Introduction

Myopia remains a critical public health issue in China and is projected to reach approximately 84% by 20501,2. In Asian regions, 10–20% of high school students are currently affected by high myopia3,4. Once progressing to moderate or high myopia, individuals face significantly elevated risks of developing glaucoma, cataracts, retinal detachment, and myopic macular degeneration5,6,7. Uncorrected refractive errors remain the leading cause of visual impairment in China, with severe vision impairment cases increasing by 147% over the past three decades8. For individuals with severe visual impairment to blindness, the annual productivity loss is estimated at US$9.4 billion (range: US$8.5–10.4 billion)9.

While myopia is prevalent during adolescence, commonly cited factors that may be associated with myopia include reduced physical activity10, insufficient outdoor activity11, insufficient sleep time12, a high level of education13, and excessive digital screen use14. However, the causal relationships between these factors and myopia progression remain incompletely understood. Current evidence on the risk factors and etiological mechanisms of moderate-to-high myopia in adolescents remains particularly limited. Existing evidence indicates significant alterations in adolescent behavioral patterns during the COVID-19 pandemic15. Prolonged home confinement was associated with persistent rapid onset and progression of myopia in pediatric populations, showing no signs of improvement16,17. This also implies that the effects of previous prevention and control strategies focusing on individual factors still need to be improved. Middle school students in economically developed regions face intense academic competition pressure18, significantly elevating their risk of developing moderate-to-high myopia. The age range of 12–18 years represents a critical developmental transition, during which health risk behaviors are either adopted or abandoned, often clustering due to evolving social and physiological contexts19,20. Globally, over one-third of adolescents report three or more health risk behaviors, with prevalence increasing significantly with age21. Characterizing these behavioral patterns could inform clinical management strategies for moderate-to-high myopia. Previous trajectory studies on myopia have primarily focused on either single-behavior analyses or ocular biometric trajectories22,23. In contrast, cluster analysis of multiple behavioral factors has been extensively utilized in cardiovascular disease research, generating innovative findings24,25. Are myopia-related behaviors also clustered in the population? If so, what are these characteristics and how do they relate to myopia?

The present study hypothesized that adolescents exhibited clustered patterns of myopia-related lifestyle risk factors, which demonstrated specific characteristics and quantifiable associations with myopia progression and ocular biometric parameters. The aim of this study was to evaluate the epidemiology of myopia-linked behavioral patterns over a five-year period among middle school students in Chinese economically developed areas and investigate its associations with myopia.

Methods

Study population

This study was a prospective cohort study based on the Surveillance and Intervention of Common Diseases and Health Influencing Factors of Students in Shanghai, focusing on middle school students (aged 12–17) in Jiading District from 2019 to 202326. The dataset was obtained from the Jiading District Center for Disease Control and Prevention. This ongoing prospective cohort study used a multi-stage stratified cluster sampling scheme to select 10 schools (including kindergartens, elementary schools, middle schools, academic high schools, vocational high schools, and colleges) within the district‌ as national surveillance sites for ‌student common diseases and health risk factors‌. Follow-up surveys were conducted during the first semester (September to December) of each school year‌. Baseline data were collected in 2019, with annual follow-up examinations through 2023; only participants with two or more follow-up records were included‌. After excluding 4 kindergartens and elementary schools due to age ineligibility, 1,945 students aged 12–17 years were recruited from six schools, representing the following grades: junior high school preparatory class (12 years of age), junior high school year 1 (13 years of age), junior high school year 2 (14 years of age), senior high school year 1 (16 years of age), senior high school year 2 (17 years of age), vocational high school year 1 (16 years of age) and vocational high school year 2 (17 years of age).

The flow diagram is shown in Fig. 1. ‌This study protocol was approved by the Ethics Committee of the School of Public Health, Fudan University (Approval No. IRB#2024–05-1115) and adhered to the tenets of the Declaration of Helsinki. Informed consent was obtained from a parent and/or legal guardian for study participation.‌

Fig. 1
figure 1

Flowchart of study participants. “Sufficient numbers of lifestyle measurements” was defined as least two lifestyle monitoring data measured during 2019 to 2023. The 19 participants removed was specifically excluded due to missing of eye examination data.

Full size image

Eye examinations

This study utilized‌ data from the participants’ baseline and final vision examination results. Participants underwent comprehensive ophthalmic examinations annually from September 2020 to September 2023. An experienced team comprising one ophthalmologist, five optometrists, and five trained ophthalmic assistants performed the ophthalmic examinations. Uncorrected distance visual acuity (UCVA) was measured using a standard logarithmic visual acuity E chart (adhering to the National Standard of People’s Republic of China, GB 11,533–1989) under indoor lighting. Individuals with UCVA ‌ ≤ 5.0‌ in either eye underwent ‌additional assessment of best-corrected visual acuity‌. Axial length (AL) was measured three times per eye using an IOL Master (version 5.02; Carl Zeiss, Jena, Germany). Non-cycloplegic autorefraction was conducted using a closed-field desk-mounted auto-refractor (KR-8900; Topcon, Tokyo, Japan) to determine spherical equivalent (SE) and capture corneal radius (CR) measurements in two meridians (CR1 was calculated for the horizontal meridian with an axis range of 180–45° and 135–180°, while CR2 was assessed for the vertical meridian within an axis range of 45–135°). Three consecutive readings were documented, with automatic averaging for each eye‌. ‌If two of the three measurements differed by > 0.02 mm for AL/CR or > 0.50 D for refraction, the eye was re-examined‌.

The primary outcomes of this study were incident myopia, changes of SE, AL, and AL/CR ratio. Secondary outcomes included prevalent myopia, myopia degree and other ophthalmic parameters(AL, AL/CR ratio) measured at the final follow-up‌. SE was defined as‌ sphere power plus half of the cylinder power. At the final follow-up, prevalent myopia was defined as SE < −0.50 D in eyes with UCVA < 20/20, low myopia, moderate myopia, high myopia were defined as SE > −3.00 D, −3.00- −5.99 D, ≤ −6.00 D27. Incident myopia was defined as new-onset myopia in baseline non-myopies. Longitudinal changes‌ in SE (ΔSE), AL (ΔAL), and AL/CR ratio (ΔAL/CR) ‌among myopic students were computed as final minus baseline values‌. Using P50 as the cut-off value, ΔSE, ΔAL and ΔAL/CR ratio values were divided into the rapid progression group and the slow progression group28,29.

Questionnaires

Based on previous literature11,12,13, we selected ‌the following lifestyle indicators‌ for trajectory analysis: sleep time, moderate and vigorous physical activity (MVPA), outdoor time, screen time and extracurricular study time (hereafter “extra-study time”). The validated questionnaire‌ demonstrated a Cronbach’s α coefficient of 0.90 and a content validity coefficient of 0.86, adapted from‌ the ‌2019 National Survey on Monitoring and Intervention of Common Diseases and Health Influencing Factors Among Students30.

Sleep time

Sleep time was students’ self-reported average sleep time on school days and weekends over the past week. “On school days (Monday to Friday) over the past week, what was your average daily sleep time?” and “On weekends (Saturday and Sunday) over the past week, what was your average daily sleep time?” The formula for calculating the daily average sleep time was: daily sleep time = sleep time on school days * 5 + sleep time on weekends * 2)/7.

MVPA

Physical activity was assessed using a self-reported questionnaire that measured the number of days per week students engaged in MVPA for at least one hour. The specific questionnaire item was “On how many days in the past week did you engage in at least one hour of moderate to vigorous physical activity (can be accumulated)?”.

Outdoor time

“On school days (Monday to Friday) over the past week, what was your average daily outdoor activity time?” and “On weekends (Saturday and Sunday) over the past week, what was your average daily outdoor activity time?” The formula for calculating the daily average outdoor time was: daily outdoor time = outdoor time on school days * 5 + outdoor time on weekends * 2)/7.

Screen time

Screen time was defined as the cumulative amount of time students used electronic screens on an average day in the past week. The specific questionnaire item was “In the past week, how much time on average per day did you spend watching TV, using a computer, mobile phone, game console, or other electronic screens?”.

Extra-study time

Extra-study time was defined as the cumulative amount of time students attended cram school classes in the past week. The specific questionnaire item was “In the past week, how much total time did you spend attending learning extracurricular classes (e.g., English, math, writing, etc.)”.

Covariates

Covariates included age and height ‌at the time of refraction measurement, sex (male/female), school type (vocational school/academic school), grade, self-perceived family income (poor/fair/fine), parental myopia (none/only one/both) and maternal education and paternal education (junior high school or below/high school/college degree or above). All covariates were concurrently collected alongside primary variables using standardized student self-report questionnaires.

Statistical analysis

Conventional statistical practice generally falls far short of taking full advantage of the information available in multivariate longitudinal data for tracking the course of the outcome of interest. Group-based multi-trajectory modeling (GBTM), introduced by Nagin31 in 2018, is a generalization of univariate group-based trajectory modeling. Trajectory models for sleep time, MVPA, outdoor time, screen time, and extra-study time were developed using the‌ traj ‌plugin in Stata 15.1‌. All variables ‌were modeled under censored normal distributions‌. The number of classes that best fit was selected based on the acceptability of the available fit-criteria indices (i.e., average posterior probability of assignment [APPA], Bayesian Information Criterion [BIC], Akaike Information Criterion [AIC], Entropy, Supplementary Table 1). We consider the optimal grouping model as following criteria: (1) the absolute values of BIC and AIC are minimized. (2) The Entropy value exceeds 0.7. (3) The APPA is above 0.8. (4) The sample size percentage of each group is greater than 2%. (5) The model must be one that is substantially useful and explainable by professional knowledge.

Logistic regression examined associations between ‌lifestyle trajectories‌ and prevalent/incident myopia and rapid progression of optometric indicators. Ordinal logistic regression models assessed relationships between myopia levels and the lifestyle indicators. Generalized linear regression models evaluated associations between the lifestyle indicators and AL, CR, and AL/CR ratio. To test model robustness, ‌a subsample‌ (n = 996) ‌with ≥ 3 longitudinal lifestyle measurements‌ underwent ‌replicated GBTM analysis‌, followed by myopia association testing. Based on established literature13, we included the following covariates: student demographics, educational attainment, parental myopia history, parental demographics, and household socioeconomic status.

Significance tests were 2-tailed, and P < 0.05 was considered to be statistically significant. Since the lifestyle indicators were compared five times across five waves, we applied ‌Bonferroni correction‌ with a ‌significance threshold‌ of P < 0.01 (0.05/5) ‌for multiple testing adjustment‌. All analyses were performed using R, version 4.2.2 (R Group for Statistical Computing) and Stata, version 15.1 (StataCorp LLC).

Results

A total of 1,945 middle school students were included from five waves of longitudinal lifestyle monitoring‌. Participants were followed for a median of ‌3 years (IQR = 2–4 years). At baseline, the mean (SD) age was 14.21 ± 2.00 years, with 59.3% (1154/1945) being boys, 73.4% (1427/1945) having myopia. At the final visit, the mean age was 16.18 ± 1.58 years, 81.4% (1583/1945) students were myopic. From 2019 to 2023, the incidence of myopia was 37.3% (193/518). Among myopic children, mean changes of SE, AL, AL/CR ratio were – 0.56 ± 1.05 diopters, 0.27 ± 0.45, 0.01 ± 0.08, respectively.

Trajectories classification of lifestyle indicators: descriptive results

Group-based trajectory modeling‌ identified ‌three distinct lifestyle patterns‌ (trajectory classifications are shown in Fig. 2, final model fit statistics are shown in ESM Table 1, class selection models fit statistics are shown in ESM Table 2, descriptive analysis of lifestyle measures are shown in ESM Table 3). The ‌three-class solution‌ was selected based on its low BIC value, sufficient subgroup sample size and interpretability. Then we designated them based on the comparison of the distributional traits of lifestyle indicators between groups as follows:

Fig. 2
figure 2

3 Latent classifications of lifestyle trajectory from 2019 to 2023. “General”, account for 38.4% (n = 747); “Rapidly declining sleep time and prolonged extra-study time”, included 53.3% (n = 1036) of participants; “Persistently low MVPA and prolonged extra-study time”, included 8.3% (n = 162) of participants; MVPA = moderate to vigorous physical activity.

Full size image

(1) The “General” group, account for 38.4% (n = 747), exhibited longer sleep duration (around 8 h/day), engaged in MVPA around 3 days a week, had increased and consistent outdoor time (approximately 2 h/day), spent less time on studying (around 0.5 h/day), and more time on screens (about 2 h/day).

(2) The “Rapidly declining sleep time and prolonged extra-study time” group, included 53.3% (n = 1036) of participants, showed a variation in sleep duration (from around 8.5 h/day to 7.15 h/day) with age. They engaged in more MVPA (about 3 days/week) while spending less time outdoors (reducing from 1.8 h/day to 1.5 h/day), had less screen time (around 1 h/day), and allocated more time to extracurricular studies (approximately 2.5 h/week).

(3) The “Persistently low MVPA and prolonged extra-study time” group included 8.3% (n = 162) of participants, exhibiting fluctuating sleep time (approximately 6.6–8.4 h/day), minimal MVPA (approximately 0.6 days/week), moderate outdoor time (around 1.5 h/day), moderate screen time (about 1.7 h/day), and spent a moderate amount of time on studies (approximately 2 h/day).

Demographic characteristics of 1945 participants are presented in Table 1. Compared to the other two groups, students in “General” group were more likely to be boys, older, taller, in vocational schools, less likely to have myopic, educated parents, or a good family economy (χ2 = 26.53–519, all P values < 0.001). The prevalence of myopia degrees‌ at baseline and endpoint ‌for‌ the three trajectories ‌is‌ shown in Fig. 3;‌ ‌a comparison‌ of lifestyle trajectory groups with ophthalmic parameters ‌is‌ presented in ESM Table 4. At the endpoint, the “Rapidly declining sleep time and prolonged extra-study time” group exhibited the highest prevalence of myopia and high myopia (83.98%, 13.90%), higher than “General” group (77.38%, 12.05%), P = 0.001, P = 0.013; but not higher than the “Persistently low MVPA and prolonged extra-study time” group (82.10%, 11.11%), all P values > 0.05. The “Rapidly declining sleep time and prolonged extra-study time” group had more rapid SE progression (52.53%) than the “General” group (45.65%), P = 0.031. Conversely, there were no notable distinctions in AL, AL/CR ratio at the final visit, ΔAL, ΔAL/CR ratio, and myopia incidence, with all P values > 0.05.

Table 1 Baseline characteristics of the study participants.
Full size table
Fig. 3
figure 3

Myopia degree prevalence of participants in each trajectory group from baseline to endpoint. Each panel presents the prevalence of myopia degree for each model group. Group 1, “General”; Group 2, “Rapidly declining sleep time and prolonged extra-study time”; Group 3, “Persistently low MVPA and prolonged extra-study time”. For myopia and high myopia prevalence in endpoint, P < 0.05 between “General” and “Rapidly declining sleep time and prolonged extra-study time” groups.

Full size image

Associations between lifestyle indicators trajectories and ophthalmic outcomes

Table 2 shows the comparison of individual lifestyle indicators across five waves from 2019 to 2023 between myopia and AL. The differences of all lifestyle indicators between myopia and no myopia groups did not reach significance, with all P values > 0.01. However, ‌outdoor time ‌and‌ screen time in 2021 were negatively associated with AL, with all P values < 0.01.

Table2 Comparison of lifestyle measures across five waves with myopia and AL among middle school students in 2019–2023.
Full size table

The results of associations between lifestyle indicators trajectories and myopia are shown in Table 3. The “Rapidly declining sleep time and prolonged extra-study time” lifestyle increased the risk of myopia in the final visit, with an adjusted OR 1.30 (95%CI, 1.01 to 1.67, P = 0.039), compared to the “General” group. The “Rapidly declining sleep time and prolonged extra-study time” group had a 10% increased risk of worsening myopia degree (95%CI, 1.01 to 1.20, P = 0.045). Longer AL (β coefficient: 0.17; 95%CI = 0.05 to 0.29) and a positive AL/CR ratio (β coefficient: 0.02; 95%CI = 0.01 to 0.03) were ‌observed in the “Rapidly declining sleep time and prolonged extra-study time” group. Furthermore, the “Rapidly declining sleep time and prolonged extra-study time” lifestyle increased the risk of rapid SE progression, with an adjusted OR of 1.35 (95%CI = 1.06 to 1.72, P = 0.016). The difference between “Rapidly declining sleep time and prolonged extra-study time” and “General” in incident myopia was close to statistical significance, P = 0.104. However, no statistically significant difference was observed in the progression of AL and AL/CR ratio among the three groups, with all P values > 0.05.

Table 3 OR/β (95% CI) for associations between lifestyle trajectory and prevalent myopia, incident myopia, myopia progression.
Full size table

A sensitivity analysis that produced broadly consistent findings is provided in Electronic supplementary material 1(ESM Fig. 1, ESM Table 5).

Discussion

In this cohort study, we identified three distinct lifestyle trajectories: the “General” group, accounting for 38.2%; the “Rapidly declining sleep time and prolonged extra-study time” group, comprising 52.2% of participants; the “Persistently low MVPA and prolonged extra-study time” group, comprising 8.3% of participants. Boys, vocational school students, with less myopic and less educated parents, poorer families were more likely to have the “General” lifestyle. Compared to those with “General” lifestyle, the “Rapidly declining sleep time and prolonged extra-study time” group students had higher risk of myopia especially moderate to high myopia, longer AL, higher AL/CR ratio, and more rapid SE progression.

There are 3 specific clustering patterns of myopia-related behaviors

We identified three lifestyle trajectories among adolescents in this economically developed city along the eastern coast of China. This categorization is robust and was validated in trajectory subgroups with three or more recorded follow-up visits. One-third of them predominantly adhered to the current guidelines27,32, except for MVPA, which fell significantly below recommendation standards. Screen time increased with age across all trajectory groups, but adolescents leading the “General” lifestyle‌ spent the longest periods (mean time: 2 h) on screens. Additionally, around half of the adolescents lacked sufficient sleep and had persistently long extra-study time. Less than ten percent maintained the “Persistently low MVPA and prolonged extra-study time” lifestyles, characterized by low MVPA and relatively high extracurricular academic demands, while other factors fell within moderate ranges. Adolescents who were male or attended‌ vocational high schools, were more likely to adopt healthy lifestyles. Compared to academic students, vocational school students in China have more recreational breaks and a lighter study burden, which may partly explain the lower prevalence of myopia in this group33,34. Within intensive education systems, parents from affluent and well-educated backgrounds often enroll their children in extracurricular cram schools35. This practice‌ imposes a heavy study burden on children, leaving them with little time for sleep and making it difficult to maintain a healthy lifestyle36.

Middle school students with a “Rapidly declining sleep time and prolonged extra-study time” lifestyle were at a higher risk for myopia

The “Rapidly declining sleep time and prolonged extra-study time” trajectory exhibited the highest prevalence of myopia and showed the longest AL. Moreover, this lifestyle was‌ associated with a higher risk of moderate to high myopia and rapid progression of myopia. Since 73% of participants already had myopia, the association between lifestyle trajectory and incident myopia in this study was not significant. Once myopia sets in, there is an extended window for progression toward high myopia37. Limited research has been conducted on lifestyle indicators associated with higher myopia38. The present study suggests that health behaviors can still be improved after the onset of myopia to protect eye health and slow its progression. The “Persistently low MVPA and prolonged extra-study time” lifestyle had a myopia prevalence second only to that of the “Rapidly declining sleep time and prolonged extra-study time” group, although the difference compared to the “General” group was not statistically significant. Our analysis suggests this may stem from the exceptionally small sample size, and the better performance of extra-study time compared to the “Rapidly declining sleep time and prolonged extra-study time” group.

The results of this study indicate that only SE progression was significantly associated with lifestyle trajectories, while AL progression showed no significant association. While non-cycloplegic refraction shows reasonable reliability after age 12, when refractive status typically stabilizes, it may overestimate myopia39,40. Consequently, observed rapid SE progression could reflect measurement error or refractive accommodation, not true axial elongation. Verifying this effect necessitates longitudinal studies with larger sample sizes and extended follow-up and recommends the use of cycloplegic refraction.

Aggregated patterns of multiple behaviors help identify new associations

Few studies have analyzed the effects of multiple behavioral aggregation patterns on myopia, which may be closer to the real, complex world. In contrast, the multi-GBTM approach integrates multiple factors across repeated measurements, effectively identifying susceptible populations and clarifying the role of individual factors within aggregated behavioral patterns. Insufficient epidemiological evidence exists to establish associations between sleep time41, MVPA42,43, screen time44 and myopia, these associations are confounded when considered in isolation. This instability may stem from confounding factors, as distinct clustering patterns of behaviors can compromise the stability of single variable-outcome associations. In the present study‌, we suggest that ‌such unhealthy lifestyle may negatively affect well-being and increase susceptibility to common chronic non-communicable diseases21,45. Multiple studies in North America, Asia, and Europe have found that unhealthy lifestyles, characterized by increased risk factors, are associated with a higher risk of non-communicable diseases and all-cause mortality45,46,47. It is possible that MVPA, sleep time and screen time have little effect on myopia when considered in isolation, but their impact may become more apparent when combined with other factors. Guggenheim et al48 suggested that the effects of lifestyle risk factors on myopia have been underestimated in previous studies. A similar cumulative effect ‌has‌ been identified in multifactorial myopia study49. As for outdoor activities and extra-study time, outdoor activities exhibit a robust protective impact against myopia13,50, while education stands out as a risk factor for myopia, and has been confirmed in many studies13,51,52. We found that outdoor activity levels were consistently low among all middle school students, with minimal variance. ‌Numerous studies demonstrate that outdoor activities protect against myopia onset, yet evidence regarding their effect on myopia progression remains lacking50,53. Further, the lack of significant association between outdoor time and myopia in middle school students may stem from unaccounted clustering of other myopia-related behaviors at the individual level.

Among these clustered behaviors, some remain relatively stable with minimal variation, while others exhibit rapid changes over time. ‌Notably‌, extra-study time and MVPA ‌demonstrated‌ substantial population variability and ‌served as key discriminators‌ of behavioral patterns. These behaviors ‌may‌ accumulate over time, ‌potentially exerting‌ cumulative effects54. ‌Our trajectory analysis‌ reveals the ‌longitudinal‌ impact of myopia-related behaviors on visual outcomes, ‌underscoring‌ the need for timely interventions—‌particularly‌ targeting ‌deteriorating‌ behavioral patterns like excessive extra-study time.

Strengths and limitations

Most studies on myopia-related factors have been cross-sectional and focused on a single factor. In contrast, this longitudinal study aggregated multiple behavioral factors, exploring the epidemiological trajectory of several factors over a five-year period. This allowed us to explore the role of certain factors in aggregating behavioral patterns for myopia. In addition, by integrating a wide range of optometric data, including AL, CR, and others, this study provided a more objective reflection of the refractive status of the subjects.

Lifestyle factors were collected via self-reported questionnaires‌, which may introduce potential recall bias into the study. Due to the high dropout rate in routine monitoring programs, we accounted for the potential loss of accuracy in models, the high variance in follow-up duration, and depicting a narrower age range within the trajectories. Although cycloplegic autorefraction was omitted, the refractive status of adolescents over 12 years of age was considered relatively stable. Regional differences also constrained the generalizability of the findings ‌among‌ middle school ‌Chinese students‌.

Conclusion

Myopia-related factors tended to cluster among middle school students in eastern China’s economically developed areas‌. The lifestyle characterized by prolonged‌ extra-study time during ages 12 to 17 was associated with increased risks for multiple domains of myopic outcomes over 5 years. Focusing solely on individual factors may offer limited effectiveness in prevention and control. It is vital to adopt a multifactorial and multi-disease approach, tailoring prevention and control strategies to individuals based on various behavioral patterns.