The survey samples were obtained from the monitoring program of students’ myopia and health influencing factors conducted from September to December 2021 by the Ningxia Center for Disease Control and Prevention. Ningxia is located in the northwest of China, with 9 districts, 11 counties and 2 cities under its jurisdiction, covering a total area of 66,400 square kilometers.¹ Ningxia is also an autonomous region of China known as the Ningxia Hui Autonomous Region. We used stratified cluster random sampling to select schools for monitoring, and a total of 120 schools were selected. The detailed spatial distribution of surveyed schools in each district or county is shown in Figure 1. About two classes per school per grade were taken as the basic unit of investigation. Given responsiveness, our study only focused on in-school adolescents (fourth grade through high school). A total of 34,777 adolescents participated in the survey, with each participant receiving a physical examination and a structured questionnaire. The questionnaires were paper-based and completed by the adolescents themselves, with completion taking approximately 10 to 30 min. After excluding 1,617 (4.65%) participants with missing covariates, a total of 33,160 adolescents were included in the final analysis. The study was approved by the Ethics Committee of the People’s Hospital of Ningxia Hui Autonomous Region ([2022]-NZR-093) and was conducted in accordance with the guidelines of the Declaration of Helsinki.

In this study, myopia was diagnosed following the national standard “Technical Specifications for Student Health Examination” (GB/T 26343) and the health industry standard “Specification for Myopia Screening in Primary and Secondary School Students” (WS/T 663-2020). Briefly, ophthalmologists or trained investigators used automated refractometers to perform eye examinations without cycloplegia. All refractometers were compliant with the ISO 10342:2010 standard for ophthalmic instruments. Three measurements were taken in each eye and averaged. If the spherical equivalent (SE) of any two measurements differed by ≥0.50 diopters (D), the eye was re-measured and a new mean value was calculated. SE was calculated by the following formula: SE = sphericity degree + 0.5 × cylindrical degree. SE was used to assess refractive error in this study, consistent with previous epidemiologic surveys of children (34, 35). Based on previous literature and industry standards (WS/T 663-2020), myopia was defined as a SE of ≤−0.5 D in at least one eye (36, 37). In addition, students with suspected eye abnormalities were referred to a specialist for further evaluation.

During the survey, the residential address of each participant’s school was collected and subsequently the longitude and latitude of each participant’s school were extracted. Outdoor ALAN data were sourced from NASA/NOAA’s Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) low-light imaging data, which provides high-quality nighttime light products² (18, 38). The intensity of outdoor ALAN was quantified in raster units, ranging from 0 to 472.86 nanoWatts/cm²/sr, with a spatial resolution of 500 m. Taking Supplementary Figure S1 as an example, we presented the annual average outdoor ALAN exposure for 2020 in Ningxia province. Considering our population data collection occurred in 2021, we downloaded outdoor ALAN data for the years 2018–2020. Then, using ArcGIS version 10.2 software, we extracted the annual mean values of outdoor ALAN for each school buffer (1 km, 3 km, and 5 km) for the years 2018–2020. Considering that the majority of the included schools follow a three-year curriculum, for the main analysis, we used the 1 km buffer and 2-year average of outdoor ALAN exposure as the main exposure variable. Also, the effects of multiple definitions of outdoor ALAN were explored.

The assessment of the schools’ green space was conducted using the normalized difference vegetation index (NDVI). NDVI values range from −1 to 1 (3). Higher NDVI values indicate higher levels of green space around the school. NDVI data were sourced from Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance data.³ Considering the seasonality of vegetation, we extracted the summer three-month average (June to August) from the MODIS VI product (MOD13Q1) as the primary exposure, with a spatial resolution of 250 m. Taking Supplementary Figure S2 as an example, we also presented the annual average NDVI exposure for 2020 in Ningxia province. For the main analysis, we also used the 1 km buffer and 2-year average of NDVI exposure as the main exposure variable.

The daily data of particulate matter ≤2.5 mm (PM_2.5) at a 1 km spatial resolution were obtained from ChinaHighAirPollutants database⁴ (39). PM_2.5 was ultimately selected as the primary exposure variable given the generally high correlation of air pollution indicators. Additionally, using ArcGIS version 10.2 software, the 1 km data were resampled to 500 m for subsequent matching purposes. For the main analysis, the 1 km buffer and 2-year average of PM_2.5 exposure was used as the main exposure variable.

Based on previous studies (3, 40, 41), covariates were selected as follows: demographic characteristics [age, sex, residence, nationality, school level, exposure time, body mass index (BMI), and parental myopia]; behavioral and lifestyle factors (nighttime sleep, outdoor exercise time per day, frequency of sugar per week, frequency of oil per week, frequency of fruit per week, frequency of vegetable per week, computers and TVs spend time per day, after-school homework time per day, after-school tutoring time per day, eye exercises per day, and annual frequency of visual inspections); and regional macro-indicators [gross domestic product (GDP) per capita, population density, and health technical personnel].

Among these factors, exposure time represented time exposed to the school environment: grade 4 to 6, indicating >3 year of exposure time; grade 7 and grade 10, indicating 1 year of exposure time; grade 8 and grade 11, indicating 2 year of exposure time; grade 9 and grade 12, indicating 3 year of exposure time, respectively. BMI was calculated by dividing the weight (kg) of each participant by the square of height (m). Data on demographic and behavioral/lifestyle factors were collected using a structured questionnaire. Behavioral/lifestyle factors were primarily assessed through questions about the past week. For example, participants were asked: “How much time is spent on outdoor exercise per day over the past week?” with response options including: less than 1 h, 1–2 h, and 2 h or more. Data on regional macro-indicators were obtained from Ningxia Statistical Yearbook of 2020.⁵

Continuous variables that were normal or approximately normal were described by the mean [standard deviation (SD)], while those that were not normally distributed were expressed as the median [interquartile spacing (IQR)]. Descriptive statistics for categorical variables were expressed as numbers (percentages, %). Depending on the distribution of the variables, student’s t-tests, Mann–Whitney U-tests, or chi-square tests were used to compare differences in individual characteristics between non-myopic and myopic participants.

Multiple logistic regression was used to estimate the associations between outdoor ALAN (per 10 units or per quartile increase) exposure and the prevalence of myopia. We constructed three models. Model 1 was not adjusted for any covariates. Based on univariate analysis (p < 0.1), Model 2 adjusted for age, sex, residence, nationality, school level, exposure time, body mass index, parental myopia, nighttime sleep, outdoor exercise time, fruit, vegetable, computers and TVs spend time, after-school homework time, after-school tutoring time, eye exercises, annual frequency of visual inspections, health technical personnel, NDVI, and PM_2.5. A stepwise selection approach was used to build a parsimonious model. Thus, Model 3 (main model) adjusted for age, sex, residence, nationality, school level, exposure time, body mass index, parental myopia, fruit, vegetable, after-school homework time, eye exercises, annual frequency of visual inspections, and PM_2.5.

In examining the relationship between outdoor ALAN exposure and the prevalence of myopia, we further investigated the shape of the exposure-response curve using restricted cubic spline (RCS) regression. For the RCS, the knots were set to the 5th, 25th, 50th, 75th and 95th percentiles, and the reference value was 0. The nonlinear p-value of the RCS was calculated by the Wald test.

Based on prior literature and our database, we selected five individual characteristics to explore the modifying effects, which were associated with myopia with moderate to high strength of evidence (3, 40). These characteristics were: (1) age (≤10 years and >10 years), (2) sex (male and female), (3) residence (rural areas and urban areas), (4) school level (primary, junior, and senior), and (5) outdoor exercise time (<1 h, 1–2 h, and ≥2 h). We used Z-tests to examine whether the differences between the subgroup effect estimates were significant or not. In addition, multiple comparisons existed for school level and outdoor exercise time. Thus, the Benjamini–Hochberg FDR method was calculated to correct for multiple comparisons.

We also performed a series of sensitivity analyses. First, we redefined four ALAN levels based on a variety of buffers (e.g., 3 km and 5 km) and annual averages (e.g., 1-year and 3-year average). Second, we excluded exposures with a duration of ≤1 year. Third, considering potential data clustering in schools, we used a generalized linear mixed model to test main results. All analyses were completed using SPSS for Windows (version 23.0) and R software (version 4.1.3). A two-tailed p-value of less than 0.05 was considered statistically significant.

Supplementary Table S1 shows the characteristics of 33,160 Chinese adolescents. The age of the students participating in the study ranged from 9 to 18 years (mean ± SD: 13.51 ± 2.55 years). Supplementary Table S1 also shows that the overall prevalence of myopia among students was 61.4% (20,325 out of 33,160). Table 1 further illustrates the variations in myopia prevalence across different groups. For example, a higher prevalence of myopia was observed among older adolescents, females, individuals from urban areas, and those of Han Chinese ethnicity, as well as among individuals with higher levels of education. Briefly, with the exception of five variables (outdoor exercise time, frequency of sugar, frequency of oil, GDP per capita, population density), all other variables showed differences between the two groups (p < 0.05).

The radiance of LAN exposure exhibited significant variation across Ningxia Province, as illustrated in Supplementary Figure S1. In general, most areas experienced lower-intensity outdoor ALAN, while higher-intensity outdoor ALAN was notably concentrated in northern cities. The distribution of outdoor ALAN among the 120 study sites (schools) displayed a leftward skew, and the median (IQR) ALAN exposure for all study sites was 12.90 (2.10, 25.04) nanoWatts/cm²/sr, as depicted in Supplementary Figure S3.

Significant differences in outdoor ALAN levels were observed between the non-myopia group and the myopia group (p < 0.001; see Figure 2). Outdoor ALAN levels were higher in the myopia group [median (IQR): 14.44 nanoWatts/cm²/sr (3.88–26.56)] than in the non-myopia group [median (IQR): 6.95 nanoWatts/cm²/sr (1.21–21.74)].

Table 2 presents the ORs and 95% CIs for the associations between outdoor ALAN exposure and the prevalence of myopia. In Model 1, without adjusting for any covariates, there was an 18% increase in myopia prevalence for every 10-unit change in outdoor ALAN. In Model 2, after adjusting for covariates identified through univariate analysis (p < 0.1), a 4% increase in myopia prevalence was observed for every 10-unit change in outdoor ALAN. Similar results were obtained using stepwise regression approaches in Model 3.

To preliminarily explore non-linear relationships, we categorized outdoor ALAN into quartiles (Q1, Q2, Q3, and Q4). As shown in Table 2, association estimates were calculated for the top three quartiles compared to the lowest outdoor ALAN quartile. In Model 1, statistically significant positive associations were observed between outdoor ALAN and myopia prevalence (OR_{Q4 vs. Q1}: 2.08, 95% CI: 1.95–2.22). Model 2 and Model 3 found similar association estimates (OR_{Q4 vs. Q1}: 1.21, 95% CI: 1.08–1.35; OR_{Q4 vs. Q1}: 1.20, 95% CI: 1.08–1.34, respectively). Considering the simplicity of the model, we use Model 3 as the main model for subsequent analysis.

We further investigated the shape of the exposure-response curve in examining the relationship between outdoor ALAN exposure and myopia prevalence (Figure 3). The relationship between outdoor LAN exposure (reference concentration: 0 nanoWatts/cm²/sr) and the prevalence of myopia exhibited a nonlinear exposure-response curve (p for nonlinear <0.001). The effect steeply increased at low outdoor ALAN levels, remained stable at medium levels, and gradually increased at high levels. Overall, the RCS analysis indicated a nonlinear positive dose-response association between long-term outdoor ALAN exposure and myopia in adolescents.

Table 3 shows the associations of outdoor ALAN with myopia prevalence modified by age, sex, residence, parental myopia, school level, and outdoor exercise time. Compared to the outdoor ALAN in Q1, the outdoor ALAN in Q4 was significantly associated with increased odds of myopia in adolescents older than 10 years of age, with no myopic parents, attending junior school, and spending more time outdoors. Both males and females showed a significant increase in the prevalence of myopia when exposed to outdoor ALAN in Q4 compared to Q1. Compared to the outdoor ALAN in Q1, the outdoor ALAN in Q2 was significantly associated with increased odds of myopia in rural areas. It is worth noting that we only found statistically significant differences between urban and rural areas (p = 0.029).

In sensitivity analyses, the associations between outdoor ALAN and myopia prevalence remained stable (see Supplementary Tables S2, S3). First, defining four ALAN levels was based on a variety of buffers (e.g., 3 km and 5 km) and annual averages (e.g., 1-year and 3-year average). The associations varied to some extent in different definitions, but the direction remained stable. Second, excluding exposure time ≤1 year, we observed a slight decrease in the effects (OR_{Q4 vs. Q1}: 1.16, 95% CI: 1.02–1.31). Third, considering potential data clustering in schools, we used a generalized linear mixed model and observed the increased effects. For instance, a 8% increase in myopia prevalence was observed for every 10-unit change in outdoor ALAN. Compared to the outdoor ALAN in Q1, the outdoor ALAN in Q4 was significantly associated with increased odds of myopia (OR_{Q4 vs. Q1}: 1.40, 95% CI: 1.00–1.96).

This study is the first to identify a positive nonlinear association between long-term exposure to outdoor ALAN and myopia in adolescents. Although a stronger correlation was not found in a specific population (e.g., sex), the findings still offer important insights. Given the rising rates of ALAN and myopia in contemporary society, the study offers crucial first-hand evidence of ALAN’s potential impact on eye health.

Holden et al. (4) reported a global myopia prevalence of 22.9% (1.406 billion people) in 2000, projecting a significant increase to 49.8% (4.758 billion) by 2050 (4). A comprehensive review underscored a non-coincidental correlation between the widespread adoption of LEDs and the escalating prevalence of myopia (14). While daytime LEDs used for general lighting may impact visual performance and eye strain, ALAN, characterized by exposure to artificial light during nighttime hours, may have distinct and potentially more significant effects on myopia development (14, 25). Research on the association of ALAN with myopia is very limited, as only a few epidemiologic studies conducted abroad have shown inconsistent results (30–33). Quinn et al. (33) identified a strong link between childhood myopia development and exposure to nighttime lighting during the first 2 years of life. However, a follow-up study by Gwiazda et al. (32) found no significant relationship between nighttime lighting exposure—whether in the first 2 years or later—and the prevalence of myopia. Similarly, Czepita et al. (30) observed a high incidence of astigmatism associated with fluorescent lighting. In a subsequent study, Czepita et al. (31) found no association between myopia and nighttime light use in Polish children without a family history of myopia. These inconsistent results may partly be attributed to the accuracy of exposure measurements, statistical methods, and the limitation on specific populations.

Our study extends previous work in three ways. First, unlike previous studies that used questionnaires to assess artificial light (30–33), we used relatively high-precision data of outdoor ALAN (38). By focusing on outdoor ALAN, our study introduces a new perspective on environmental influences on myopia, offering insights beyond the commonly studied indoor lighting. Second, we used a variety of methods, including logistic regression and RCS methods, to mutually validate the results. These methods demonstrated a positive nonlinear dose–response relationship between ALAN and myopia, providing a more comprehensive understanding of the association. Third, to our knowledge, this is the first and largest study examining the link between ALAN and myopia in adolescent. These three extensions further add support for a positive association between ALAN and myopia.

The exact mechanisms between ALAN and myopia are not clear. One possible mechanism is that long-term exposure to ALAN may lead to severe disruption of circadian rhythms, thereby altering the circadian rhythm of the retina. Several systematic reviews have detailed this mechanism, including Lei et al. (27), Regmi et al. (42), and Nickla (25). In addition, several other systematic reviews have noted that retinal circadian rhythm disturbances may specifically affect the onset and progression of myopia, such as Chakraborty et al. (20), Stone et al. (21), and Zhang et al. (14). Another possible mechanism is that ALAN inhibits melatonin production and deprives sleep, both of which are also important factors in the development of myopia, as supported by Chakraborty et al. (19) and Touitou et al. (22). These mechanisms form, in part, the biological basis for the link between ALAN and myopia.

Our study found that long-term exposure to outdoor ALAN may be positively associated with myopia. This association provides important insights for clinical practice and public health. Ophthalmologists can advise adolescents to take a series of measures, such as avoiding overexposure to strong light sources at night, which are expected to reduce the incidence of myopia. Eye care professionals can advise adolescents to avoid or minimize exposure to ALAN, which may help reduce the incidence of myopia. For high-risk groups in light-polluted environments, especially students, regular eye health checkups are especially important, and early detection and intervention will improve the effectiveness of myopia prevention and control.

In addition, understanding the association between outdoor light pollution and myopia provides an important knowledge base for the public. Although no specific sensitive groups were identified, the effect did diminish after adjusting for other known risk factors such as lifestyle factors. This suggests that individuals may be able to take steps to mitigate the effects of ALAN. Educational institutions and health departments can work together to conduct publicity campaigns to raise public awareness of outdoor light pollution in order to reduce its harmful effects on eye health. It is crucial to incorporate light pollution prevention and control measures in urban planning, such as rationalizing lighting facilities and reducing glare exposure, to improve the urban environment and reduce eye health risks. Taken together, this study provides substantial guidance for the development of more effective eye health strategies that not only contribute to the eye health of individuals, but also have a positive impact at the public health level.