Predicting future myopia progression from prior myopia progression

Title: Annual myopia progression and subsequent year progression in Singaporean children

Authors: Noel A. Brennan(1); Saiko Matsumara(2); Hla Myint Htoon(2,3); Biten K Kathrani(1); Chuen Seng Tan(4); Carla Lanca(2); Donald Tan(3,5); Charumathi Sabanayagam(2,3); Seang-Mei Saw(2,3)

(1) Johnson & Johnson Vision, Johnson & Johnson Vision Care, Jacksonville , Florida, United States
(2) Singapore Eye Research Institute, Singapore
(3) Duke-NUS Medical School, Singapore
(4) Saw Swee Hock School of Public Health, National University of Singapore, Singapore, Singapore
(5) Singapore National Eye Centre, Singapore

Date: June 2020

Source: ARVO 2020 Abstracts - video presentation

Summary

Annual rate of myopia progression from an existing dataset is used to determine if rate of subsequent years of myopia progression can be reliably predicted, to reveal a statistically significant though only modest clinical relevance of myopia progression from one year to the next, with no improvement to the model from averaging up to three prior years of myopia progression measurement.

Clinical relevance

Reliance on prior annual rate of myopia progression alone is not a reliable enough predictor for subsequent years of progression. Consequently, myopia management strategy for a given individual should be determined based on multiple patient-specific factors including myopia progression in the previous year, higher initial SE, age of myopia onset and parental myopia.

Limitations and future research

  • Meeting abstract so not fully peer-reviewed
  • Dataset restricted to Singapore children - results don’t necessarily extrapolate to other nationalities / ethnicities
  • Gives promise of more accurate prediction models hopefully in the near future

Full story

Singapore Cohort Study of the Risk Factors for Myopia (SCORM) was modelled to ascertain whether myopia progression data from a given year could be used to predict future myopia progression.

Previous research from Donovan et al(1) shows that on average myopia progresses faster in younger children compared to older children, however Chua et al(2) reveals that there is considerable variability across individuals - some children who became myopic at young age hardly progressed over subsequent years and remained with low myopia, while others progressed to high myopia, meaning that age of myopia onset is not a good predictor of future myopia progression.

There also appears to be a perception amongst eye care practitioners towards being more likely to apply myopia management strategies to children with fast progression, as shown by Leshno et al(3), than other indicators such as any sign of myopia, age that myopia was first detected and degree of myopia.

Results for using first year to predict progression in the next year

  • Year 1 progression correlated with year 2, but r only 0.47, suggests that year 1 only accounts for about 20% of variance in year 2
  • For each D increase in year 1 there was 0.35D increase in year 2
  • There was a weak association between baseline refractive error and year 2 myopia progression
  • Age of myopia onset had no influence on year 2 progression

From their constructed model the authors also measured area under the curve to find that when other related variables were included the model revealed 73.5% sensitivity and 64% specificity 64% of using year 1 progression to predict year 2 progression, meaning that:

  • 1 in 4 fast Progressor’s in year 2 would not have been predicted by the model
  • 1 in 3 would have been falsely predicted to have fast progression.
    Using baseline refractive error, age of myopia onset (years) or parental myopia instead of annual progression at year 1 to predict myopia progression in year 2 did not improve the model.

When stratified by degree of progression (<-0.50; ≤-0.50 to >-0.75; ≤-0.75 to >-1.00; ≤-1.00 to >-1.25; ≤-1.25D) a reasonably good correlation was reported across the averaged data - e.g. average progression of ≤-0.50 to >-0.75 in year 1 lead to good correlation with average progression of around 0.62D in year 2. Despite this however, there was considerable variability across each progression group indicating that some within each group either progressed more or less in the second year than had occurred in the first year.

Results for using first year to predict subsequent years of progression
Year 1 progression was reported to correlate better to year 2 than year 3 and year 4 with the correlation reduces for increasing interval between years. The reported area under the curve measures were 0.76 for year 1 to predict year 2, 0.69 for year 1 to predict year 3, and 0.63 for year 1 to predict year 4.

Cumulative progression over prior years to predict the next year was also assessed- built these into models including other relevant data to calculate sensitivity and specificity:

  • One year prior: Sensitivity 72.9%, Specificity 59.5%
  • Two years prior: Sensitivity 72.9%, Specificity 59.6%
  • Three years prior: Sensitivity 76.3%, Specificity 52.6%
    These data reveal that when trying to predict myopia progression for the following year, no improvement in the model is gained from including more than the prior year of myopia progression.

Conclusions
Measurement of myopia progression across 1 year provides statistically significant but only modest clinically relevant prediction of myopia progression for the subsequent year. Therefore, myopia progression as a standalone factor alone is not sufficient to accurately predict likely myopia progression in the next year on an individual basis.

Patient compliance with yearly measurements is critical to determine reliable measures. Also, other factors need to be taken into consideration when considering whether to start or modify myopia treatment, including age of patient, progression in previous year, higher initial refractive error, parental myopia, as these may all play Into the following years progression.

Abstract

Purpose: To investigate predictors of myopia progression for subsequent year amongst myopic children in the Singapore Cohort Study of the Risk Factors for Myopia (SCORM).

Methods: A total of 674 myopic children (353M, 321F) aged 7 to 10 (mean 8.0 ± 0.9) years from 3 schools at baseline with at least 2 follow-up visits in SCORM were included. Cycloplegic autorefraction (RK5 autokeratorefractometer) and axial length (AL) measurement (US-800 Echo scan) were performed at every visit. Multiple linear regression analysis was performed with annual future myopia progression as the dependent variable. Receiver operating characteristic (ROC) curves from multiple logistic regressions were used to derive prediction scores for future fast myopia progression defined by the median cut at different durations of different follow up years.

Results: Myopia progression in Year 1 correlated with Year 2 progression (r = 0.47; see figure). For every 1 D increase in annual myopia progression in Year 1, Year 2 progression increased by 0.35 D (p < 0.001), in a multivariate linear regression model. Children with slow myopia progression during the first year (Year 1) (>-0.50 D/year) had the slowest mean Year 2 progression (-0.44 ± 0.44 D/Year), while children with fast myopia progression (<-1.25 D) in Year 1 had the fastest mean progression (-1.01 ± 0.39 D/year) in Year 2. There was a dose-response relationship (p for trend < 0.001). Year 1 myopia progression had the highest AUC for predicting fast Year 2 progression [0.76 (95% CI 0.73-0.80)] when compared to baseline SE [0.70 (95% CI 0.66-0.73)] or age of myopia onset [0.70 (95% CI 0.66-0.73)] or parental myopia [0.70 (95% CI 0.66-0.73)], after adjusting for confounders. For Year 1 myopia progression, AUC for predicting fast Year 2 progression was 0.76 [95% CI 0.73-0.80] and higher than those for fast Year 3 [0.69 (95% CI 0.65-0.73)] or Year 4 [0.63 (95% CI 0.57-0.68)] progression.

Conclusions: One-year annual myopia progression correlates with immediate subsequent year myopia progression. However, annual progression as a single factor cannot fully predict subsequent year or long term myopia progression. Strategic management to a given individual should be determined based on multiple patient-specific factors including myopia progression in the previous year, age of myopia onset and parental myopia.

Abstract link is here

Paul-profile-photo

About Paul

Dr Paul Gifford is a research scientist and industry innovator based in Brisbane, Australia, and co-founder of Myopia Profile.

References

  1. Donavan L, Sankaridurg P, Ho A, Naduvilath T, Smith III EL, Holden B. Myopia progression rates in urban children wearing single-vision spectacles. Optometry and Vision Science. 2012;89:27-32.
  2. Chua S, Sabanayagam C, Cheung Y-B, Chia A, Valenzuela RK, Tan D, Wong T-Y, Cheng C-Y, Saw S-M. Age of onset of myopia predicts risk of high myopia in later childhood in myopic Singapore children. Ophthalmic and Physiological Optics. 2016;36:388-94.
  3. Leshno et al Variations between respondents from different geographical regions in implementation of criteria for initiating treatment to reduce myopia progression. British Journal Ophthalmology. 2019;):1-6 doi:10.1136/bjophthalmol-2019-314752

Leave a comment