Development and validation of a novel nomogram for predicting the occurrence of myopia in schoolchildren: A prospective cohort study

•Myopia onset was closely related to ocular indicators.•Single spherical equivalent refraction performed poorly in predicting myopia among Chinese schoolchildren.•An online myopia risk calculator was developed to predict incident myopia in schoolchildren. PurposeMyopia is a major public health issue and occurs at young ages. Apart from its high prevalence, myopia results in high costs and irreversible blinding diseases. Accurate prediction of the risk of myopia onset is crucial for its precise prevention. We aimed to develop and validate an effective nomogram for predicting myopia onset in schoolchildren.DesignSchool-based prospective cohort study.MethodsA total of 1073 schoolchildren were enrolled from November 2014 to May 2019 in China, and were divided into the training and validation cohorts. Myopia was defined as a spherical equivalent refraction (SER) ≤−0.5 diopters. Predictors of myopia were determined through the least absolute shrinkage and selection operator regression and multivariable Cox proportional hazard model based on the training cohort. The predictive performance of the nomogram was validated internally through time-dependent receiver operating characteristic (ROC) curves, calibration plot, decision curve analysis, and Kaplan-Meier curves.ResultsIndependent predictors at baseline including gender, SER, axial length, corneal refractive power, and positive relative accommodation were included in the nomogram prediction model. This nomogram demonstrated excellent calibration, clinical net benefit, and discrimination, with all the area under the ROC curves (AUCs) between 0.74 and 0.86 in the training and validation cohorts. The Kaplan-Meier curves showed that 3 distinct risk groups stratified through X-tile analysis were well discriminated and robust among subgroups. The Harrell's C-index and net reclassification improvement demonstrated that the nomogram substantially improved compared with previous models. An online myopia risk calculator was generated for better individual prediction.ConclutionsThe nomogram provides accurate and individual prediction of myopia onset in schoolchildren. External validation is needed to verify the generalizability of this nomogram. Myopia is a major public health issue and occurs at young ages. Apart from its high prevalence, myopia results in high costs and irreversible blinding diseases. Accurate prediction of the risk of myopia onset is crucial for its precise prevention. We aimed to develop and validate an effective nomogram for predicting myopia onset in schoolchildren. School-based prospective cohort study. A total of 1073 schoolchildren were enrolled from November 2014 to May 2019 in China, and were divided into the training and validation cohorts. Myopia was defined as a spherical equivalent refraction (SER) ≤−0.5 diopters. Predictors of myopia were determined through the least absolute shrinkage and selection operator regression and multivariable Cox proportional hazard model based on the training cohort. The predictive performance of the nomogram was validated internally through time-dependent receiver operating characteristic (ROC) curves, calibration plot, decision curve analysis, and Kaplan-Meier curves. Independent predictors at baseline including gender, SER, axial length, corneal refractive power, and positive relative accommodation were included in the nomogram prediction model. This nomogram demonstrated excellent calibration, clinical net benefit, and discrimination, with all the area under the ROC curves (AUCs) between 0.74 and 0.86 in the training and validation cohorts. The Kaplan-Meier curves showed that 3 distinct risk groups stratified through X-tile analysis were well discriminated and robust among subgroups. The Harrell's C-index and net reclassification improvement demonstrated that the nomogram substantially improved compared with previous models. An online myopia risk calculator was generated for better individual prediction. The nomogram provides accurate and individual prediction of myopia onset in schoolchildren. External validation is needed to verify the generalizability of this nomogram. Myopia, the most common reason for distance vision impairment,1Repka M. Prevention of myopia in children.JAMA. 2015; 314: 1137-1139Crossref PubMed Scopus (14) Google Scholar,2Holden BA Fricke TR Wilson DA et al.Global prevalence of myopia and high myopia and temporal trends from 2000 through 2050.Ophthalmology. 2016; 123: 1036-1042Abstract Full Text Full Text PDF PubMed Scopus (2263) Google Scholar has become a major public health problem worldwide.3Dolgin E. The myopia boom.Nature. 2015; 519: 276-278Crossref PubMed Scopus (549) Google Scholar The global prevalence of myopia was approximately 22.9% in 2000 and is projected to reach 49.8% by 2050,2Holden BA Fricke TR Wilson DA et al.Global prevalence of myopia and high myopia and temporal trends from 2000 through 2050.Ophthalmology. 2016; 123: 1036-1042Abstract Full Text Full Text PDF PubMed Scopus (2263) Google Scholar and the economic burden of myopia is $202 billion per year.4Smith TS Frick KD Holden BA et al.Potential lost productivity resulting from the global burden of uncorrected refractive error.Bull World Health Organ. 2009; 87: 431-437Crossref PubMed Scopus (235) Google Scholar In East Asia, the “myopia boom” is especially severe.3Dolgin E. The myopia boom.Nature. 2015; 519: 276-278Crossref PubMed Scopus (549) Google Scholar Approximately 80% to 90% of teenagers graduating from high school are myopic in urban areas of East and Southeast Asian countries, and 10% to 20% present high myopia.5Tsai TH Liu YL Ma IH et al.Evolution of the prevalence of myopia among Taiwanese schoolchildren: a review of survey data from 1983 through 2017.Ophthalmology. 2021; 128: 290-301Abstract Full Text Full Text PDF PubMed Scopus (49) Google Scholar,6Morgan IG Ohno-Matsui K Saw SM. Myopia.Lancet. 2012; 379: 1739-1748Abstract Full Text Full Text PDF PubMed Scopus (1199) Google Scholar In China, a representative country of East Asia, only 10% to 20% of the population was myopic 60 years ago; however, close to 90% of Chinese young adults suffer from myopia today.3Dolgin E. The myopia boom.Nature. 2015; 519: 276-278Crossref PubMed Scopus (549) Google Scholar Myopia is most likely to occur during adolescence, with its incidence peaking during primary school.7Mutti DO Zadnik K Fusaro RE et al.Optical and structural development of the crystalline lens in childhood.Invest Ophthalmol Vis Sci. 1998; 39: 120-133PubMed Google Scholar Meanwhile, the prevalence of myopia tends to occur at a younger age than that in the past, which provides more time for progression toward high myopia until refraction stabilizes in the mid-20s.8Baird PN Saw SM Lanca C et al.Myopia.Nat Rev Dis Primers. 2020; 6: 99Crossref PubMed Scopus (190) Google Scholar McCullough and associates9McCullough S Adamson G Breslin KMM et al.Axial growth and refractive change in white European children and young adults: predictive factors for myopia.Sci Rep. 2020; 10: 15189Crossref PubMed Scopus (30) Google Scholar observed that a young age of myopia onset is related to faster progression. Furthermore, current evidence suggests that a substantial proportion of people with high myopia will finally develop pathologic myopia,10Choudhury F Meuer SM Klein R et al.Prevalence and characteristics of myopic degeneration in an adult Chinese American population: the Chinese American Eye Study.Am J Ophthalmol. 2018; 187: 34-42Abstract Full Text Full Text PDF PubMed Scopus (42) Google Scholar in which case there is a considerable risk of ocular complications, such as glaucoma, macular atrophy, or retinal detachment.8Baird PN Saw SM Lanca C et al.Myopia.Nat Rev Dis Primers. 2020; 6: 99Crossref PubMed Scopus (190) Google Scholar It is widely believed that myopia is a multifactorial disease, mainly caused by a large number of micro-effect genes and major environmental factors.3Dolgin E. The myopia boom.Nature. 2015; 519: 276-278Crossref PubMed Scopus (549) Google Scholar,6Morgan IG Ohno-Matsui K Saw SM. Myopia.Lancet. 2012; 379: 1739-1748Abstract Full Text Full Text PDF PubMed Scopus (1199) Google Scholar A randomized clinical trial has proved that additional outdoor time is effective in reducing the risk of myopia onset.11He M Xiang F Zeng Y et al.Effect of time spent outdoors at school on the development of myopia among children in China: a randomized clinical trial.JAMA. 2015; 314: 1142-1148Crossref PubMed Scopus (583) Google Scholar With the development of valid preventative measures, it is crucial to identify schoolchildren with a high risk of myopia in the early stage to provide a precise intervention plan. Several studies have reported that ocular variables are closely related to the onset of myopia. Gwiazda and associates reported the potential predictive power of the refractive error in childhood myopia onset.12Gwiazda J Thorn F Bauer J et al.Emmetropization and the progression of manifest refraction in children followed from infancy to puberty.Clin Vis Sci. 1993; 8: 337-344Google Scholar Afterward, the refractive error was proved to be a remarkable ocular indicator for myopia onset prediction, and the first prediction model of incident myopia was established with the spherical equivalent refraction (SER), axial length (AL), and corneal refractive power (CR) using the Orinda Longitudinal Study of Myopia (OLSM) database (OLSM-1999).13Zadnik K Mutti D Friedman N et al.Ocular predictors of the onset of juvenile myopia.Invest Ophthalmol Vis Sci. 1999; 40: 1936-1943PubMed Google Scholar Jones and associates14Jones LA Sinnott LT Mutti DO et al.Parental history of myopia, sports and outdoor activities, and future myopia.Invest Ophthalmol Vis Sci. 2007; 48: 3524-3532Crossref PubMed Scopus (535) Google Scholar further added the outdoor activity hours, number of myopic parents, and their interaction term to obtain a better myopia prediction model (OLSM-2007) with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.90. In addition, Zadnik and associates16Zadnik K Sinnott LT Cotter SA et al.Prediction of juvenile-onset myopia.JAMA Ophthalmol. 2015; 133: 683-689Crossref PubMed Scopus (154) Google Scholar demonstrated SER as a single best predictor for incident myopia and structured a prediction model that included SER, gender, and race, based on data from the Collaborative Longitudinal Evaluation of Ethnicity and Refractive Error (CLEERE) study with a C statistic ranging from 0.87 to 0.93 (CLEERE-2015). Although the discrimination of the aforementioned models seems to be high, few available myopia prediction models are well validated. Zhang and associates15Zhang M Gazzard G Fu Z et al.Validating the accuracy of a model to predict the onset of myopia in children.Invest Ophthalmol Vis Sci. 2011; 52: 5836-5841Crossref PubMed Scopus (31) Google Scholar demonstrated the predictive effect of their model with an external validation cohort. However, the practicability of the model is debatable because of the inclusion of a complex calculation formula.15Zhang M Gazzard G Fu Z et al.Validating the accuracy of a model to predict the onset of myopia in children.Invest Ophthalmol Vis Sci. 2011; 52: 5836-5841Crossref PubMed Scopus (31) Google Scholar,16Zadnik K Sinnott LT Cotter SA et al.Prediction of juvenile-onset myopia.JAMA Ophthalmol. 2015; 133: 683-689Crossref PubMed Scopus (154) Google Scholar As a simple and intuitive approach, the nomogram is increasingly popular in developing and applying to clinical prediction models.17Balachandran VP Gonen M Smith JJ et al.Nomograms in oncology: more than meets the eye.Lancet Oncol. 2015; 16: e173-e180Abstract Full Text Full Text PDF PubMed Scopus (1670) Google Scholar, 18Hou X Wang D Zuo J et al.Development and validation of a prognostic nomogram for HIV/AIDS patients who underwent antiretroviral therapy: data from a China population-based cohort.EBioMedicine. 2019; 48: 414-424Abstract Full Text Full Text PDF PubMed Scopus (20) Google Scholar, 19Liu L Li R Huang D et al.Prediction of premyopia and myopia in Chinese preschool children: a longitudinal cohort.BMC Ophthalmol. 2021; 21: 283Crossref PubMed Scopus (13) Google Scholar Using the nomogram, complex predictive models can be simplified to the probability of an event. Furthermore, a user-friendly interactive interface increases its value in clinical applications.17Balachandran VP Gonen M Smith JJ et al.Nomograms in oncology: more than meets the eye.Lancet Oncol. 2015; 16: e173-e180Abstract Full Text Full Text PDF PubMed Scopus (1670) Google Scholar,18Hou X Wang D Zuo J et al.Development and validation of a prognostic nomogram for HIV/AIDS patients who underwent antiretroviral therapy: data from a China population-based cohort.EBioMedicine. 2019; 48: 414-424Abstract Full Text Full Text PDF PubMed Scopus (20) Google Scholar Recently, the Nanjing Eye Study (NES) has developed a nomogram based on AL, axial length/corneal curvature radius (AL/CR), and the number of myopic parents for predicting pre-myopia and myopia onset in Chinese preschool children (NES-2021).19Liu L Li R Huang D et al.Prediction of premyopia and myopia in Chinese preschool children: a longitudinal cohort.BMC Ophthalmol. 2021; 21: 283Crossref PubMed Scopus (13) Google Scholar However, the effectiveness of the nomogram was not evaluated or validated, and it might not be applicable to Chinese schoolchildren because it lacks one of the key predictors, the baseline SER. Depending on the ocular and demographic variables, and environmental factors collected from the Wenzhou Medical University Essilor Progression and Onset of Myopia (WEPrOM) study, a prospective cohort of schoolchildren, we aimed to (1) identify the most valuable predictors of juvenile myopia onset; (2) generate a simple but effective nomogram score system to foresee the personalized risk of juvenile myopia onset; and (3) evaluate the discrimination, calibration, and clinical net benefit of the nomogram and further validate its performance in an independent validation set. The WEPrOM study is a school-based, 2-center, prospective follow-up study conducted from November 2014 to May 2019 in Wenzhou, a typical city in eastern China. To adjust for potential urban-rural differences, 2 representative urban and rural elementary schools were included in this study. To include the prime age for myopia onset and maximize recruitment efficiency and follow-up rates, 1118 schoolchildren from second and third grade were invited to participate in this study. Fifteen parents declined to participate (the nonresponse rate was 1.3%) and 30 children were further excluded because of a history of eye disease affecting vision (strabismus, amblyopia, or congenital glaucoma). Finally, 1073 participants were included in the WEPrOM study. Subsequently, regular follow-up visits were arranged annually until participants graduated from elementary school, apart from the last 2 visits, which were postponed 6 months because of severe flu and following winter vacation. Among the children enrolled, 207 were nearsighted at baseline, which was measured as an SER ≤–0.50 diopters (D). Furthermore, 30 of the remaining 866 children had no follow-up record because of school transfers during the follow-up process. Thus, 836 schoolchildren were included in the development and validation of the nomogram score system. The flowchart is summarized in Fig. S1. The WEPrOM study was approved by the Ethics Committee of the Eye Hospital of the Wenzhou Medical University (KYK[2013]34). The examination was performed according to the Declaration of Helsinki. Parental written informed consent was obtained from all participants before data collection. For each participant, a standardized questionnaire was applied to collect the characteristics mentioned below. Demographic variables collected for the study included age, gender, school, and grade. Daily activities included average time spent on near-work/outdoors per day. Parental variables included the number of myopic parents, parental early-onset (<12 years old) myopia, parental high-myopia (SER ≤–6.00 D), and paternal and maternal educational attainment. A team of professional investigators performed a comprehensive ocular standardized examination to obtain the corresponding ophthalmology metrics using equipment calibrated at each visit. Monocular uncorrected visual acuity and SER of both eyes, AL and CR in the right eye only, and clinical conventional binocular visual functions parameters were acquired after the examination. The clinical conventional binocular visual functions parameters acquired were the near-lateral heterophoria, gradient accommodative convergence to accommodation (AC/A) ratio, negative relative accommodation (NRA), positive relative accommodation (PRA), and the base-in (BI) and base-out (BO) break points of horizontal fusional convergence range at near. Consistent with previous studies, we selected the right-eye metrics for the prediction model after concluding that the SER of the left and right eyes were highly consistent at each visit.16Zadnik K Sinnott LT Cotter SA et al.Prediction of juvenile-onset myopia.JAMA Ophthalmol. 2015; 133: 683-689Crossref PubMed Scopus (154) Google Scholar Only baseline variables were applied to construct the prediction model. The details of the questionnaire and measurement methods were reported in Expanded Methods in the Supplementary Material. Myopia onset was defined as SER ≤˗0.50 D in the right eye at any visit after baseline, which is the most common definition of myopia worldwide.2Holden BA Fricke TR Wilson DA et al.Global prevalence of myopia and high myopia and temporal trends from 2000 through 2050.Ophthalmology. 2016; 123: 1036-1042Abstract Full Text Full Text PDF PubMed Scopus (2263) Google Scholar Specifically, SER was obtained from noncycloplegic subjective refraction by an experienced optometrist in a dimly lit room while children read the distance optotypes. The optometrist started with a working distance lens of +2.00 D maintained in the trial frame after retinoscopy, and determined the sphere and cylinder refractive errors using trial lenses, trying to relax accommodation to the greatest extent. SER was calculated as the sphere power plus half of the cylindrical. As missing values in this study were considered as missing at random, multiple imputations of the chained equation method were conducted to achieve appropriate values in the baseline characteristics with missing rates less than 30%. The effect of missing value filling was good according to the balanced analysis (Supplementary Table S1). To improve the reliability of our model and prevent data overinterpretation, we randomly divided the 836 subjects into a training set and another independent validation set at a ratio of 7:3 without replacement. The continuous variables were presented as the mean ± SD. Other categorical variables were presented as frequency (proportion). The comparability of the 2 sets was further assessed through t tests, χ2 tests, or Mann-Whitney U tests, as appropriate. The annual Kaplan-Meier cumulative incidences of myopia were described as frequency (proportion, 95% CI). The incidence density was obtained by dividing the number of myopia onset cases by the total person time during the observation period. The training set was applied for baseline variable selection and nomogram development. To minimize the potential collinearity of variables and overfitting of the model, the least absolute shrinkage and selection operator (LASSO) regression and 10-fold cross-validation were applied for predictor selection. After that, a multivariable Cox proportional hazard regression model was constructed to further select the most valuable variables (P < .05) and estimate their corresponding weights based on Efron approximation. Moreover, the multivariable Cox model for various combinations of reserved predictors was compared based on the Bayesian information criterion. Depending on the most valuable predictors, a candidate nomogram model was eventually constructed with excellent predictive performance. We further developed an online myopia risk calculator for easy access to the nomogram system.20Jalali A Alvarez-Iglesias A Roshan D et al.Visualising statistical models using dynamic nomograms.PLoS One. 2019; 14e0225253Crossref PubMed Scopus (67) Google Scholar The discrimination ability of the nomogram and independent risk factors to predict myopia onset in each follow-up year were quantified using time-dependent ROC analysis in the training set and further verified in the validation set. The calibration illustrated how close the nomogram-estimated risk is to the actual risk and was evaluated through calibration plots.17Balachandran VP Gonen M Smith JJ et al.Nomograms in oncology: more than meets the eye.Lancet Oncol. 2015; 16: e173-e180Abstract Full Text Full Text PDF PubMed Scopus (1670) Google Scholar Furthermore, the clinical usefulness, which measured whether model-assisted decisions improve subject outcomes, was also carefully assessed using decision curve analysis (DCA).21Vickers AJ Elkin EB. Decision curve analysis: a novel method for evaluating prediction models.Med Decis Making. 2006; 26: 565-574Crossref PubMed Scopus (2730) Google Scholar,22Fitzgerald Mark Saville et al.Decision curve analysis.JAMA. 2015; 313: 409-410Crossref PubMed Scopus (367) Google Scholar DCA evaluated the net benefit provided by the nomogram based on the threshold probability (ie, the probability that triggers the preventive intervention), without the need of measuring actual preferences for a particular patient. The net benefit is defined as follows:Weightingfactor=thresholdprobability/(1−thresholdprobability)(1) Netbenefit=truepositiverate−(falsepositiverate×weightingfactor)(2) The comparisons between the nomogram and other existing prediction models for myopia in children were conducted by calculating the change in the Harrell C index, continuous net reclassification improvement (continuous NRI), and integrated discrimination improvement (IDI). In addition, X-tile software (version 3.6.1; Yale University) was applied in the training set to find optimal cutoff values, which were defined as the total risk scores that produced the highest χ² values in the Mantel-Cox test and divided patients into low-, middle-, and high-risk groups.23Camp RL Dolled-Filhart M Rimm DL. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization.Clin Cancer Res. 2004; 10: 7252-7259Crossref PubMed Scopus (2524) Google Scholar The grouping effect was measured through Kaplan-Meier curves, log-rank test, and hazard ratios (HRs) in both training and validation sets. Subgroup analyses of the baseline grade and school region were performed to assess the robustness of the model. Stata/MP 15.1 for Windows (Stata Corp LLC) was applied for data management. R, version 4.0.5 (The R Foundation for Statistical Computing), was used for statistical analyses with the RStudio software (version 1.4.1717). All tests set a 2-sided P < .05 as the significant level. All baseline features of the training (N = 585) and validation (N = 251) sets showed similar characteristics as shown in Table 1. Overall, 444 (53.1%) of 836 children were male, 231 (27.6%) lived in rural areas, and 458 (54.8%) were grade 2 at baseline. The average age of the children was 7.8 ± 0.7 years at baseline. The Kaplan-Meier cumulative incidence rates of myopia over time are shown in Table 2. During the follow-up period, the incidence density of myopia in schoolchildren was 196.0 per 1000 person-years in the training cohort and 191.3 per 1000 person-years in the validation cohort. The average time to myopia onset for the training and validation cohorts was 3.37 and 3.40 years, respectively. Of 836 schoolchildren, 825 (98.7%), 805 (93.0%), 779 (90.0%), and 405 (88.4%, baseline third-grade schoolchildren had graduated) completed respective annual follow-up examination. Eleven of these children did not attend the first follow-up examination but attended subsequent ones.TABLE 1Baseline characteristics of the children in the training and validation cohorts.VariablesTotal(N = 836)Training Cohort(n = 585)Validation Cohort(n = 251)P ValueDiscrete variables Gender, n (%).727 Male444 (53.1)313 (53.5)131 (52.2) Female392 (46.9)272 (46.5)120 (47.8) School, n (%).819 Rural231 (27.6)163 (27.9)68 (27.1) Urban605 (72.4)422 (72.1)183 (72.9) Baseline grade, n (%).938 2458 (54.8)321 (54.9)137 (54.6) 3378 (45.2)264 (45.1)114 (45.4) No. of myopic parents, n (%).103 0333 (39.8)238 (40.7)95 (37.8) 1315 (37.7)229 (39.1)86 (34.3) 2188 (22.5)118 (20.2)70 (27.9) Parental high-myopia, n (%).790 Neither730 (87.3)512 (87.5)218 (86.9) Either106 (12.7)73 (12.5)33 (13.1) Parental early-onset myopia, n (%).336 Neither775 (92.7)539 (92.1)236 (94.0) Either61 (7.3)46 (7.9)15 (6.0) Paternal educational attainment, n (%).534 High school or less450 (53.8)319 (54.5)131 (52.2) College or beyond386 (46.2)266 (45.5)120 (47.8) Maternal educational attainment, n (%).209 High school or less444 (53.1)319 (54.5)125 (49.8) College or beyond392 (46.9)266 (45.5)126 (50.2)Continuous variables Age, y7.8 ± 0.77.8 ± 0.77.8 ± 0.7.813 Uncorrected visual acuity, logMAR–0.06 ± 0.09–0.06 ± 0.09–0.07 ± 0.09.081 Spherical equivalent refraction, D0.10 ± 0.390.11 ± 0.400.10 ± 0.37.676 Axial length, mm22.95 ± 0.7022.96 ± 0.7222.92 ± 0.65.438 Corneal refractive power, D43.15 ± 1.4143.11 ± 1.4443.26 ± 1.35.169 Near lateral heterophoria, PD–3.37 ± 5.08–3.21 ± 4.92–3.73 ± 5.41.177 AC/A ratio, PD/D1.76 ± 2.711.72 ± 2.711.85 ± 2.71.509 Negative relative accommodation, D2.69 ± 0.782.72 ± 0.792.63 ± 0.74.146 Positive relative accommodation, D–3.86 ± 2.06–3.92 ± 2.07–3.70 ± 2.03.152 Base-in break point, PD22.51 ± 5.8722.35 ± 5.6622.90 ± 6.32.234 Base-out break point, PD25.39 ± 6.5225.26 ± 6.2625.71 ± 7.08.378 Time spent on near-work, h/d2.4 ± 2.02.3 ± 2.02.5 ± 2.1.267 Time spent outdoors, h/d2.2 ± 1.82.3 ± 1.82.2 ± 1.7.745AC/A = accommodative convergence to accommodation, D = diopter, PD = prism diopter.Note: Results are presented as the mean ± SD unless otherwise indicated. Open table in a new tab TABLE 2The Kaplan-Meier cumulative incidence rates of myopia over time in the training and validation cohorts.VariablesTotal(N = 836)Training Cohort(n = 585)Validation Cohort(n = 251)P Value1-y cumulative incidence of myopia114 (13.6) [11.4-16.1]82 (14.0) [11.3-17.0]32 (12.7) [9.0-17.2].6242-y cumulative incidence of myopia237 (28.8) [25.7-31.9]167 (29.0) [25.4-32.8]70 (28.3) [22.8-34.0].8463.5-y cumulative incidence of myopia441 (55.1) [51.6-58.5]308 (55.2) [50.9-59.2]133 (55.0) [48.5-61.1].9284.5-y cumulative incidence of myopia496 (68.5) [64.4-72.1]348 (69.1) [64.2-73.4]148 (67.1) [59.5-73.5].888Note: Results are presented as n (%) [95% CI]. Open table in a new tab AC/A = accommodative convergence to accommodation, D = diopter, PD = prism diopter. Note: Results are presented as the mean ± SD unless otherwise indicated. Note: Results are presented as n (%) [95% CI]. Twenty-one baseline variables (Table 1) were included in the LASSO regression for the training cohort. Based on minimum (λmin) criteria (Supplementary Fig. S2), 18 variables remained candidate indicators of incident myopia, including gender, school area, baseline grade, number of myopic parents, parental high-myopia, paternal educational attainment, maternal educational attainment, average time spent on outdoors, uncorrected visual acuity, SER, AL, CR, near-lateral heterophoria, AC/A ratio, NRA, PRA, BI, and BO break points at baseline. To further screen the most valuable predictors, the indicators were incorporated into a multivariable Cox proportional hazard regression model and 5 indicators were retained as independent predictors with P <.05 (Supplementary Table S2, Fig. 1). In terms of the Bayesian information criterion, the model based on the 5 reserved predictors performed best. Multivariable Cox proportional hazard regression analysis was performed in both the training and internal validation cohorts based on 5 predictors (Fig. 1). Five predictors were significantly associated with myopia onset in both cohorts. Overall, female children (HR 1.94, 95% CI 1.59-2.37), more negative SER values (0.30, 0.20-0.47), longer AL (3.94, 3.12-4.97), larger CR (1.74, 1.56-1.93), and lower magnitude of PRA (1.10, 1.05-1.15) at baseline had a greater risk for incident myopia. As shown in Supplementary Fig. S3, the individual AUCs of SER, AL, and PRA were better than CR and gender in predicting myopia onset in each follow-up year. We finally generated a nomogram for predicting the probability of myopia in schoolchildren with 5 predictors containing gender, SER, PRA, AL, and CR in the training cohort (Fig. 3). As the ophthalmologic predictors were continuous and the graphical nomogram might not be valuable enough for practical applications, an online myopia risk calculator was further developed and was allowed free access at https://myopia-risk-calculator.shinyapps.io/weprom/ (Supplementary Fig. S4). Researchers could easily determine the predicted myopia probability of children by inputting the 5 predictors mentioned and reading the output generated by the webserver. Discrimination of the nomogram was assessed carefully. The AUCs of the nomogram in the training cohort were 0.857 for 1 year, 0.820 for 2 years, 0.787 for 3.5 years, and 0.772 for 4.5 years, whereas the AUCs in the validation cohort were 0.826, 0.788, 0.768, and 0.740, respectively (Fig. 2). The calibration plots demonstrated an excellent consistency between actual observations and nomogram predictions in both training and validation cohorts (Supplementary Figs. S5, 3). After ascertaining that the nomogram had goo