The Association between childhood myopia prevalence and environmental factors in China: A metaregression analysis

Authors: Zou Y, Sekar U, Read S, Davis B, Collins M.

Geographical coverage: Mainland China

Sector: Burden of disease

Sub-sector: Prevalence

Equity focus: Primary and secondary school children.

Study population: Primary and secondary school children in mainland China.

Review type: Other review

Quantitative synthesis method: Meta-analysis

Qualitative synthesis method: Not applicable

Background: Myopia prevalence is increasing worldwide, with the greatest increases in developed South-East Asian countries, including China, Japan, South Korea, Singapore and Taiwan. These increases have been attributed to various environmental factors, including greater near-work, less time outdoors and urbanisation.

Objectives: “to quantify the variability in regional myopia prevalence (among primary and secondary school children in China) that can be accounted for by recorded environmental factors”.

Main findings:

Review findings showed that annual sunshine hours are negatively associated, and mean annual temperature and population density are positively associated with myopia prevalence among school children in China,

In total 49 were included in the analysis (31 reported overall prevalence and 18 for secondary children specifically). All included studies had samples of 600+ and used standard modes of data collection and definition of myopia.  Pooled estimate of myopia prevalence, reported by the authors, was 32.88% (23.55% for primary and 60.05% for secondary). Review authors identified high level of heterogeneity across all studies and when primary and secondary school children’s studies were analysed separately.

In the univariate meta-regression, when all studies were combined, 27.97% of the variance in myopia prevalence was accounted for by cumulative annual sunshine hours. Mean annual temperature explained 24.66% of the variance, and population density accounted for 7.06%. Authors observed that seasonal changes in sunshine hours were not significantly linked to overall myopia prevalence, but they did account for 12.54% of the variance among primary school children. Furthermore, authors note that year of study publication did not have a significant association with myopia prevalence.

In the multivariate meta-regression analysis, authors found that 28.01% of the variation in myopia prevalence could be attributed to factors such as annual sunshine hours, annual temperature, and population density. Among these, only the annual sunshine hours emerged as a statistically significant predictor in the multivariate model, both overall and specifically for primary and secondary school children.

Authors found that areas with low and moderate annual sunshine hours showed a comparable prevalence of myopia (low: 41.00%; moderate: 39.64%), which was greater than the prevalence in areas with high annual sunshine hours (18.86%).

The authors note that while the exact causes of higher myopia prevalence in regions of China with less sunshine and higher population density are unknown, spending more time outdoors or being exposed to bright outdoor light may help reduce or delay the onset of myopia. They highlight the need to understand the mechanisms behind these observed relationships and their implications for lifetime myopia risk.

Methodology:

Inclusion criteria consisted of studies published from 1979-2019 that report myopia prevalence among school children in mainland China. Research papers in English and Chinese were included (since exclusion of latter might lead to bias).

Authors conducted a search through PubMed, Web of Science and China Science Periodical Databases in April 2019. The article lists the full range and combinations of search terms used. A literature search was conducted by two authors independently, and the extracted studies were compared, and any inconsistencies were resolved by consensus. Reference lists from included studies were searched to identify other relevant studies.

The quality of the studies included was evaluated using the methodological criteria for prevalence studies developed by Leboeuf-Y de and Lauritsen and Walker.

As most studies did not report environmental factors, authors accessed the Chinese National Earth System Science Data Sharing Infrastructure website to obtain past records of sunshine hours and temperature, recorded at meteorological stations for each of the study locations. Population density data was obtained from the records of the National Bureau of Statistics of China for each of the study locations.

Analyses were conducted using Stata to calculate pooled myopia prevalence estimates with 95% confidence intervals using a random-effects meta-analysis. The I2 statistic tested for heterogeneity across studies. Univariate meta-regression examined sources of heterogeneity in myopia prevalence estimates, analysing factors like annual sunshine hours, annual temperature, population density, and seasonal sunshine variation. These factors were then used in a multivariate meta-regression model to estimate overall variance. The studies were stratified based on these factors to recalculate pooled myopia prevalence estimates. The Egger test assessed potential publication bias.

Applicability/external validity: The authors compared their results, which are specific to China, with those from other regions, particularly higher-income countries like Denmark and Finland. They pointed out that the applicability of findings from one country to another may be limited due to variations in seasonal patterns within each country.

Geographic focus: Study focuses on mainland China.

Summary of quality assessment:

While the approaches to selecting and extracting the data are relatively robust, it appears that the literature search was not sufficiently thorough. It is unclear whether items published both in English or Chinese were considered for inclusion, or whether this was the case for unpublished material. It is also unclear whether relevant experts were contacted. The data analysis was generally robust; however, there is no evidence that the quality of different studies (assessed by the authors) was accounted for in the analysis. For these reasons, we have medium confidence in the results of this study.