.001).

Table 1 Frequency of AL shortening in each follow-up visit of the RLRL group and SVS group by different criteria
Full size table

Figure 1 shows the associations of AL shortening (> 0.05 mm) at 12-months follow-up with age, AL, or SER at baseline in each group. Frequency of AL shortening was significantly associated with baseline age, AL, and SER (Fig. 1 and Table 2). Myopic children in the RLRL group had the highest frequency of AL shortening at 1 month follow-up, with new recorded AL shortening not as frequent at other follow-up time points. Boys had higher incidence of AL shortening at 3, 6, and 12 months compared with girls, although these sex differences were not statistically significant (Table 2). Sensitivity analyses at cutoffs > 0.10 mm (Supplementary Table S1) and > 0.20 mm (Supplementary Table S2) showed trends consistent with these sex differences.

Fig. 1
figure 1

Scatter plots with line of best fit showing the associations between rate of AL change during follow-up with age, AL, and SER at baseline in each group. A Rate of AL change versus age at baseline; B rate of AL change versus AL at baseline; C rate of AL change versus SER at baseline. RLRL repeated low-level red-light, SVS single vision spectacle, AL axial length, SER spherical equivalent refraction, D diopter

Full size image
Table 2 Frequency of AL shortening (> 0.05 mm) in each follow-up visit stratified by age, sex, and SER at baseline in RLRL and SVS groups
Full size table

Changes of Axial Length (AL) in Children with AL Shortening

For all participants, the RLRL group had consistent and significant AL progression from baseline to each visit compared with the SVS group (Supplementary Table S3). For children with AL shortening at each follow-up visit, the changes in AL of between intervention groups are summarized in Table 3. Over 12 months, mean changes in AL among children with AL shortening (SD) were −0.156 (0.086) mm in the RLRL group, while the two children in the control group with AL shortening had a mean of −0.06 mm. This was likely due to measurement error. Changes in AL among children with AL shortening was most pronounced in the first month of treatment (−0.095 mm) and subsequent reductions were less in magnitude but constant at 3, 6, and 12 months follow-up (Table 3). AL across follow-up visits stratified by the presence of AL shortening (> 0.05 mm and > 0.10 mm) at 1 month follow-up for RLRL participants is shown in Fig. 2. For those with AL shortening at 1 month, the AL remained stable at subsequent follow-up visits. As expected, the AL change was consistent and significant related to SER change [β =  −0.0869 (95% CI −0.131, −0.0431), p < 0.0001) (Supplementary Fig. S1).

Table 3 Magnitude of AL reduction for subjects with AL shortening (> 0.05 mm) after RLRL treatment
Full size table
Fig. 2
figure 2

Trajectory of AL at each follow up visit for each child in RLRL group stratified by early responses to treatment. A AL shortening > 0.05 mm from baseline to 1 month. B AL shortening > 0.10 mm from baseline to 1 month. Each single line represents the trajectory of AL change in a year for each child in the RLRL group. The red line and blue line represent the children who had AL shortening or not at 1 month. To account for the measurement error (0.05 mm), AL shortening was defined as AL reduced by at least 0.05 mm, and a reference line of y =  −0.05 mm is presented. RLRL repeated low-level red-light, SVS single vision spectacle, AL axial length. Red represents the children that achieved AL shortening at 1 month. Blue represents the children that did not have AL shortening at 1 month

Full size image

Factors Associated with the Presence of Axial Length (AL) Shortening

Table 4 summarizes the baseline factors associated with frequency of AL shortening in multivariable models. After adjusting for other confounding factors, older age and female gender were significantly associated with a greater odds of AL shortening > 0.05 mm/year in model 1. For multivariate model 2, older baseline age was significantly associated with the presence of AL shortening after adjusting for other factors. Sensitivity analyses showed that older baseline age and lower SER were significantly associated with higher odds of AL shortening of > 0.10 mm/year (Supplementary Table S4). Table 5 summarizes the characteristics of the children with > 0.20 mm AL shortening across the 12-month period.

Table 4 Multivariable logistic regression models of baseline associators with presence of AL shortening (> 0.05 mm) after RLRL treatment
Full size table
Table 5 Incident cases of AL shortening > 0.20 mm from baseline to 12 months after RLRL treatment
Full size table

Choroidal Thickness in Relation to Axial Length (AL) Shortening

The subgroup of children with OCT data did not differ from those without OCT data with respect to age, sex, AL, and AL change (Supplementary Table S5). A total of 72 patients in the RLRL group underwent OCT measurement. Participants were included in the analysis if they had valid ChT at least at baseline and the 12-month visit (56 in the RLRL group). There were no significant differences in characteristics between included and excluded participants (Supplementary Table S6–7).

The change in AL and subfoveal ChT from 12-months follow-up to baseline were −0.20 mm and 0.056 mm of the right eye, and −0.23 mm and 0.083 mm of the left eye among children with AL shortening in the RLRL, respectively. After adjusting for other covariates, the presence of AL shortening was significantly associated with older age and greater ChT thickening (Supplementary Table S8). Figure 3 shows the associations of changes in AL with baseline parameters, and changes in ChT during the 12-month period among children with AL shortening in the RLRL group. Changes in AL of children with AL shortening was observed with increasing baseline ChT and changes in ChT at follow up. The ChT at baseline and ChT changes could only explain 5.4% and 28.3% of the AL changes, respectively.

Fig. 3
figure 3

Scatter plots showing the associations between rate of AL change with ChT at baseline and ChT changes during the 12-month period in each group. A total of 72 patients in the RLRL group and 90 patients in the SVS group underwent OCT measurement at baseline. Participants were included in the analysis if they provided valid choroidal thickness and had complete covariate data (56 in the RLRL group and 51 in the SVS group). RLRL repeated low-level red-light, SVS single vision spectacle, SS-OCT swept-source optical coherence tomography, ChT choroidal thickness

Full size image

Sensitivity Analysis

Sensitivity analysis was performed by using the 1-month AL measurements as baseline. The results of the sensitivity analysis did not change the conclusions from the primary analysis (Table S9). Over 1/10 of children receiving RLRL therapy experienced AL shortening but this phenomenon was scare in the SVS control group at each visit (p < 0.05), indicating the consistency and robustness of the primary analysis.

Discussion

This study systematically reported AL shortening over the course of months among the children on myopia control treatments. AL shortening > 0.05 mm occurred in a quarter of children using RLRL treatment across the 12-month duration of this study, and this was extremely rare among children in the control arm. AL shortening was also more common among older children. OCT examination confirmed that choroidal thickening only explained 28.3% of the AL shortening, indicating that scleral remodeling may involve the observed AL shortening as response to RLRL.

Axial Length (AL) Shortening in Previous Studies

Clinical reports specifically looking into AL shortening have been extremely rare to date. One study reported AL shortening in the children following orthokeratology treatment, but this was transient and only observed in the first week of treatment at a negligible level (average, 0.026 mm) [11]. In the present analysis, RLRL therapy achieved AL shortening of > 0.05 mm in 21.85% of children at the 12-month visit. A cutoff of 0.05 mm was selected because this was considered the range of measurement error of the IOL-Master.

Scleral Remodeling or Choroidal Thickening?

In the AL measurement, the optical biometric device (IOL-Master) measures the distance from the anterior surface of cornea to the retinal pigment epithelium (RPE). Therefore, the “observed” AL shortening could be due to choroidal thickening that would push the RPE forward. In our analysis, we confirmed significant choroidal thickening after RLRL treatment, but this could only explain 28.3% of AL shortening. This indicates the observed AL shortening might be related to true scleral remodeling or shortening, rather than an apparent shortening due to choroidal thickening.

Factors Associated with Axial Length (AL) Shortening

Age was strongly correlated with AL shortening from RLRL therapy and was more common in older children, indicating that the treatment response of RLRL therapy is weaker among young children. This phenomenon was also observed in other studies utilizing other myopia treatments, including the low-concentration atropine for myopia progression (LAMP) and Defocus Incorporated Multiple Segments (DIMS) studies [12, 13]. In general, young children tend to progress more rapidly than older children or adults, which often means children who are diagnosed earlier in childhood are likely to have worse visual prognosis. Approximately 80% of the DIMS lens wearers with rapid myopia progression were young children aged 8–9 years compared with 20% in children aged 10–13 years [13]. In children using 0.05%, 0.025%, and 0.01% atropine drops, for every 1 year younger, the AL elongation accelerated by −0.10, −0.11, and −0.12 mm/year, respectively [12]. Therefore, the effect of age on response to RLRL might be due to the different retinal profile or intuitive greater myopia progression in younger children. These treatments are likely fighting an uphill battle against factors that threaten concomitant AL elongation, so in future it would be of interest to determine if a bolus dose of RLRL is likely to curtail myopia progression further in children and change their course of disease.

Extent of Axial Length (AL) Shortening

The greatest AL shortening was observed in the 1-month visit, from there on, the extent of AL shortening gradually reduced and maintained at the same level until the end of the study, albeit in a reduced manner. From the inception of RLRL therapy to the 12-month follow-up visit, AL shortening ranged from −0.05 to −0.31 mm, with a mean (SD) of −0.156 ± 0.086 mm at the 12-month visit among children with AL shortening in the intervention group. These data suggest that AL shortening is more likely to be maintained at a certain level (−0.156 ± 0.086 mm) rather than to further increase over time. This implies that this AL shortening will unlikely fully recover a myopic eye to an emmetropia; however, longer follow-up durations are paramount to exploring the long-term response to RLRL therapy.

Insights into Mechanisms Underlying Eye Growth Regulation

It is unknown how RLRL therapy shortens AL, although animal models have observed ocular shape in infant rhesus monkeys recover, to some extent, from experimentally induced myopia once the myopic stimulus is removed [14]. After removal, the form deprivation or hyperopic defocus self-corrects and AL shortening can be achieved in chicks, guinea pigs, tree shrews, and nonhuman primates [14,15,16,17,18]. It was proposed that an altered scleral and choroidal mRNA expression was a possible mechanism underlying scleral remodeling, which led the shortening observed among these animals [19,20,21,22]. The nonhuman primate studies in response to imposed optical defocus suggested that the altered expressions of human myopia quantitative trait loci improved choroidal perfusion and choroidal vascular permeability, and dilated nonvascular smooth muscle, and may contribute to choroidal thickening [23,24,25,26,27]. In the chicken response during recovery from induced myopia, increased expression of the hyaluronan synthase-2 gene and avian thymic hormone were noted [28, 29]. In addition, choroidal changes coincide with a reduced synthesis of scleral glycosaminoglycan, and blocked human scleral fibroblast cell attachment to type I collagen may contribute to AL changes [30,31,32]. This leads to the idea that refractive development is influenced by both visual and nonvisual cues, which presumably interact with physiological mechanisms that alter eye length. By this principle, RLRL is a visual cue that alters the natural course of myopia. It is not clear whether red light interferes with any of these previously mentioned processes, but laboratory studies should seek to identify possible molecular sites of action.

Clinical Significance of Axial Shortening

The observed AL shortening among a quarter of children after RLRL treatment, at a mean of 0.156 mm, is clinically important for the future of myopia treatment. Myopia is expected to reach a population burden of 50% globally in the next 30 years, so the achievement of partial reversal in this study is a breakthrough that brings AL shortening treatments into the realm of possibility. In future, RLRL may reverse myopia in its early phases so the newly onset myopic children can be reverted to emmetropia, which would mean children might not ever require wearing glasses or specialist interventions. This study also observed higher magnitudes of AL shortening in moderate myopia children compared with mild myopia, which is promising because each diopter of reduction would reduce myopia-induced visual impairment by 40% [33]. Currently, existing RCT data on low-dose atropine drops and optical treatments (orthokeratology, bifocal spectacles, and multifocal spectacles) for myopia control are designed to delay AL progression rather than reverse or cure. This is the first treatment design to have a potentially curative future focus, and has unlocked potential to influence future clinical trials and laboratory studies.

Strengths and Limitations

The strengths of this study include its multicentered, randomized controlled design with 12-months follow-up, and standardized measurements at each study site. The data from the SVS group confirms that AL shortening is unlikely to occur among myopic children without treatment, leaving RLRL as the stimulus for the observed changes in AL. In addition, the measurements of both AL and ChT excludes choroidal thickening as the sole contributor to AL shortening, and invites new speculation for how the structure and shape of the eye may contribute to this phenomenon. Notably, SS-OCT imaging was following complete pupil mydriasis, which enables the obtainment of high-quality choroidal images and automatic ChT measurement.

Despite these breakthroughs, some limitations should also be acknowledged. First, all analyses were conducted in an RCT population with strict inclusion and exclusion criteria, which may not be representative of the general myopic population. Second, only 1-year AL shortening outcomes were reported for RLRL treatment compared with SVS. The frequency and extent of long-term AL shortening, including how AL changes after RLRL treatment needs further evaluation in future. Third, participants were limited to Chinese children, and studies with multiethnic populations are needed. In addition, understanding the long-term prospects for AL shortening over time and its contribution to myopia control will require a longer follow-up study to confirm. Fourth, given the fact that the right and left eyes were highly correlated and symmetrical (Supplementary Table S10), only data of the right eye were analyzed. The effect of RLRL on children with anisometropia needed further study. Fifth, OCT data were only available in a small subset of subjects. Further SS-OCT studies of large sample sizes are needed to clarify the correlations between AL, SER, and ChT, and their trends overtime. Sixth, the information of age of onset, progression rate before enrollment, and environmental factors such as near work and time outdoors were not collected in this study, which limited our ability to determine the potential impact of these factors on responses to the RLRL therapy. Future studies with sensors are needed to quantified theses myopia-genic factors. Lastly, biometry was performed without cycloplegia, which might be influenced by the accommodation, but it would not significantly change the conclusions.

Conclusions

Nearly a quarter of children presented with AL shortening following RLRL therapy, an effect that could not be attributed to measurement error and/or ChT. Age and baseline AL were related to the degree of benefit from RLRL therapy. This novel finding highlights the disease-modifying implications of RLRL and for the first time shows myopia, to some extent, may be reversible in humans in the year timeframe. Such findings present RLRL as a therapy which actively prevents sight-threatening complications associated with AL elongation due to myopia development. Further studies are needed to confirm the effect of RLRL on AL in non-Chinese populations and elucidate mechanisms underlying AL shortening.