| Author: | Liang, Dong |
| Title: | Astigmatism and axial length in humans |
| Advisors: | Kee, Chea-su (SO) Zhou, Yongjin (SO) |
| Degree: | Ph.D. |
| Year: | 2025 |
| Subject: | Astigmatism Eye -- Refractive errors Retina Hong Kong Polytechnic University -- Dissertations |
| Department: | School of Optometry |
| Pages: | 180 pages : color illustrations |
| Language: | English |
| Abstract: | Purpose: Astigmatism plays an important role in emmetropization and refractive development, but its impact on ocular structural changes remains largely unexplored in humans. The main purpose of this thesis was to investigate the associations between astigmatism - including its magnitude and axis orientation - and ocular axial length in humans. Specifically, the meta-analysis aimed to determine the association between baseline astigmatism and myopia development in children. Study I aimed to examine the impact of baseline astigmatism on axial elongation in school-age children. Study II aimed to characterize retinal thickness in eyes with different types of astigmatism. Study III aimed to develop prediction models for axial elongation and explore how astigmatism affects these models. Meta-Analysis - Impact of Astigmatism on Refractive Development Methods: A systematic literature search was conducted using PubMed, Embase, Cochrane Library, and Web of Science databases from inception through January 31, 2023. Longitudinal studies reporting the progression rates of spherical-equivalent refraction (SER) in children (<18 years) with different astigmatism magnitudes (lower vs. higher) or axis orientations (With-the-Rule [WTR] vs. Against-the-Rule [ATR]) were included. SER progression rates were extracted and pooled with random-effects model meta-analysis. Results: Eight studies comprising 3,964 children (WTR, n = 3,021; ATR, n = 943) were included in the meta-analysis on the impact of astigmatism axis orientation. Children with baseline ATR astigmatism (SER, -0.44 ± 0.21 D/year) exhibited significantly faster myopic progression compared to those with WTR astigmatism (SER, -0.36 ± 0.16 D/year; P = 0.008). Regarding the impact of astigmatism magnitude, ten studies involving 7,554 children (lower, n = 2,933; higher, n = 4,621) were included, and revealed no statistically significant difference in myopia development between lower (SER, -0.36 ± 0.15 D/year) and higher magnitudes of baseline astigmatism (SER, -0.32 ± 0.14 D/year; P = 0.510). Study I - Impact of Astigmatism on Axial Elongation: A Five-Year Population-Based Study Methods: Annual vision screenings were conducted at seven schools in Tianjin, China, from 2018 to 2022. Ocular biometry and non-cycloplegic autorefraction were collected. Children aged 5-16 years without any myopia interventions were included and categorized by their baseline astigmatism magnitude (Controls vs. Low vs. High) and axis orientation (WTR vs. Oblique vs. ATR). Additionally, children were classified by baseline spherical ametropia (Compound Hyperopic vs. Compound Myopic vs. Other). Annual progression rates of axial length (AXL) were calculated using regression models and compared across different types of astigmatism and spherical ametropia, adjusting for baseline age, sex, AXL, and follow-up duration as covariates. Only the right eyes were analyzed. Results: A total of 10,732 Chinese children (baseline age, 9.26 ± 2.42 years; follow-up, 2.63 ± 1.01 years; 53.2% male) were included and divided into a younger cohort (age < 11 years; n = 7,880) and an older cohort (age ≥ 11 years; n = 2,852). Across both age groups and all astigmatic magnitudes, ATR astigmatism exhibited the most rapid AXL progression, followed by oblique and WTR astigmatism. Two-way ANCOVA of the combined cohort revealed that both high-magnitude and ATR astigmatism were significantly associated with increased AXL progression (P ≤ 0.018). However, the impact of astigmatism on AXL progression varied depending on baseline spherical ametropia, as high-magnitude and ATR astigmatism increased AXL progression in compound myopic eyes but decreased progression in compound hyperopic eyes. Study II - Impact of Astigmatism on Retinal Thickness: A Case-Control Study Methods: This case-control study included 101 Chinese young adults (age, 31.67 ± 7.84 years; 38% male) from the Hong Kong Polytechnic University (PolyU) Optometry Clinic. Non-cycloplegic subjective refractions, optical coherence tomography (OCT), and best-corrected distance visual acuity (BCDVA) were collected. Participants were categorized by different types of astigmatism (WTR, n = 41; ATR, n = 25; Controls, n = 35). Inclusion criteria were ages between 18-45 years, BCDVA no worse than 0.10 LogMAR, SER ≥ -10.00 D, and astigmatism ≤ -0.75 D for WTR and ATR groups or astigmatism ≥ -0.25 D for controls. OCT-based retinal thickness and BCDVA were compared across three astigmatism groups, adjusting for age, sex, and AXL. Only the right eyes were analyzed, and groups were matched for age, sex, SER, AXL, and corneal curvature. Results: Significant differences were found across the astigmatism groups in both retinal thickness (P = 0.028) and BCDVA (P = 0.039), with thicker retina and poorer BCDVA found in eyes with WTR astigmatism. Bonferroni's post-hoc test revealed significant between-group differences in BCDVA (WTR vs. Controls, P = 0.041), as well as in retinal thickness at the inner-nasal (WTR vs. ATR, P = 0.034) and outer-temporal subfields (WTR vs. Controls, P = 0.042). BCDVA was positively associated with macular retinal thickness (r = 0.206, P = 0.041) after adjusting for age, sex, and AXL. Study III - Impact of Astigmatism on Predicting Axial Elongation: A Machine-Learning Study Methods: This study included longitudinal eye examination data from 8,296 children in the Tianjin Vision Screening Programme (Centre 1) and the PolyU Optometry Clinic (Centre 2). Inclusion criteria were ages between 5-16 years, with at least three visits over more than one year of follow-up, and no myopia control interventions. Baseline variables, including age, sex, SER, astigmatism, AXL, and previous AXL progression rate, were used to develop two machine-learning algorithms: 1) AXL-Estimator, to predict future axial length; and 2) AXL-Classifier, to identify children with progressive myopia (axial elongation ≥ 0.30 mm/year) over the next three years. Random forest algorithms with ten-fold cross-validation were employed to train the models on the training dataset (n = 5,734 from Centre 1). Independent validations were conducted on the internal validation (n = 1,419 from Centre 1) and external validation (n = 1,143 from Centre 2) datasets. The performance of the AXL-Estimator and AXL-Classifier was evaluated by mean absolute error (MAE) and the area under the ROC curve (AUROC), respectively. The impact of astigmatism on these prediction models was assessed using Gini feature importance and stratification analysis. Only the right eyes were analyzed. Results: The AXL-Estimator demonstrated high prediction accuracy across all datasets, with one-year MAEs of 0.126-0.146 mm, two-year MAEs of 0.222-0.277 mm, and three-year MAEs of 0.292-0.310 mm. The corresponding R² values ranged from 0.884 to 0.985, indicating a good fit for our models. The AXL-Classifier also yielded significant and robust diagnostic performance in differentiating progressive myopia, achieving AUROCs from 0.893 to 0.974 for up-to-three-year predictions across all datasets (all P < 0.001). While both astigmatic magnitude (6.66% to 13.10%) and axis orientation (1.85% to 3.32%) had generally relatively low feature importance in these models, stratification analysis revealed consistent and significant decreases in models' performance when predicting for eyes with astigmatism, particularly high astigmatism. Conclusions: These studies collectively highlight the significant impact of astigmatism on refractive development (meta-analysis) and axial length in humans. Both the magnitude and axis orientation of baseline astigmatism significantly influence axial elongation in school-age children (Study I), potentially leading to significant alterations in retinal thickness and visual acuity (Study II). Furthermore, while our machine learning models can accurately predict axial elongation, their performance may decrease in children with significant astigmatism (Study III). These findings not only improve our understanding of how astigmatism influences ocular structural growth but also provide valuable insights for personalized management and interventions for refractive development involving astigmatism. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13651