Table 1.
A summary of prediction models for myopia.
| Article | Myopia definition | Outcome definition | Data source | Participants (age, gender) | Predictors | Prediction/analysis method | Evaluation |
|---|---|---|---|---|---|---|---|
| Zadnik et al. [34] | ≤−0.75D in the right eye (cycloplegic) | Later onset of juvenile myopia | The Orinda Longitudinal Study of Myopia (OLSM) | 554 children in grade 3 | Refractive error, CP, LP, and AL | Logistic regression model | AUC = 0.893 |
| Jones et al. [35] | ≤−0.75D in the right eye (cycloplegic) | Later onset of myopia | The OLSM | 514 children in grade 3 | SE, AL, CP, sports/outdoor activity hours, number of myopic parents | Logistic regression model | AUC = 0.90 |
| Jones et al. [46] | ≤−0.75D in the right eye (cycloplegic) | Onset of myopia from grade 2–8 | The Collaborative Longitudinal Evaluation of Ethnicity and Refractive Error (CLEERE) Study | 1854 children in grade 1 | SE at grade 1, number of myopic parents | Discrete-time survival model | Sensitivity of 62.5%; specificity of 81.9% |
| Zhang et al. [47] | <−0.75D in either eye (cycloplegic) | 3-year incident myopia | Study cohort from Xiamen and Singapore | 236 Xiamen children (mean 7.82 years); 1979 Singapore children (mean 7.83 years) | Gender, height, presenting VA, AL, ACD, LT, VCD, CC | Generalized estimating equations | AUC = 0.974 in Xiamen; 0.815 in Singaporea |
| French et al. [20] | <−0.50D in either eye (cycloplegic) | Later onset of myopia | The Sydney adolescent Vascular and Eye Study (SAVES) | 2059 children | SE, time outdoors, parental myopia, ethnicity, near work | Logistic regression model | AUC = 0.89 |
| Zadnik et al. [37] | ≤−0.75D in the right eye (cycloplegic) | Onset of myopia from grade 2–8 | The CLEERE Study | 4512 children in grade 1–6 (6–11 years, 49.6% female) | Age, gender, ethnicity, SE, parental myopia, AL, CP, LP, AC/A ratio, horizontal/vertical astigmatism magnitude, and VA. | Discrete-time survival analysis | AUC = 0.84–0.89 for younger cohort; =0.88 for older cohort |
| Chua et al. [38] | High myopia: ≤−5D in the right eye (cycloplegic) | Onset of high myopia | The Singapore Cohort Of the Risk factors for Myopia (SCORM) | 928 children (7–9 years) | age at myopia onset, gender, race, school, parental myopia | Linear regression model | AUC = 0.87 |
| Chen et al. [49] | High myopia: ≤−6D (cycloplegic) | Onset of high myopia before the age of 15 years in GTES | The Guangzhou Refractive Error Study in Children (RESC) and the Guangzhou Twin Eye Study (GTES) | 4218 children (5–15 years) and 354 first-born twins (7–15 years) | Reference centile curves for refraction based on the RESC data | Quantile regression model | The 5th centile was most effectively (sensitivity 92.9%, specificity 97.9%, positive predictive value 65.0%) |
| Ma et al. [45] | ≤−0.50D in the right eye (cycloplegic) | 2-year and 4-year incident myopia | A 4-year school-based cohort study in Baoshan, Shanghai, China | 1856 children in grade 1–3 (mean 8.1 years, 47.6% female) | Age, gender, baseline SE, parental myopia, AL/CR ratio | Multiple linear regression model | AUC = 0.880 (2 year) and 0.861 (4 year) |
| Ghorbani Mojarrad et al. [27] | ≤−1.00D (non-cycloplegic) | Incident myopia | The UK Biobank and the Avon Longitudinal Study of Parents and Children (ALSPAC) | 3320 children aged 7 years; 2273 children aged 15 years | Age, gender, number of myopic parents, PRS from CREAM and 23andMe Inc, and the UK Biobank | Linear mixed models, cox proportional hazards model | Inclusion of genetic risk score in the survival model improved the model fit compared to using number of myopic parents alone (p < 0.0001)a |
| Lin et al. [26] | High myopia: ≤−6D in the right eye (cycloplegic) | Presence of high myopia at 18 years; onset of high myopia over 10 years | Data from 8 ophthalmic Centres in China, and 2 datasets from population-based cohort studies | Electronic medical records of 129242 individuals (6-20 years); 17113 follow-up records of 3215 participants | Age, SE, annual progression rate | Random forest model | Onset of high myopia for 8 years: AUC = 0.852–0.888 in internal validation; 0.802-0.886 in external validation, 0.752-0.869 in multi-resource testing (4 years); Onset of high myopia by the age of 18: AUC = 0.801 to 0.837 (over 8 years)a |
| Williams et al. (2018) | ≤−0.75D, mean of two eye (non-cycloplegic) | Adolescent myopia | The UK-based Twins Early Development Study | 1991 twin participants (born between 1991 and 1996) | Age, gender, maternal education, fertility treatment, summer birth, hours spent playing computer games | Logistic regression model | AUC = 0.68, R square = 4.4% |
| Tideman et al. [54] | ≤−0.50D (cycloplegic) | Incident myopia | The Generation R Study | 4734 children (6–9 years, 50.7% female) | Parental myopia, 1 or more books read per week, time spent reading, no participation in sports, non-European ethnicity, time spent outdoors, and baseline AL/CR ratio | Linear regression model | AUC = 0.78 |
| Chen et al. [22] | High myopia: ≤−6D in the right eye (cycloplegic) | Presence of high myopia at 18 years old | The GTES | 1063 children (7–15 years, 51.4% female) | Age, gender, parental SE, outdoor and near work time, PRS from CREAM | Linear mixed effect model | AUC = 0.9845 |
| Diez et al. [25] | High myopia: ≤−0.50D in the right eye (cycloplegic) | 3-year incident high myopia in 226 children | A study cohort from Wuhan, China | 12780 children (5–16 years, 48.2% female) | Axial length growth curves based on 12554 children | Analysis of variance (ANOVA); Kruskal–Wallis tests; Least significant difference (LSD) test | 9 years or older: 25th percentile had AUC = 0.876 for female and 0.781 for male; 6–9 years: 50th percentile had AUC = 0.875 for female and 0.835 for malea |
| Yang et al. [21] | Not given | Presence of myopia at grade 6 | A primary school in Henan, China | 3112 children in grade 1–6 (42.2% female) | No. of parents wearing glasses, gender, indoor and outdoor activities, AL, CC, frequency of eating white meat, red meat and eggs, close and remote adjustment ability, pulse, frequency of drinking carbonated drinks, dioptres | Support vector machine | Accuracy = 0.93, precision = 0.95, sensitivity = 0.94, f1-score = 0.94, AUC = 0.98 |
| Matsumura et al. [48] | ≤−0.5D in the worse eye (cycloplegic) | Fast subsequent 2-year myopia progression, defined based on median cut-off | The SCORM study, Singapore | 618 myopia children (7–9 years, 46.8% female) | Year 1 myopia progression, baseline SE, age of myopia onset. All models include age, gender, ethnicity, maternal education, parental myopia, near work and outdoor time | Logistic regression model | AUC = 0.77 (year 1 myopia progression), AUC = 0.70 (baseline SE), AUC = 0.65 (age of myopia onset) |
D dioptre, SE spherical equivalence, VA visual acuity, AL axial length, ACD anterior chamber depth, LP lens power, CP corneal power, CC corneal curvature, VCD vitreous chamber depth, ROC receiver-operating curve, AUC area under the ROC curve, AC/A ratio of accommodative convergence to accommodation, AL/CR axial length to corneal radius, PRS polygenic risk score, MAE mean absolute error, CNN convolution neuron network.
aAvailability of independent external validation.