Trunk-rotation and posture-screening markers in relation to refractive anisometropia among schoolchildren in Nantong, China: a cross-sectional school health surveillance study

Abstract

Background

School health programmes often collect vision, growth and posture-screening data in the same children, but these streams are usually analysed separately. We examined whether trunk-rotation and posture-screening markers were associated with refractive anisometropia burden in school surveillance.

Methods

This cross-sectional analysis used 2025 school health surveillance records from Nantong, China. The primary outcome was non-cycloplegic screening-defined refractive anisometropia, defined as an interocular spherical-equivalent difference of 1.00 dioptre (D) or greater. Primary posture-screening markers were maximum angle of trunk rotation (ATR), ATR thresholds and operationally defined inspection or movement-screening indicators. Adjusted prevalence ratios (aPRs) were estimated with modified Poisson regression adjusted for age, sex, body mass index and sampled school-class cluster fixed effects.

Results

Among 2605 students, 713 (27.4%) had anisometropia of 1.00 D or greater. The main continuous marker, maximum ATR per 1 degree, was modestly associated with anisometropia (aPR 1.04, 95% CI 1.01 to 1.07). ATR of 7 degrees or greater showed a higher crude prevalence (40.0% versus 26.9%) and a directionally similar but less precise adjusted estimate (aPR 1.30, 95% CI 1.01 to 1.67). Broad myopia and axial-myopia markers showed little corresponding positive pattern. Secondary severity, school-stage and segment-direction analyses were descriptive and exploratory.

Conclusions

Trunk-rotation and posture-screening markers were modestly associated with concurrent refractive anisometropia burden in school surveillance data. These findings support integrated analysis of school posture and vision screening data, but should not be interpreted as evidence of temporal order, causality, individual-level prediction, or posture-based ophthalmic referral.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28630-4.

Background

School health systems are increasingly expected to identify more than one child and adolescent health risk during the same surveillance cycle. The recently issued national Children and Adolescents’ “Five Health” Promotion Action Plan (2026–2030) places spinal-health promotion and visual-health promotion within the same child and adolescent health agenda in China [1]. Vision screening remains central to this agenda because refractive error is common in school-age populations and has risen rapidly in many settings. Global modelling projected that 49.8% of the world’s population could have myopia and 9.8% could have high myopia by 2050 [2]. China-specific evidence also indicates a large and heterogeneous school-age burden, with recent systematic and school-based studies documenting high age-specific prevalence, pandemic-era shifts and axial-length distributions relevant to surveillance [3–6].

Spinal and postural problems represent a parallel school-health burden. Recent syntheses show that scoliosis and spinal-curvature abnormalities occur across childhood and adolescence and vary by sex, age, region and subtype [7, 8]. These findings make posture, growth stage and musculoskeletal health population-level school-surveillance concerns rather than only specialist orthopaedic issues.

Although school vision programmes commonly summarise myopia as a child-level condition, refractive anisometropia is a between-eye visual-risk phenotype that can be hidden by such summaries. It refers to a clinically meaningful difference in refractive status between the two eyes. In children and adolescents, this imbalance can leave one eye with more uncorrected blur than the fellow eye, disturb binocular fusion and stereopsis, increase amblyopia risk and contribute to visual fatigue during near work [9–11]. School-based and population studies in China and other countries have reported heterogeneous anisometropia estimates, partly because age range, refraction method and threshold definitions differ [12–17]. For school-health surveillance, anisometropia is therefore a measurable child visual-risk phenotype rather than only a narrow ophthalmic mechanism.

The origins and correlates of anisometropia are not fully defined. Between-eye differences may reflect asymmetric ocular growth, refractive status, emmetropisation, amblyopia, strabismus and unequal visual experience [9]. Posture is relevant to this uncertainty, but the available evidence remains mixed. Optical reasoning suggests that short working distances and markedly asymmetric head postures may alter patterns of retinal defocus during near work [18]. More recent objective photogrammetric evidence has reported associations between specific postural misalignments, including head tilt and writing posture, and anisometropia or inter-eye ocular differences [19]. In contrast, school-screening evidence has not consistently supported a direct relationship between refractive errors and scoliosis-screening findings [20]. Taken together, these mixed findings support integrated analysis of routinely collected posture-screening and bilateral refractive data to determine whether the two domains show consistent co-occurrence in school surveillance.

Routine posture screening makes this question testable at scale. Inspection, Adam’s forward-bend testing and angle of trunk rotation (ATR) measurement are widely used in school-age populations because they are feasible and supported by evidence on reproducibility and screening accuracy [21–23]. In the present study, these records were treated as screening markers of observed trunk rotation and visible postural asymmetry, not as diagnostic scoliosis measures or habitual-posture exposure measures. Linking them with bilateral vision data provides a practical way to characterise cross-domain co-occurrence within existing school-health workflows. Comparison with broad myopia and axial-myopia markers shows whether any corresponding pattern extends to general refractive burden.

Using routine school surveillance records from Nantong, China, we therefore estimated cross-sectional associations between trunk-rotation/posture-screening markers and non-cycloplegic screening-defined refractive anisometropia. The primary aim was to determine whether greater maximum ATR and other posture-screening markers were associated with higher concurrent anisometropia burden and whether corresponding patterns were present for broad myopia or axial myopia. Secondary analyses described variation by anisometropia severity, school stage and anatomical segment or direction.

Methods

Study design and setting

The Nantong School-aged Children Eye Study (NSES) is a registered observational school-eye study designed to investigate refractive errors, ocular biometry and related health factors among school-aged children in Nantong, Jiangsu Province, eastern China (Chinese Clinical Trial Registry: ChiCTR2300077367; registered on 6 November 2023). Nantong is a prefecture-level coastal city in the Yangtze River Delta. According to the latest official municipal statistical communique, the permanent resident population of Nantong was 7.749 million at the end of 2024 [24]. In 2025, per-capita disposable income was RMB 57,971 in Jiangsu Province and RMB 57,242 in Nantong City, indicating that Nantong was close to the provincial socioeconomic level [25, 26]. The sampling frame and school-stage context were defined using the latest publicly available Nantong education statistics [27].

This 2025 cross-sectional analysis used de-identified Student Health Surveillance Program records collected from October to November 2025. Students underwent anthropometric assessment, distance visual-acuity testing, non-cycloplegic refraction, ocular biometry and trunk and posture screening within the routine surveillance workflow. Because the present analysis focused on anisometropia, the sample-size check used the previous Nantong school-based anisometropia prevalence estimate of 25.6% [12]. Using the single-proportion precision formula (n0 = Z^2p(1-p)/d^2), with Z approximating 2 for an approximate 95% confidence level and absolute precision set at d = 0.1p [28], the unadjusted sample size was approximately 1163 students. After multiplying by a design effect of 1.5 and dividing by 0.95 to allow for an anticipated 5% nonresponse rate, the required minimum was approximately 1836 students.

The 2025 wave used a citywide, grade-stratified, school-based class-cluster surveillance framework. According to the Nantong student health surveillance plan, 60 schools were selected across primary, junior high and senior high school stages. Within each selected school, one intact teaching class was sampled, and all enrolled students in that class were invited to participate in the routine surveillance examination. Because most schools contain multiple teaching classes, the sampled class should be interpreted as a school-class cluster rather than as a school-wide census. The audited school, grade and class fields generated an anonymised school-class code for the 60 sampled clusters. Because one class was sampled from each selected school and the de-identified analytic export did not retain school identifiers separately from class identifiers, school-level and class-level components could not be modelled separately. Field logs recorded 2736 invited students, 12 absences, no recorded refusals or guardian non-consent, and 2724 students who attended field screening. Before generation of the audited de-identified export, 81 students were excluded because of ocular-treatment or ocular-condition criteria likely to affect refraction or keratometry, and 35 additional records were not retained because of incomplete or unusable records, information loss or field-level quality-control exclusions. The final analytic sample exceeded the minimum calculated sample size. The sample should be interpreted as a routine surveillance extract from selected school-class clusters in Nantong, China, not as a probability sample of all students in Nantong. Reporting followed STROBE principles for observational studies [29].

Participants and data cleaning

Students were considered for the primary analytic sample if they attended the 2025 field screening, had an age-grade combination compatible with the recorded school stage, and had bilateral non-cycloplegic screening refraction sufficient to calculate interocular spherical-equivalent difference. These criteria were chosen because anisometropia is an interocular outcome and cannot be defined from unilateral or missing SE data. The field workflow had already excluded students with recorded ocular treatment or ocular conditions likely to affect screening refraction or keratometry, including orthokeratology wear, contact lens wear likely to affect measurement, high-concentration atropine treatment, previous ocular disease or surgery, ocular trauma, and obvious eyelid or corneal abnormalities.

After the audited de-identified export, records were excluded from the primary analytic sample for implausible age-grade combinations or missing bilateral recorded SE. Complete-case subsets were then defined separately for each model according to the exposure, outcome and covariates required for that model; variable-specific denominators are summarised in Additional file 1: Table S7. AL/CR-specific analyses excluded records with implausible absolute interocular AL/CR difference greater than 0.50 after audit, a conservative distributional data-quality rule applied only to AL/CR analyses; those records were retained for non-AL/CR analyses. The surveillance dataset did not contain a full clinical examination for amblyopia, strabismus or ocular torticollis; these factors therefore could not be used as exclusion criteria and were handled as unmeasured limitations.

Vision screening and ocular biometry

Technicians completed centralised NSES training in visual-acuity testing, non-cycloplegic autorefraction, ocular biometry and field quality control before participating in the survey. The examination team included certified optometrists and ophthalmologists from the First People’s Hospital of Nantong City. Visual-acuity testing and site setup followed the Chinese National Standard for Standard Logarithmic Visual Acuity Charts (GB 11533 − 2011) [30]. Examination rooms were arranged to maintain standardised testing distance and illumination. Non-cycloplegic autorefraction was performed without spectacle or contact-lens correction in place, using three valid readings per eye from a WSRMK-8000 autorefractor (Biobase, Shandong, China), with the arithmetic mean used for spherical and cylindrical values and repeat checks when sphere or cylinder differed by more than 0.50 D between readings. Ocular biometry and keratometry were obtained using a Swan700-series anterior-segment and ocular-biometry platform (Changsha Aiscan Medical Technology Co., Ltd., Changsha, China); scans were repeated when fixation, alignment, eyelid interference or device quality indicators were inadequate. Autorefractors and ocular-biometry devices were powered on and checked each examination day according to manufacturer guidance before student testing began.

Spherical equivalent (SE) was calculated as sphere plus cylinder/2 using the signed cylinder value. Interocular asymmetry variables were defined as absolute right-left differences. AL/CR was calculated as axial length divided by mean corneal radius. SE, axial length, AL/CR and interocular biometric differences are commonly used to describe paediatric refractive status, axial myopia and ocular asymmetry [31–36]. Because refraction was performed without cycloplegia, all SE-based outcomes were interpreted as non-cycloplegic screening phenotypes rather than cycloplegic diagnostic classifications; interocular SE difference could be misclassified if accommodative response differed between eyes.

Outcome definitions

The primary outcome was non-cycloplegic screening-defined refractive anisometropia, defined as an absolute interocular SE difference of 1.00 D or greater. A 0.50 D threshold was analysed as a sensitivity outcome. Severity categories were none (< 1.00 D), mild (1.00 to < 2.00 D), moderate (2.00 to < 3.00 D) and severe (3.00 D or greater). These categories were used to describe the distribution of posture-screening markers across anisometropia severity levels. Myopia phenotypes followed standard SE threshold logic [31]. Manifest hyperopia was considered exploratory because cycloplegia was not used.

Additional refractive-error outcomes were included as comparison phenotypes. These included any-eye, mean-SE and bilateral myopia; moderate or high myopia; high myopia; axial-length and AL/CR thresholds; any-eye manifest hyperopia; any-eye astigmatism of 1.50 D or greater; and interocular cylinder difference of 1.00 D or greater. For descriptive prevalence only, the refractive-domain count summed the recorded screening domains represented by myopia, axial-myopia, astigmatism and manifest-hyperopia indicators. The astigmatism and interocular-cylinder outcomes were operational screening descriptors used to assess whether posture-screening associations extended beyond SE-defined anisometropia to other refractive-error patterns. They did not define the primary endpoint.

Trunk and postural asymmetry screening

Trunk and postural asymmetry variables were taken from routine school posture-screening fields. Trained school-health examiners completed standardised standing inspection and forward-bend testing as part of the surveillance workflow. Standing inspection recorded visible shoulder, scapular, waist, iliac-crest and spine-line asymmetry. Forward-bend screening recorded back asymmetry, rib hump or muscle contracture when present, and ATR was recorded separately for thoracic, thoracolumbar and lumbar regions. These fields followed the structure of the school posture-screening form and the screening framework described in Chinese and international school-age scoliosis-screening literature [21–23].

Maximum ATR was defined as the largest recorded ATR across thoracic, thoracolumbar and lumbar measurements. ATR of 5 degrees or greater and ATR of 7 degrees or greater were evaluated as screening thresholds rather than diagnostic cut-offs. Inspection-positive postural asymmetry was a rule-based composite: any recorded shoulder, scapular, waist, iliac-crest or spine-line asymmetry; any recorded forward-bend back asymmetry, rib hump or muscle contracture; or any thoracic, thoracolumbar or lumbar segment-direction abnormality was classified as positive. Movement-screen positivity was taken from the recorded abnormal/positive movement-screen field. These operational definitions are summarised in Additional file 1: Table S8.

For exploratory direction-specific analyses, source records were grouped as no side-bending record, right-low/left-high or left-low/right-high within the thoracic, thoracolumbar and lumbar regions. Direction-specific models compared each directional category with the corresponding no side-bending category from the same segment. These variables were analysed as exploratory surveillance markers rather than direct measures of habitual posture or spinal diagnosis.

Statistical analysis

Continuous variables were summarised as mean and standard deviation or median and interquartile range, and categorical variables as n (%). Models used complete-case analysis; no imputation was performed because missingness varied by measurement domain and the primary endpoint had a near-complete denominator. Adjusted prevalence ratios were estimated using modified Poisson regression with robust sandwich standard errors [37]. Continuous outcomes were analysed using linear regression with robust standard errors. Logistic odds ratios were generated for severity-specific analyses using the non-anisometropia group as the reference.

The main adjusted model included age, sex, BMI and fixed effects for the sampled school-class cluster. In the analytic dataset, the anonymised school-class code represented this sampled cluster. Because each selected school contributed one sampled class, this fixed-effect term adjusted for differences between sampled school-class clusters, including school context, grade composition, screening workflow and background health-risk structure, but it could not separate school-level from class-level variation. Sampling weights were not applied because the de-identified analytic export did not retain the complete school-level class sampling frame or class-selection probabilities. The estimates should therefore be interpreted as unweighted surveillance associations within the sampled school-class clusters. Cluster-robust standard errors by the same sampled cluster were examined in sensitivity analysis. Each trunk or postural asymmetry exposure was entered into adjusted models one at a time because the markers were correlated components of the same screening examination. The primary outcome was anisometropia of 1.00 D or greater, with maximum ATR per 1 degree treated as the primary continuous screening exposure and ATR of 7 degrees or greater as a higher-threshold screening contrast. Mild, moderate and severe anisometropia were secondary severity analyses, and anisometropia of 3.00 D or greater was evaluated as an exploratory high-burden endpoint. Segment-specific ATR, segment-direction and school-stage-stratified models were exploratory.

Planned sensitivity analyses evaluated the robustness of this model hierarchy. The first sensitivity model removed sampled school-class cluster fixed effects and adjusted only for age, sex and BMI. The second sensitivity model extended the main model by adding systolic blood pressure, diastolic blood pressure, mean heart rate and best recorded vital capacity, treating these as broader school-health indicators rather than primary confounders. The third sensitivity model added height to age, sex, BMI and sampled school-class cluster fixed effects, allowing linear growth and relative weight status to be considered simultaneously. An additional cluster-robust sensitivity analysis for core exposure-outcome pairs used robust standard errors clustered by the sampled school-class cluster. Height and weight from routine surveillance were used to calculate BMI. Mean heart rate was calculated from the two recorded heart-rate readings, and best vital capacity was defined as the larger of two recorded vital-capacity values.

Formal multiple-testing correction was not applied because the analysis followed a hierarchy defined for this study: primary anisometropia outcome, secondary severity outcomes, exploratory severe, school-stage-stratified, segment-specific and direction-specific outcomes, and refractive comparison outcomes. Secondary and exploratory analyses were interpreted descriptively using effect sizes, 95% confidence intervals and p values, with emphasis placed on consistency across related models rather than isolated p values. These analyses were not treated as confirmatory tests. Sparse threshold and segment-direction estimates were reviewed for instability, and separation-like estimates in very small cells were suppressed in supplementary reporting. Analyses were performed in Python with pandas, scipy, statsmodels, matplotlib and seaborn.

Use of AI-assisted tools.

AI-assisted tools were used during revision to support language editing and restructuring of manuscript text. They were not used to generate or analyse study data, perform statistical analyses or determine scientific conclusions. All AI-assisted output was checked against the source data and literature, revised by the authors and approved by all authors, who take full responsibility for the final content.

Results

Study population and surveillance burden

The primary analytic sample included 2605 students after one age-grade exclusion and two exclusions for missing bilateral recorded SE (Fig. 1). Girls accounted for 1367 students (52.5%), the mean age was 12.38 years and the mean BMI was 20.90 kg/m². The sample included 1238 primary-school students, 526 junior high school students and 841 senior high school students (Table 1). Anisometropia of 1.00 D or greater was present in 190/1238 primary-school students (15.3%), 171/526 junior high school students (32.5%) and 352/841 senior high school students (41.9%).

Fig. 1.

Participant flow. Flow of students from invitation and fieldwork attendance to the primary analytic sample

Table 1.

Participant characteristics and refractive phenotype distribution

Characteristic	Overall	Denominator
Students, n	2605	2605
Girls	1367 (52.5%)	2605
Age, years, mean ± SD	12.38 ± 3.70	2605
BMI, kg/m², mean ± SD	20.90 ± 5.22	2605
Primary	1238 (47.5%)	2605
Junior high school	526 (20.2%)	2605
Senior high school	841 (32.3%)	2605
Any-eye myopia (SE ≤ -0.50 D)	1979 (76.0%)	2605
Mean-SE myopia (mean SE ≤ -0.50 D)	1820 (69.9%)	2605
Bilateral myopia	1603 (61.5%)	2605
Moderate/high myopia (any eye SE ≤ -3.00 D)	927 (35.6%)	2605
High myopia (any eye SE ≤ -6.00 D)	208 (8.0%)	2605
Any-eye manifest hyperopia ≥ + 2.00 D	40 (1.5%)	2605
Any-eye astigmatism ≥ 1.50 D	814 (31.2%)	2605
Anisometropia ΔSE ≥ 1.00 D	713 (27.4%)	2605
Anisometropia ΔSE ≥ 0.50 D	1410 (54.1%)	2605
Aniso-astigmatism Δcylinder ≥ 1.00 D	264 (10.1%)	2605
Any-eye AL ≥ 24.0 mm	1523 (58.5%)	2605
Any-eye AL ≥ 26.0 mm	295 (11.3%)	2605
Any-eye AL/CR ≥ 3.0	1712 (66.2%)	2588
Any refractive error domain	2086 (80.1%)	2605
Refractive domain count, 0	519 (19.9%)	2605
Refractive domain count, 1	929 (35.7%)	2605
Refractive domain count, 2	855 (32.8%)	2605
Refractive domain count, ≥3	302 (11.6%)	2605
Inspection-positive postural asymmetry	241 (9.3%)	2605
Maximum ATR ≥ 5°	214 (8.2%)	2605
Maximum ATR ≥ 7°	90 (3.5%)	2605
Movement-screen positivity	226 (8.7%)	2605
Thoracic ATR ≥ 5°	44 (1.7%)	2605
Thoracic ATR ≥ 7°	9 (0.3%)	2605
Thoracolumbar ATR ≥ 5°	131 (5.0%)	2605
Thoracolumbar ATR ≥ 7°	41 (1.6%)	2605
Lumbar ATR ≥ 5°	143 (5.5%)	2605
Lumbar ATR ≥ 7°	62 (2.4%)	2605
Maximum ATR, degrees, median (IQR)	0.00 (0.00, 1.00)	2605
Maximum ATR, degrees, mean ± SD	1.01 ± 2.03	2605
Thoracic ATR, degrees, median (IQR)	0.00 (0.00, 0.00)	2605
Thoracolumbar ATR, degrees, median (IQR)	0.00 (0.00, 0.00)	2605
Lumbar ATR, degrees, median (IQR)	0.00 (0.00, 0.00)	2605

Note. Data are n (%) unless otherwise specified. Refraction was non-cycloplegic. Abbreviations: AL, axial length; AL/CR, axial length/corneal radius ratio; ATR, angle of trunk rotation; BMI, body mass index; D, dioptre; IQR, interquartile range; SD, standard deviation; SE, spherical equivalent

Anisometropia of 1.00 D or greater was present in 713 students (27.4%). Severity categories comprised 1892 students without anisometropia, 474 mild cases, 158 moderate cases and 81 severe cases. School-detected trunk and postural asymmetry markers included inspection-positive postural asymmetry in 241 students (9.3%), maximum ATR of 5 degrees or greater in 214 (8.2%), maximum ATR of 7 degrees or greater in 90 (3.5%) and movement-screen positivity in 226 (8.7%). Refractive phenotype prevalence is summarised in Fig. 2.

Fig. 2.

Refractive phenotype prevalence. Prevalence of anisometropia, myopia, axial-myopia markers, astigmatism and manifest hyperopia in the primary analytic sample

Ocular-biometric characterisation of anisometropia

Ocular biometry showed a graded interocular asymmetry pattern across anisometropia severity (Fig. 3). Median interocular axial-length difference was 0.13 mm in students without anisometropia, 0.49 mm in mild anisometropia, 0.93 mm in moderate anisometropia and 1.40 mm in severe anisometropia. Median interocular vitreous-chamber-depth difference showed a nearly parallel pattern, increasing from 0.13 mm to 0.48 mm, 0.91 mm and 1.38 mm across the same severity categories.

Fig. 3.

Interocular biometry by anisometropia severity. Distributions of interocular axial-length and vitreous-depth differences across anisometropia severity categories; metric-specific denominators differ because complete-case availability varied by biometric measure

Interocular SE difference correlated most strongly with interocular axial-length difference (Spearman rho = 0.682), interocular vitreous-chamber-depth difference (rho = 0.669) and audited interocular AL/CR difference (rho = 0.543) (Fig. 4). Figure 4 shows these key correlations between interocular refractive and biometric differences.

Fig. 4.

Key correlations among interocular asymmetry measures. Spearman correlations between interocular spherical-equivalent difference and the three strongest interocular biometric asymmetry variables. Smooth lines, where shown, are visual aids rather than formal nonlinear model estimates

Trunk and postural asymmetry markers and anisometropia

The unadjusted prevalence of anisometropia of 1.00 D or greater was 34.4% versus 26.6% for students with and without inspection-positive postural asymmetry, 34.1% versus 26.8% for ATR of 5 degrees or greater, 40.0% versus 26.9% for ATR of 7 degrees or greater, and 35.0% versus 26.6% for movement-screen positivity.

After adjustment for age, sex, BMI and sampled school-class cluster fixed effects, the estimate for maximum ATR per 1 degree was 1.04 (95% CI 1.01 to 1.07; p = 0.012). The higher-threshold ATR contrast was directionally similar but less precise: ATR of 7 degrees or greater was associated with an adjusted PR of 1.30 (95% CI 1.01 to 1.67; p = 0.043). The adjusted estimates were 1.16 for inspection-positive postural asymmetry (95% CI 0.96 to 1.40), 1.16 for ATR of 5 degrees or greater (95% CI 0.95 to 1.41) and 1.18 for movement-screen positivity (95% CI 0.98 to 1.43) (Table 2; Fig. 5).

Table 2.

Key adjusted associations of trunk and postural asymmetry markers across refractive phenotypes. Panel a. Interocular asymmetry outcomes. Panel b. Myopia and axial-myopia comparison outcomes

Exposure	Anisometropia ΔSE ≥ 1.00 D (713/2605), aPR	ΔSE, β	ΔAL, β	ΔVT, β
Inspection-positive postural asymmetry	1.16 (0.96 to 1.40); p = 0.125	0.09 (-0.03 to 0.22); p = 0.141	0.038 (-0.020 to 0.095); p = 0.198	0.039 (-0.016 to 0.095); p = 0.164
Maximum ATR ≥ 5°	1.16 (0.95 to 1.41); p = 0.148	0.08 (-0.05 to 0.21); p = 0.237	0.038 (-0.022 to 0.097); p = 0.217	0.040 (-0.018 to 0.097); p = 0.179
Maximum ATR ≥ 7°	1.30 (1.01 to 1.67); p = 0.043	0.13 (-0.07 to 0.32); p = 0.201	0.054 (-0.027 to 0.134); p = 0.189	0.045 (-0.031 to 0.120); p = 0.248
Movement-screen positivity	1.18 (0.98 to 1.43); p = 0.082	0.11 (-0.02 to 0.24); p = 0.094	0.048 (-0.011 to 0.107); p = 0.113	0.049 (-0.008 to 0.107); p = 0.094
Maximum ATR per 1°	1.04 (1.01 to 1.07); p = 0.012	0.01 (-0.01 to 0.03); p = 0.156	0.005 (-0.004 to 0.013); p = 0.260	0.005 (-0.003 to 0.013); p = 0.228

Exposure	Any-eye myopia (1979/2605), aPR	High myopia (208/2605), aPR	Any-eye AL ≥ 24.0 mm (1523/2605), aPR	Any-eye AL ≥ 26.0 mm (295/2605), aPR	Any-eye AL/CR ≥ 3.0 (1712/2588), aPR
Inspection-positive postural asymmetry	1.00 (0.95 to 1.05); p = 0.954	0.87 (0.56 to 1.36); p = 0.548	0.98 (0.91 to 1.06); p = 0.673	0.83 (0.58 to 1.18); p = 0.295	0.98 (0.91 to 1.04); p = 0.469
Maximum ATR ≥ 5°	0.99 (0.93 to 1.05); p = 0.754	0.97 (0.62 to 1.53); p = 0.900	0.97 (0.89 to 1.05); p = 0.426	0.90 (0.63 to 1.29); p = 0.559	0.97 (0.90 to 1.04); p = 0.431
Maximum ATR ≥ 7°	1.05 (0.98 to 1.13); p = 0.174	0.32 (0.10 to 0.96); p = 0.043	1.00 (0.89 to 1.14); p = 0.948	0.50 (0.23 to 1.06); p = 0.070	0.95 (0.86 to 1.06); p = 0.349
Movement-screen positivity	0.99 (0.94 to 1.05); p = 0.799	0.96 (0.61 to 1.49); p = 0.839	0.98 (0.90 to 1.06); p = 0.612	0.90 (0.63 to 1.28); p = 0.554	0.97 (0.90 to 1.04); p = 0.369
Maximum ATR per 1°	1.00 (0.99 to 1.01); p = 0.690	0.95 (0.88 to 1.03); p = 0.196	1.00 (0.98 to 1.01); p = 0.631	0.97 (0.91 to 1.02); p = 0.235	1.00 (0.99 to 1.01); p = 0.820

Note. Binary outcomes are adjusted prevalence ratios; continuous outcomes are beta coefficients. Models adjusted for age, sex, BMI and sampled school-class cluster fixed effects. Delta denotes an absolute interocular difference. Abbreviations: AL axial length, AL/CR axial length/corneal radius ratio, aPR adjusted prevalence ratio, ATR angle of trunk rotation, BMI body mass index, CI confidence interval, D dioptre, SE spherical equivalent, VT vitreous chamber depth

Fig. 5.

Screening associations across refractive phenotypes. Panel a shows adjusted prevalence ratios for trunk and postural asymmetry markers for the primary anisometropia endpoint. Panel b shows PR-type binary outcome models for maximum ATR per 1 degree; threshold comparison estimates are provided in Table 2. Asterisks mark estimates for which the 95% CI excluded 1.0

Exploratory school-stage-stratified models are shown in Additional file 1: Tables S1 and S10. Primary-school students had the lowest baseline anisometropia prevalence but showed positive adjusted estimates for maximum ATR per 1 degree and ATR of 7 degrees or greater. Senior high school students had the highest anisometropia prevalence and also showed positive adjusted patterns. Junior high school students had a high baseline anisometropia prevalence (171/526, 32.5%), but the same posture-screening markers did not show positive adjusted associations; for ATR of 7 degrees or greater, the exposed junior-high stratum included 30 students and 7 anisometropia cases. These stratified results were descriptive, and no formal interaction test was used to claim stage modification.

Severity, segment and direction analyses

Severity-stratified descriptive rates are shown in Additional file 1: Table S9, and severity-specific models are shown in Table S2. Posture-screening positivity was not monotonic across severity categories: ATR of 7 degrees or greater was present in 54/1892 students without anisometropia (2.9%), 25/474 with mild anisometropia (5.3%), 5/158 with moderate anisometropia (3.2%) and 6/81 with severe anisometropia (7.4%). Inspection-positive postural asymmetry showed a similar descriptive pattern of higher positivity in mild and severe categories but not in the moderate category.

Exploratory segment-specific and direction-specific analyses are shown in Additional file 1: Tables S3-S5 and in Table 3. Regional ATR estimates and directional contrasts varied by segment, and several directional strata were small. These analyses were exploratory.

Table 3.

Exploratory segment-direction models for anisometropia of 1.00 D or greater

Segment	Direction	Directional group ΔSE ≥ 1.00 D	Reference no side-bending record ΔSE ≥ 1.00 D	Adjusted PR versus reference	p value
Thoracic	Right-low/left-high	5/19 (26.3%)	696/2559 (27.2%)	1.03 (0.54 to 1.98)	0.927
Thoracic	Left-low/right-high	12/27 (44.4%)	696/2559 (27.2%)	1.48 (0.96 to 2.31)	0.079
Thoracolumbar	Right-low/left-high	20/58 (34.5%)	670/2469 (27.1%)	1.27 (0.89 to 1.82)	0.180
Thoracolumbar	Left-low/right-high	23/78 (29.5%)	670/2469 (27.1%)	0.98 (0.70 to 1.39)	0.926
Lumbar	Right-low/left-high	34/81 (42.0%)	658/2457 (26.8%)	1.32 (1.02 to 1.70)	0.035
Lumbar	Left-low/right-high	21/67 (31.3%)	658/2457 (26.8%)	0.96 (0.68 to 1.37)	0.838

Note. Direction-specific models compared each right-low/left-high or left-low/right-high segment record with the corresponding no side-bending record from the same segment. Models adjusted for age, sex, BMI and sampled school-class cluster fixed effects. These estimates are exploratory and should not be interpreted as segment-specific or side-specific mechanistic evidence. Abbreviations: ATR angle of trunk rotation, BMI body mass index, CI confidence interval, D dioptre, PR prevalence ratio, SE spherical equivalent

The most visible direction-specific contrast was lumbar right-low/left-high, but the direction and lower-side laterality patterns were not consistent across segments.

Among students with anisometropia and directional records, concordance between the recorded lower trunk side and the more myopic or longer-axial-length eye varied by segment and was close to or below chance in several comparisons (Fig. 6b; Additional file 1: Table S4).

Fig. 6.

Exploratory segment direction and lower-side ocular laterality. Panel a shows anatomical segment-direction examples for thoracic, thoracolumbar and lumbar right-low/left-high and left-low/right-high records, with anisometropia prevalence, reference prevalence and adjusted prevalence ratios. Panel b summarises whether the recorded lower side was concordant with the more myopic eye or longer-axial-length eye among directional cases with anisometropia. Spine drawings are original schematic line illustrations created for this figure to represent recorded school-screening segment and direction categories; they are not diagnostic imaging and do not imply a deterministic ocular pathway

Refractive comparison and sensitivity analyses

For maximum ATR per 1 degree, adjusted estimates were close to null for any-eye myopia (aPR 1.00, 95% CI 0.99 to 1.01) and high myopia (aPR 0.95, 95% CI 0.88 to 1.03). Estimates were also close to null for the axial-length thresholds: any-eye axial length of 24.0 mm or greater (aPR 1.00, 95% CI 0.98 to 1.01) and any-eye axial length of 26.0 mm or greater (aPR 0.97, 95% CI 0.91 to 1.02). The estimate for any-eye AL/CR of 3.0 or greater was likewise close to null (aPR 1.00, 95% CI 0.99 to 1.01). The same was true for astigmatism of 1.50 D or greater (aPR 0.98, 95% CI 0.95 to 1.01) and manifest hyperopia of + 2.00 D or greater (aPR 0.96, 95% CI 0.80 to 1.15) (Table 2; Fig. 5). For ATR of 7 degrees or greater, an isolated inverse estimate was observed for high myopia (aPR 0.32, 95% CI 0.10 to 0.96; p = 0.043), whereas maximum ATR per 1 degree for high myopia was close to null.

Associations with continuous interocular-asymmetry outcomes were small and imprecise for maximum ATR per 1-degree increment. At the lower anisometropia threshold of 0.50 D, the maximum-ATR estimate was 1.01 (95% CI 0.99 to 1.02). Planned sensitivity models are shown in Additional file 1: Tables S1 and S6. When sampled school-class cluster fixed effects were removed, the maximum-ATR estimate was 1.04 (95% CI 1.01 to 1.07) and the estimate for ATR of 7 degrees or greater was 1.35 (95% CI 1.05 to 1.74). Further adjustment for systolic blood pressure, diastolic blood pressure, mean heart rate and best vital capacity gave corresponding estimates of 1.04 (95% CI 1.01 to 1.07) and 1.32 (95% CI 1.02 to 1.70). Additional adjustment for height with BMI gave corresponding estimates of 1.04 (95% CI 1.01 to 1.07) and 1.29 (95% CI 1.00 to 1.66). Cluster-robust models for core exposure-outcome pairs gave a similar maximum-ATR estimate for the primary endpoint (aPR 1.04, 95% CI 1.01 to 1.07) and the same ATR-threshold point estimate with a wider interval (aPR 1.30, 95% CI 0.98 to 1.72) (Additional file 1: Table S6).

Discussion

This cross-sectional school-surveillance study found a modest association between maximum ATR and concurrent non-cycloplegic screening-defined refractive anisometropia, with directionally similar but less precise estimates for other posture-screening markers. Each 1-degree increment in maximum ATR was associated with a 4% higher adjusted prevalence of anisometropia of 1.00 D or greater (aPR 1.04, 95% CI 1.01 to 1.07), with little change across planned sensitivity models. ATR of 7 degrees or greater provided a complementary contrast: anisometropia prevalence was 40.0% among students at or above the 7-degree threshold and 26.9% among those below it, while the cluster-robust interval around the adjusted estimate included the null. Other posture markers showed positive but imprecise estimates, whereas broad myopia and axial-myopia markers showed no corresponding pattern. Together, the findings show how linking posture records with bilateral vision data can reveal between-eye burden patterns that child-level myopia indicators may miss.

Anisometropia merits separate attention in school health because it is an interocular imbalance rather than simply a more severe form of overall myopia, with implications for binocular function, stereopsis, amblyopia risk and visual fatigue [9–11]. In this sample, 713 of 2605 students (27.4%) met the 1.00 D threshold. This burden was close to earlier Nantong data, although comparisons across studies depend on age, definition, cycloplegia and population structure [12–17]. Ocular biometry showed meaningful between-eye variation: median interocular axial-length difference increased from 0.13 mm without anisometropia to 1.40 mm in severe anisometropia, with an almost parallel vitreous-chamber-depth gradient, and interocular SE difference correlated most strongly with these measures. Together, these biometric patterns reinforce the relevance of bilateral refractive outcomes in school surveillance [38, 39].

Anisometropia and child-level myopia describe different dimensions of refractive status. A child-level myopia label can encompass either little or marked interocular difference, while mean SE does not show how refractive error is distributed between the eyes [9, 31]. In this study, maximum ATR was modestly associated with anisometropia, but estimates for any-eye myopia, high myopia and axial-myopia markers were close to null. This contrast suggests that the association was not simply a feature of greater overall myopic burden and that between-eye imbalance may be the more relevant dimension. Because the comparison outcomes differed in prevalence and precision, and the continuous interocular and 0.50 D estimates were small or imprecise, this interpretation remains provisional. Even so, retaining bilateral eye-level data in school surveillance can reveal patterns that aggregate myopia categories may miss and support focused longitudinal investigation.

Prior evidence gives a plausible reason to examine the posture-vision relationship, but does not point in one causal direction. Charman proposed that short working distances and asymmetric head positions may alter retinal defocus during near work [18]. A recent school-based photogrammetric study using cycloplegic refraction reported associations of head tilt and poor writing posture with anisometropia [19], whereas another Nantong school study found no direct association between refractive errors and scoliosis-screening status [20]. Differences in posture construct, refractive assessment and age range mean that these results are not directly interchangeable. Together, they remain compatible with three explanations: habitual head tilt or asymmetric viewing distance could expose the eyes to different blur and accommodative demands; refractive imbalance or binocular discomfort could produce compensatory posture; or both could arise from shared visual-behavioural and developmental circumstances. Near work, digital-device use and time outdoors may form part of that context [40–43], while parental myopia may shape refractive susceptibility; writing posture, spectacle use and binocular-vision function were additional unmeasured factors. A cross-sectional snapshot cannot distinguish these pathways, but defining them clarifies the measurements needed in longitudinal research.

A methodological contribution is the linkage of several posture-screening fields with bilateral vision data recorded in the same school-health workflow. The source records included continuous and thresholded maximum ATR, visible inspection and movement findings, and segment-direction fields, providing more detail than a self-reported or single binary posture variable. Their operational definitions are specified in the Methods and Additional file 1: Table S8. These are surface screening markers rather than radiographic scoliosis measures, objective three-dimensional posture or cumulative exposure to habitual near-work posture [21–23]; examiner reading, positioning, cooperation, transient muscle tension and functional posture may affect classification. Their strength lies in enabling scalable analysis of cross-domain co-occurrence within existing school-health systems, while their measurement boundaries limit biomechanical and diagnostic interpretation.

The severity and school-stage analyses add context to the primary association. ATR of 7 degrees or greater was present in 2.9% without anisometropia, 5.3% with mild, 3.2% with moderate and 7.4% with severe anisometropia; with only 81 students in the severe group, this non-monotonic pattern does not support a simple dose-response sequence. Baseline anisometropia prevalence increased from 15.3% in primary school to 32.5% in junior high and 41.9% in senior high school. Positive adjusted patterns appeared in primary and senior high school but not in junior high school, where the 7-degree stratum contained 30 exposed students and 7 anisometropia cases. Exposed-event counts, exposure distribution, sampled school-class composition and unmeasured behaviour may all contribute to these differences. Without formal interaction testing, the observations frame developmental heterogeneity for longitudinal study rather than establish school-stage modification.

Segment and direction fields provided limited but informative anatomical context. The clearest contrast was for lumbar right-low/left-high records, with 42.0% anisometropia prevalence and an aPR of 1.32 (95% CI 1.02 to 1.70), while other estimates varied by segment and directional cells were small. Across segments, the lower trunk side showed no consistent alignment with the more myopic or longer-axial-length eye. This absence of stable side concordance narrows the anatomical hypotheses for prespecified testing in larger samples with direct head- and body-posture measurements.

The principal public-health implication is analytic integration of information already collected in schools. WHO guidance supports coordinated school health services rather than isolated organ-specific programmes [44, 45]. Linking vision and posture records from the same children can characterise cross-domain burden patterns, make fuller use of existing data and define questions for longitudinal follow-up. The 40.0% versus 26.9% crude contrast at the 7-degree threshold illustrates a group-level burden difference and supports further programme-level evaluation of linked records; individual referral should continue to follow standard eye-care assessment. Strengths of this study include the real-world surveillance setting, linkage of both domains in 2605 students, bilateral refraction and ocular biometry, broad refractive comparison outcomes, planned sensitivity analyses and detailed posture fields. The most transferable contribution is an approach for using routine cross-domain records and bilateral outcomes to examine patterns that aggregate or child-level myopia reports may miss.

Several limitations affect causal interpretation, prevalence inference and outcome classification. The cross-sectional design provides no temporal sequence and leaves reverse causation and shared causes unresolved [29]. The sample was a selected school-class cluster extract, with one intact class from each selected school, and does not represent every class, all Nantong schoolchildren or other populations. Class-selection probabilities and numbers of eligible classes were unavailable, preventing design-weighted citywide prevalence estimation; sensitivity models cannot confer population representativeness. Refraction was non-cycloplegic, so accommodation may shift SE and misclassify interocular differences, particularly if responses differ between eyes, with an uncertain direction of bias. These are screening-defined phenotypes rather than cycloplegic diagnostic estimates [31]. Full clinical assessments of ocular motility, binocular vision, strabismus and amblyopia were also unavailable, limiting evaluation of accompanying or alternative explanations.

The exposure and secondary analyses introduce further uncertainty. Surface-measured ATR and operational posture fields do not quantify habitual head tilt, reading distance, writing posture or duration of asymmetric visual exposure. Near-work duration, screen use, outdoor time, parental myopia, spectacle history and detailed binocular-vision measures were also unavailable, leaving residual confounding. Severity, school-stage, segment, direction, multisegment and laterality analyses involved multiple comparisons and sparse cells; some isolated secondary findings, including the inverse high-myopia threshold estimate, may therefore represent chance associations and require independent replication. Extended ocular-biometry analyses used available complete cases without imputation. Future longitudinal studies should combine cycloplegic refraction with repeated interocular measurements, objective head and trunk posture, habitual near-work behaviour and binocular-function assessment, using prespecified interaction and segment-direction hypotheses and adequate exposed-event counts. Replication in other settings will determine whether the observed cross-domain co-occurrence is stable over time and transferable across school-health systems.

Conclusions

In this cross-sectional school-surveillance dataset, maximum ATR was modestly associated with concurrent refractive anisometropia, while other posture-screening markers showed directionally similar but less precise estimates and broad myopia and axial-myopia markers showed little corresponding positive pattern. The public-health value of these findings lies in showing that linkage of routinely collected posture records with bilateral vision data can characterise between-eye burden patterns that child-level myopia summaries may miss. The associations do not establish temporal order or causality and should not be used for individual prediction or posture-based ophthalmic referral. Longitudinal studies with cycloplegic refraction, objective posture and near-work measures and binocular-vision assessment are needed to determine whether this cross-domain pattern persists over time and replicates in other school-health systems.

Abbreviations

AL

Axial length

AL/CR

Axial length/corneal radius ratio

aOR

Adjusted odds ratio

aPR

Adjusted prevalence ratio

ATR

Angle of trunk rotation

BMI

Body mass index

CI

Confidence interval

D

Dioptre

SE

Spherical equivalent

VT

Vitreous chamber depth

Author contributions

Yue Zhou and Yue Guo contributed equally to this work and contributed to data curation, analysis verification, interpretation and manuscript drafting. Yaojia Xiong, Yan Zhu, Xiaobo Huang, Lele Li and Yexun Gong contributed to fieldwork coordination, data acquisition, quality control and interpretation. ZhiMin Sun and Yong Wang conceived and supervised the study. Yong Wang led the analysis plan and manuscript revision. All authors read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No. 82571194), the Natural Science Foundation of Jiangsu Province (Grant No. BK20251839), and the People’s Livelihood Directive Program of the Nantong Municipal Science and Technology Bureau (Grant No. MS2024031). The funders had no role in study design, data collection, analysis, interpretation, manuscript preparation or the decision to submit the work for publication.

Data availability

The individual-level surveillance dataset contains health information from minors and cannot be deposited publicly because unrestricted sharing is not covered by the ethics and local school-health approvals. Aggregated, non-identifiable source data supporting the main tables, figures and supplementary tables are provided as Additional file 2. A de-identified minimum analysis dataset sufficient to reproduce the main analyses is available from the corresponding author on reasonable request, subject to ethics approval, local education/health authority permission and a signed data-use agreement.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of the First People’s Hospital of Nantong City, China (approval number 2023KT122; approval date 31 July 2023). All procedures were conducted in accordance with the Declaration of Helsinki and relevant local regulations. Written informed consent was obtained from at least one guardian for participants younger than 18 years and directly from participants aged 18 years or older.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.