A natural experiment in school lighting: observational evidence linking far-red light exposure to slower myopia progression in Chinese Uyghur schoolchildren

Abstract

Background

Childhood myopia is a growing problem worldwide. In a primary school in Xinjiang, China, some classrooms had been installed with far-red/near-infrared (FR/NIR) lighting, while others kept conventional lights. This natural experiment let us test whether everyday exposure to FR/NIR light is linked to slower myopia progression in Uyghur children—a group often understudied despite high myopia rates.

Methods

In this one-year prospective observational cohort study, we invited all eligible Uyghur children in grades 2 and 3. Exposure groups were defined by pre-existing classroom lighting conditions (FR/NIR-enriched vs. Conventional-lit). Primary outcome was change in axial length (AL); secondary outcome was change in spherical equivalent refraction (SER, calculated as sphere + ½ cylinder). Propensity score matching (PSM) was employed to address confounding, with sensitivity analyses (ANCOVA and IPTW-ANCOVA) conducted to assess robustness.

Results

After PSM, 82 participants (41 per group) were included. AL increase was significantly smaller in the FR/NIR-enriched group (0.19 ± 0.09 mm) than Conventional-lit group (0.34 ± 0.12 mm), with a mean difference of –0.15 mm (95% CI: –0.20 to –0.10; P < 0.001). SER myopic shift was also attenuated in the FR/NIR-enriched group (–0.12 ± 0.46 D vs. –0.52 ± 0.59 D), yielding a mean difference of 0.40 D (95% CI: 0.16 to 0.64; P < 0.001). Subgroup analyses suggested age-related differences.

Conclusions

Over one year, children in classrooms with FR/NIR-enriched lighting showed slower eye growth and less myopic shift. With careful statistical adjustment, these findings suggest that changing classroom lighting could be a simple, passive way to help control myopia at the population level.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-27883-3.

Introduction

Childhood myopia is a growing global public health challenge [1, 2]. According to the World Health Organization World report on vision [3], at least 2.2 billion people globally have a vision impairment, among whom at least 1 billion have a condition that could have been prevented or has yet to be addressed. Within this population, uncorrected refractive error is a leading cause [4]. In China, a 2020 national survey revealed a myopia prevalence of 52.7% among children and adolescents aged 6–18 years, based on a nationwide school-based screening of 8,604 schools and 2.477 million students using non-cycloplegic autorefraction [5]. Early-onset myopia strongly predicts high myopia and sight-threatening complications including cataract, glaucoma, and macular degeneration [6–9].

The pathophysiology involves complex gene-environment interactions. Extended near-work under inadequate illumination is a well-established risk factor [10]. These behaviors can trigger accommodative lag [11] and reduce choroidal thickness [12], that promote scleral remodeling and axial elongation [13, 14]. While increased time outdoors demonstrates a clear dose–response relationship for myopia risk reduction in children [15], ensuring adequate outdoor exposure is challenging in academically intensive systems globally, and it primarily delays onset rather than slowing progression once established [16].

Current mainstay interventions, including low-concentration atropine and orthokeratology lenses, face significant constraints [17–21]. Pharmacological options such as atropine are limited by dose-dependent side effects such as photophobia and accommodative insufficiency, with higher concentrations offering greater effect size but bringing with more adverse effects [17–19]. Orthokeratology, while effective, carries risks of microbial keratitis and other corneal complications, necessitating rigorous hygiene and monitoring [20]. Other optical interventions, such as multifocal soft contact lenses or spectacles, show promise but are challenged by variable effect size and compliance issues [21]. This landscape highlights a critical gap: the need for scalable, equitable, and low-burden population-level strategies that can be seamlessly integrated into children's daily environments.

Repeated low-level red-light (RLRL) therapy has emerged as a promising intervention for myopia control, with clinical trials demonstrating efficacy in reducing axial elongation and myopic shift. A 6-month randomized controlled trial comparing RLRL with 0.01% atropine in 91 children aged 6–12 years reported more favorable outcomes with RLRL in terms of axial elongation (− 0.09 mm vs. + 0.13 mm) and myopic shift (+ 0.25 D vs. − 0.25 D) [22, 23].

In contrast to individualized therapies, modifying the school built environment offers a complementary public health strategy. Such interventions have been associated with various child health outcomes [24], but their role in myopia control remains inconclusive with limited evidence [25]. While active red-light therapies have shown promise as individual treatments [26], whether routine passive exposure to far-red/near-infrared light in classroom settings could influence myopia progression has not yet been addressed.

A unique opportunity arose at a primary school in Xinjiang, China, where some classrooms had been independently upgraded with far-red/near-infrared (FR/NIR) lighting while others retained conventional fixtures. The predominantly Uyghur student population—with high myopia burden yet underrepresented—makes this setting valuable for generating equitable evidence.

We designed a one-year prospective cohort study to assess whether FR/NIR-enriched classroom lighting is associated with slower myopia progression.

Study design and setting

This was a 1-year, prospective, assessor-blinded, school-based non-interventional comparative cohort study conducted from September 2023 to September 2024 at the First Primary School of Pishan County, Xinjiang, China. The study adopted a pragmatic, real-world approach by capitalizing on a pre-existing variation in classroom lighting infrastructure within the school. No intervention was administered; children were observed under their naturally assigned classroom lighting conditions. The class served as the unit of environmental exposure. All examiners performing axial length measurements, cycloplegic autorefraction, and other ocular biometric assessments were masked to the classroom lighting allocation. To maintain blinding given the classroom-level exposure, participants were instructed not to discuss their classroom lighting with the examiners.

The study protocol was approved by the Myopia Prevention Ethics Committee of Pishan County (Approval No. PSKY2023001). All procedures strictly adhered to the tenets of the Declaration of Helsinki.

Study participants

Participants were children enrolled in grades 2 and 3 at the school. Eligibility was assessed through voluntary registration. Inclusion criteria were: age 6 to 12 years; spherical equivalent refraction (SER, calculated as sphere + ½ cylinder) between –6.00 D and + 5.00 D in both eyes; astigmatism ≤ 2.25 D; and best-corrected visual acuity (BCVA) of 0.00 logMAR or better in each eye. Exclusion criteria included: history of myopia control treatment; irregular astigmatism, strabismus, or other active ocular diseases; previous ocular surgery; or systemic diseases potentially affecting visual outcomes (e.g., diabetes, rheumatoid arthritis). All enrolled children were of Uyghur ethnicity, reflecting the local demographic. The final analysis included 82 children (mean age 8.99 ± 0.57 years, range 7.2–10.0 years; 46.3% boys, 53.7% girls). For all analyses, data from the right eye were used unless otherwise specified. This approach was based on the high interocular correlations observed at baseline for both axial length (r = 0.97, P < 0.001) and spherical equivalent refraction (r = 0.76, P < 0.001). Selecting a single eye is a common practice in ophthalmic research [27]and avoids statistical bias arising from interocular correlation, which can inflate type I error rates and produce falsely narrow confidence intervals. [28].

Written informed consent was obtained from the parents or legal guardians of all participants. Additionally, age-appropriate child assent was obtained from each child participant.

Study environments and exposure

Lighting infrastructure

Prior to the study, as part of a separate school environment upgrade initiative, a subset of classrooms received modified lighting fixtures, while others retained the original conventional lighting. As shown in Figure S1, each classroom was equipped with nine ceiling-mounted lamps and three dedicated blackboard lamps, with a total power of 324 W (36 W per lamp). The power density was 5.8 W/m². All classrooms in both groups were located on the same floor, shared identical building orientations, and had the same architectural features including number, size, and glazing of windows. The curtains remained open during school hours throughout the study period. Each classroom had an area of approximately 55 m². The school operates a standard seat rotation policy whereby students rotate positions every two weeks (e.g., from window-side to wall-side), ensuring that over the one-year observation period, each student spent time in various locations within the classroom with differing levels of natural light.

Lighting operation

The classroom lighting was switched on based on actual illumination needs during school hours, primarily on cloudy days or when natural daylight was insufficient to meet the minimum standard for desktop illuminance (≥ 300 Lux). The exposure was thus intermittent and varied daily with weather conditions.

FR/NIR-enriched classrooms

In these classrooms, the FR/NIR light sources contained two types of LEDs: one was the conventional white light LEDs which keep the same as the following Conventional-lit classrooms, and the other was the specialized FR/NIR LEDs. The specialized FR/NIR LEDs were fabricated using the home-made far-red phosphors. As illustrated in Fig. 1a, a blue LED chip (emission peak at 450 nm, wavelength region 400–500 nm, full width at half maximum [FWHM] ~ 21 nm) served as the excitation source. The blue-emitting light was completely absorbed and down-converted by a dedicated far-red phosphor layer coated on the chip, thereby emitting target far-red light with an emission peak at 710 ± 20 nm and FWHM ≥ 60 nm [29]. These FR/NIR LED devices are of the same design as the light sources that we have previously used to approve the biological effects in experimental animal studies [29]. It should be noted that the FR/NIR lighting used in this study was modified from LED fixtures rather than using lasers, and serves as indirect ambient light rather than direct ocular exposure. Therefore, this exposure does not constitute high-intensity red light therapy

Fig. 1.

Schematic diagram of the fabrication and physical photo of the FR/NIR devices. a schematic diagram of the fabricating FR/NIR device by coating the phosphors on the 450 nm blue-emitting chip contained in a 2835 bracket to produce FR/NIR emission, b physical photo of the FR/NIR devices fixed in the braided strap, (c, d) the amplified front-view and backward of a single LED, and (e) the phosphors used to fabricate the FR/NIR LED devices

[30].

Conventional-lit classrooms

These classrooms retained the original conventional white LED lighting fixtures, with the same configuration of nine ceiling-mounted lamps and three blackboard lamps. All other school activities, schedules, and curricula remained identical to those in the FR/NIR-enriched classrooms.

Outcome measures

Primary outcome

The change in axial length (AL) over one academic year. AL was measured using a Suowei optical biometer (SW-9000), recording the average of three measurements.

Secondary outcome

The change in SER over one academic year. Cycloplegic autorefraction was performed using a Topcon autorefractor following a validated protocol: five cycles of 0.5% compound tropicamide were administered at five-minute intervals, with autorefraction conducted 30 min after the final instillation, which has been shown to induce significant changes in lens thickness and axial length in children (both P < 0.05) [31, 32]. Cycloplegia was confirmed by a pupil diameter ≥ 6.0 mm and absence of light reflex.

Other measures

Changes in other ocular biometric parameters were also assessed, including central corneal thickness (CCT), anterior chamber depth (ACD), lens thickness (LT), and vitreous chamber depth (VCD). Prespecified subgroup analyses were planned to compare outcome differences by grade (2 vs. 3) and age (7–8 vs. 9–10 years).

Baseline data

Demographic information (age, sex, grade, class) was recorded for each participant during the baseline assessment. Initial ocular biometric parameters and refractive status were recorded for all participants. To contextualize the study within a real-world school setting, informal discussions were held with school staff regarding the installation, routine operation, and any general observations related to the different lighting systems.

Statistical analysis

Sample size

Sample size estimation was based on comparing two groups. Based on published normative data for axial elongation in Asian children [33], the annual elongation rates for children aged 9–10 years ranged from 0.21 to 0.45 mm depending on refractive status, with a mean of approximately 0.34 mm/year. According to the empirical rule for normal distributions, approximately 68% of observations fall within mean ± 1 SD. Assuming that the reported range approximates this interval, the standard deviation was estimated as (0.45—0.21)/2 = 0.12 mm. With 90% power and a two-sided alpha of 0.05, accounting for the class-based exposure and potential attrition, we planned to observe no fewer than 41 students per group.

Addressing baseline differences and assessing robustness

Participants were not randomly assigned to lighting environments, and they were naturally allocated based on their classroom assignment. To minimize selection bias and enhance the comparability of these naturally formed groups for analysis, we performed 1:1 propensity score matching (PSM) using R software (version 4.3.1). The propensity score was derived from a logistic regression model that included the following baseline covariates: age, sex, grade, spherical equivalent refraction, axial length, central corneal thickness, anterior chamber depth, lens thickness, vitreous thickness, pupil diameter, and corneal astigmatism. Matching was performed without replacement using a caliper width of 0.2 standard deviations of the logit propensity score. All primary analyses were conducted on this matched cohort. [34].

Balance assessment

Balance between the two exposure groups before and after PSM was assessed using standardized mean differences (SMD) in R, with an absolute SMD < 0.1 indicating good balance. A Love plot was generated to visualize covariate balance (Figure S2).

Primary and sensitivity analyses

The primary analysis compared the changes in axial length and spherical equivalent refraction between the two exposure groups in the matched sample using independent-samples t-tests. To assess the robustness of our findings, we conducted two pre-specified sensitivity analyses: Analysis of Covariance (ANCOVA): Adjusted for baseline measurements of the respective outcome (AL or SER). Inverse Probability of Treatment Weighting (IPTW) with ANCOVA (IPTW-ANCOVA): Used the inverse probability weights derived from the propensity score model in R to create a weighted cohort, followed by ANCOVA adjusting for baseline values. Effect estimates, 95% confidence intervals (CI), and P-values from both sensitivity analyses are reported. Consistency across the primary matched analysis and the two sensitivity analyses were interpreted as evidence of robustness. (Table S1).

Subgroup and other analyses

Prespecified subgroup analyses were conducted by age and grade. For each subgroup, an independent-samples t-test was performed to calculate the between-group difference in outcome change. The difference in the magnitude of the between-group difference across subgroups was assessed by comparing the point estimates and their 95% confidence intervals. These subgroup analyses were exploratory, and no adjustment for multiple comparisons were made.

Results

Study population and baseline characteristics

A total of 100 children were initially enrolled across the different classroom environments. Figure 2 presents the study flow diagram. After applying eligibility criteria and performing PSM to enhance the comparability of the two groups, 82 participants (41 exposed to the FR/NIR-enriched classrooms and 41 exposed to the Conventional-lit classroom) formed the well-matched analytical cohort. All enrolled children were of Uyghur ethnicity, reflecting the local demographic. (Table 1).

Fig. 2.

Study Flow Diagram

Table 1.

Baseline characteristics and balance assessment before and after PSM

Characteristic	Conventional-lit Group	FR/NIR-enriched Group	SMD	Conventional-lit Group	FR/NIR-enriched Group	SMD	Balance
	Before Weighting			After Weighting
Demographic
Sample size	41	41	-	-	-	-	-
Effective sample size (weighted)	41.3	40.7	-	-	-	-	-
Age, years	8.94 ± 0.62	9.04 ± 0.51	−0.183	9.02 ± 0.61	9.04 ± 0.51	−0.028	Acceptable → Good
Female sex, n (%)	24 (58.50)	20 (48.80)	0.196	21.4 (52.30)	21.2 (51.80)	0.010	Acceptable → Good
Grade, n	2: 21, 3: 20	2: 21, 3: 20	0.000	2: 21, 3: 20	2: 21, 3: 20	0.000	Good → Good
Ocular Parameters
LogMAR visual acuity	−0.03 ± 0.12	−0.03 ± 0.14	−0.038	−0.03 ± 0.11	−0.02 ± 0.14	−0.040	Good → Good
Spherical equivalent, D	−0.15 ± 0.87	0.04 ± 1.31	−0.176	−0.10 ± 0.81	−0.13 ± 1.40	0.026	Acceptable → Good
Axial length, mm	23.09 ± 0.78	23.03 ± 0.87	0.069	23.08 ± 0.78	23.08 ± 0.95	−0.007	Good → Good
Central corneal thickness, μm	527.73 ± 30.68	531.56 ± 34.95	−0.116	529.35 ± 30.01	528.91 ± 35.68	0.013	Acceptable → Good
Anterior chamber depth, mm	2.90 ± 0.23	2.90 ± 0.23	−0.007	2.90 ± 0.22	2.90 ± 0.24	−0.022	Good → Good
Lens thickness, mm	3.57 ± 0.23	3.53 ± 0.14	0.227	3.54 ± 0.21	3.54 ± 0.14	0.010	Poor → Good
Vitreous chamber depth, mm	16.08 ± 0.74	16.07 ± 0.80	0.020	16.11 ± 0.73	16.11 ± 0.86	−0.004	Good → Good
Astigmatism, D	1.08 ± 0.59	1.05 ± 0.48	0.051	1.09 ± 0.59	1.09 ± 0.49	0.005	Good → Good
Pupillary distance, mm	6.49 ± 0.75	6.50 ± 1.01	−0.019	6.55 ± 0.75	6.55 ± 1.02	0.006	Good → Good

Data presentation: Continuous variables as mean ± standard deviation; categorical variables as n (%)

Balance criteria: Good (|SMD|< 0.1), Acceptable (0.1 ≤|SMD|< 0.2), Poor (|SMD|≥ 0.2)

Arrows indicate improvement in covariate balance after IPTW. All SMD values > 0.1 before weighting improved to < 0.1 after weighting

Abbreviations: SMD Standardized mean difference

As shown in Table 1, propensity score matching successfully balanced all measured baseline covariates between the two exposure groups. While some variables showed notable imbalance before matching (e.g., LT SMD =|0.227|), all absolute SMDs after matching were reduced to below 0.1, confirming excellent comparability of the matched cohorts for analysis. The Love plot (Supplementary Figure S2) visually confirms this improvement in balance.

Primary outcome: axial length

Following the 1-year observational period in the matched cohort, the increase in axial length was significantly lower in those exposed to the FR/NIR-enriched environment (0.19 ± 0.09 mm) compared with those exposed to the conventional environment (0.34 ± 0.12 mm). The between-group mean difference was −0.15 mm (95% CI: −0.20, −0.10; P < 0.001). Compared with the Conventional-lit group, the FR/NIR-enriched group exhibited a 44.12% reduction in axial elongation rate. (Fig. 3 and Table 2).

Fig. 3.

Change in axial length over 1 year. The FR/NIR-enriched group is indicated by red circles; the Conventional-lit group is indicated by blue circles. Values represent mean ± SEM. Exposure to the FR/NIR-enriched environment was associated with a markedly smaller increase in axial length compared with exposure to the conventional lighting environment

Table 2.

Primary and secondary outcome changes after one year

Outcome	FR/NIR-enriched Group (n = 41)	Conventional-lit Group (n = 41)	Mean Difference (95% CI)	P
Δ Axial Length (mm)	0.19 ± 0.09	0.34 ± 0.12	–0.15 (–0.20 to –0.10)	< 0.001
Δ Spherical Equivalent (D)	–0.12 ± 0.46	–0.52 ± 0.59	0.40 (0.16 to 0.64)	< 0.001

Data are presented as mean ± standard deviation. Δ = change from baseline to 1-year follow-up CI Confidence interval, D Diopters

Bold P values indicate statistical significance

Secondary outcome: change in spherical equivalent refraction

The change in SER followed a consistent pattern with the axial length findings. In the overall sample, SER progression was significantly slower in those exposed to the FR/NIR-enriched lighting environment (–0.12 ± 0.46 D) than in those exposed to the Conventional-lit environment (–0.52 ± 0.59 D), yielding a between-group mean difference of 0.40 D (95% CI 0.16–0.64; P < 0.001). Compared with the Conventional-lit group, the FR/NIR-enriched group demonstrated a 76.92% reduction in myopic shift rate. (Fig. 4 and Table 2).

Fig. 4.

Change in SER measured by cycloplegic autorefraction. The FR/NIR-enriched group is indicated by red circles; the Conventional-lit group is indicated by blue circles. Values represent mean ± SEM. Exposure to the FR/NIR-enriched environment was associated with significantly less myopic shift compared with exposure to the conventional environment

Sensitivity analyses

To assess the robustness of these associations against potential confounding, we conducted pre-specified sensitivity analyses using two different statistical approaches. The results from the two pre-specified sensitivity analyses (ANCOVA and IPTW-ANCOVA) were highly consistent with the primary matched analysis (Table S1). The effect estimates for axial length (–0.149 mm vs. –0.147 mm) and spherical equivalent (0.413 D vs. 0.441 D) were nearly identical and all reached high statistical significance (Table S1). This consistency across different analytical approaches strengthens the robustness of the observed associations and suggests they are not sensitive to the specific statistical method used to address confounding.

Subgroup analyses

Subgroup analyses showed that the association between FR/NIR exposure and attenuated myopia progression was consistently observed across all age and grade subgroups, with a trend toward greater differences in younger children. For axial elongation, the between-group difference was −0.19 mm (95% CI: −0.27, −0.12) in children aged 7–8 years, compared with −0.10 mm (95% CI: −0.16, −0.05) in those aged 9–10 years. For spherical equivalent refraction, the between-group difference was 0.41 D (95% CI: 0.14, 0.70) in children aged 7–8 years and 0.32 D (95% CI: 0.01, 0.64) in those aged 9–10 years. Detailed results of all subgroup analyses, including grade-stratified comparisons, are provided in the Supplementary Materials (Figure S3 and Table S2).

Other outcomes

Analysis of secondary ocular biometric parameters in the matched cohort identified a statistically significant between-group difference in the change of vitreous chamber depth (VCD), with an adjusted mean difference of 0.16 mm (95% CI 0.11–0.21; P < 0.001). No significant differences were observed in the changes of anterior chamber depth, or lens thickness. Detailed results of the other outcomes analyses are presented in Supplementary Table S3. No adverse events related to the environmental exposure were reported during the study period.

Feasibility and contextual observations

The FR/NIR lighting systems functioned normally throughout the study period as part of the standard classroom infrastructure. No technical failures or unintended interruptions of the daily environmental exposure were reported. Informal feedback from school staff indicated that the lighting was seamlessly integrated into the school day with no disruption to teaching activities. The systems required no special daily operation from teachers or students.

Discussion

In this school-based cohort study, we found that exposure to FR/NIR-enriched classroom lighting was feasible in a real-world setting and, over one academic year, was associated with a significantly slower rate of axial elongation and myopic refractive progression compared to conventional classroom lighting. The observed effect size—approximately a 44.12% reduction in axial elongation (0.15 mm) and a 76.92% reduction in myopic shift (0.40 D)—is not only statistically significant but also clinically meaningful. To put this in context, the annual axial elongation in the Conventional-lit group (0.34 mm) aligns with rates reported for untreated myopic children in East Asian populations [33]. In contrast, the FR/NIR-exposed group showed an axial elongation of only 0.19 mm per year, representing a 0.15 mm reduction compared to the control group. This effect size approaches the levels reported for established interventions such as 0.01% atropine (0.10 mm/year reduction) and 0.05% atropine (0.21 mm/year reduction) [35].

A key strength of this study lies in its pragmatic design leveraging a natural experiment in a real-world school setting. To address confounding inherent to non-random allocation, we employed propensity score matching, which effectively balanced baseline covariates between groups (Table 1, Supplementary Figure S2). The robustness of our findings was further confirmed by sensitivity analyses (ANCOVA and IPTW-ANCOVA), which yielded nearly identical effect estimates (Table S1).

Notably, the association between FR/NIR exposure and attenuated axial elongation was more pronounced in younger children (aged 7–8 years), with a between-group difference of −0.19 mm (95% CI: −0.27 to −0.12), compared to −0.10 mm (95% CI: −0.16 to −0.05) in children aged 9–10 years. This finding aligns with the developmental pattern of more rapid ocular growth in younger children [36] and suggests that environmental modifications may yield greater benefits when implemented during critical developmental windows, with important implications for public health planning.

Further analysis revealed that the reduction in axial elongation in the FR/NIR-enriched group was accompanied by a correspondingly smaller increase in vitreous chamber depth (VCD) (between-group difference: 0.16 mm; 95% CI: 0.11 to 0.21; P < 0.001), with a magnitude comparable to the total axial length difference (0.15 mm), suggesting that the effect of FR/NIR exposure may be predominantly posterior segment-related. This finding is consistent with the established understanding that vitreous chamber elongation constitutes the primary structural basis of axial myopia progression [37]. A recent study quantifying the relationship between ocular parameters and axial elongation in children reported that VCD changes account for 85.5% to 98.3% of total axial length increase [38]. The observed reduction in VCD elongation may be linked to choroidal mechanisms. Previous studies have demonstrated that repeated low-level red-light therapy increases choroidal thickness [23], and that choroidal thinning often precedes axial elongation [39]. Impaired choroidal perfusion is considered instrumental in myopia progression, as it may trigger the scleral hypoxia and subsequent remodeling that lead to axial elongation [40]. However, as the present study did not directly measure choroidal thickness, we cannot determine whether FR/NIR exposure exerted its effects through choroidal modulation. This specific alteration in ocular biometric proportions warrants further investigation using advanced imaging techniques, such as wide-field OCT, to delineate the precise anatomical site of effect.

The lighting systems operated normally throughout the study period without disrupting school routines, demonstrating that this passive environmental modification strategy has high feasibility and acceptability in real-world educational settings. Compared to individual therapies requiring daily adherence (e.g., atropine eye drops, orthokeratology), this "set-and-forget" environmental modification circumvents compliance barriers and may be particularly suitable for implementation in resource-limited regions.

Pishan County, Xinjiang, is situated on the southern edge of the Taklamakan Desert and the northern slope of the Kunlun Mountains. The region has a warm temperate continental arid climate, with high altitude and very limited annual precipitation—typical of desert-edge environments. By focusing on a Uyghur student population in this unique geographical setting—an ethnic group with high myopia burden yet underrepresented in research—our study provides important evidence for myopia prevention and control in arid desert climate regions while contributing data to support more inclusive public health strategies. Notably, FR/NIR-enriched classroom lighting may offer additional benefits for ocular surface health in arid regions beyond myopia control. Our own experimental research in a mouse model has confirmed that near-infrared light emitted by phosphor-converted LED devices—comparable to the technology used in our FR/NIR-enriched classrooms—can induce gland secretion in Sjögren's syndrome [41]. This raises the possibility that the same environmental exposure might offer additional benefits for ocular surface health, such as supporting tear film stability, although this remains to be directly investigated in future studies.

Limitations

This study has several limitations. First, children were not randomly assigned to different lighting environments. Although we used propensity score matching to adjust for measured covariates, unmeasured factors—such as outdoor time, near work, screen use, academic workload, socioeconomic status, or parental myopia—could still confound the results. The classroom-level nature of the exposure may also introduce clustering effects that we could not account for. Although examiners were masked to classroom lighting conditions and participants were instructed not to discuss their classroom lighting with the examiners, we did not formally assess masking success, and it is possible that some breaches occurred despite these precautions. Second, we could not precisely measure how much FR/NIR light each child actually received. The lights were used intermittently depending on weather and classroom brightness, and we did not have dosimeters to track individual exposure. While staff confirmed the systems were functioning as intended, the lack of objective exposure data precludes dose–response analysis. Third, we only studied Uyghur children and followed them for one year, so we do not know if the findings apply to other groups or whether the effects last longer. Fourth, we cannot assess whether any benefits persist after exposure ends, or whether rebound effects might occur. Fifth, because our sample included both non-myopic and already myopic children, we cannot separate potential effects on delaying onset from those on slowing progression. Larger studies with longer follow-up and stratification by baseline refraction are needed to address these questions.

Public health and school health implications

Despite the noted limitations, our findings provide preliminary evidence for a novel population-level approach to myopia control. Integrating FR/NIR lighting into classroom infrastructure may offer potential public health advantages, as it is a passive measure that requires no active effort from children. This approach has the potential to be equitable. However, formal cost-effectiveness analysis was beyond the scope of this preliminary observational study. Future research should prioritize long-term follow-up, replication in diverse settings, advanced imaging, and formal cost-effectiveness assessments before any firm conclusions can be drawn regarding public health applicability.

Conclusion

This prospective non-interventional cohort study indicates that exposure to a classroom environment illuminated by FR/NIR-enriched white lighting, compared to conventional white lighting, is associated with a significant attenuation of axial elongation and myopic refractive shift over one year in school-aged children. The findings, supported by rigorous propensity score matching and sensitivity analyses, show the great potential of passive environmental modification as a feasible and equitable public health strategy for myopia control.

Authors’ contributions

Chang Liu: Conceptualization, Methodology, Investigation, Data Curation, Formal Analysis, Writing – Original Draft, Project Administration; Lei Chen: Conceptualization, Resources, Methodology, Writing – Review & Editing; Lanlan Wang: Investigation, Data Curation, Validation, Writing – Review & Editing; Huijun Shao: Investigation, Resources, Visualization, Writing – Review & Editing; Yinling Sang: Investigation, Formal Analysis, Validation, Writing – Review & Editing; Yangzong Suolang: Investigation, Resources, Supervision, Writing – Review & Editing; Qiaoli Li: Investigation, Data Curation, Formal Analysis, Writing – Review & Editing; Wuqi Chen: Investigation, Data Curation, Formal Analysis, Writing – Review & Editing; Dongwei Liu: Conceptualization, Methodology, Resources, Supervision, Funding Acquisition, Writing – Review & Editing, Project Administration.

Funding

This research was supported by the Key Research and Development Technology project of Anhui Province (2022j11020013), the Natural Science Foundation of Tibet Autonomous Region (XZ202501ZR0090), the Natural Science Foundation for Group–style Medical Aid Project of Tibet Autonomous Region (XZZR202402034[W]), and the College Students' Innovation and Entrepreneurship Training Project of Anhui Province(S202410366013). The research presented in this study is funded by the National Natural Science Foundation (21875058), the Natural Science Foundation of Anhui Province (2208085J13), and the Major Science and Technology Project of Zhongshan City of Guangdong Province on the Strategic Emerging Industries Technology Research Topic (2022A1007), China.

Data availability

All data that support the findings of this study are included within the article (and any supplementary files).

Declarations

Ethics approval and consent to participate

This study was approved by the Myopia Prevention Ethics Committee of Pishan County (Approval No. PSKY2023001) and conducted in accordance with the Declaration of Helsinki. The informed consent process was tailored for children. Written informed consent was obtained from the parents or legal guardians of all participants. Additionally, age-appropriate child assent was obtained from each child participant using a simplified, child-friendly assent form and verbal explanation to ensure their understanding and voluntary participation.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.