Abstract
Objective The rates of myopia (−0.50 diopter), and high myopia (≥ −6.0 diopter) have been increasing in East Asian populations; the reasons for which may include the combinations of genetic, environment and behavioural factors. The most affected demography point to the young elite population of intellectuals produced from universities. Of the several recommendations to address the myopia epidemic, the influence of Chinese eye exercises (CEE) have been examined. However, reports have been inconsistent, prompting a meta-analysis to obtain more precise estimates.
Methods Eight articles were included in the meta-analysis where we operated on the hypothesis that CEE either increased or reduced myopia control. We compared the subjects that performed CEE against those that did not. We used and estimated odds ratios [ORs] and 95% confidence intervals (CIs) using the generic inverse variance method. Subgroup analysis involved quality (high/serious and low/non-serious) and frequency (> 5 times/week) of performing CEE comprised. Heterogeneity was subjected to outlier treatment which split the findings into pre- (PRO) and post- (PSO) outlier. The strength of evidence in our findings were based on high significance (Pa < 10−5), surviving the Bonferroni correction and homogeneity (I2 = 0%). Outcomes with these features comprised our core findings.
Results Our core findings were found in the PSO overall indicating elevated myopia control (OR 0.72, 95% CI 0.61-0.86, Pa = 0.0002) and CEE subgroups (Serious: OR 0.75, 95% CI 0.68-0.84, Pa < 10−5; Frequent: OR 0.55, 95% CI 0.45-0.68, Pa < 10−5). The low quality subgroup outcome was null in PRO (OR 0.97, 95% CI 0.50-1.86, Pa = 0.92) but conveyed significantly less myopia control in PSO (OR1.57, 95% CI 1.24-2.01, Pa = 0.0002).
Conclusions This meta-analysis found that CEE afforded 28% greater control of myopia. Enabled by outlier treatment, this finding was homogeneous and consistent. Subgroup effects elevated myopia control to 62% when CEE was done up to 5 times a week. Improper CEE performance implied reduced myopia control of up to 57%.
Introduction
Myopia is one of the most commonly occurring defects of the human eye refraction. Defined as ≥ −0.50 diopter (D), this ocular defect has three levels, low (≤−3.0 D), medium (−3.0 D to −5.9 D) and high (≥ −6.0 D). About one-fifth of the myopic global population can develop high myopia (≥ −6.0 D) rendering this eye disorder a significant epidemiological concern [1,2]. This concern is highlighted with a worldwide estimate of ∼ 1.4 billion myopic individuals comprising > 20% of the global population [3]. Moreover, this percentage is projected to increase to 50% (∼ 4.7 billion) by the year 2050 [3]. This increase in myopia prevalence will have substantial social, educational, and economic consequences to modern society. While geographic distribution of myopia prevalence varies between countries, this eye defect has become epidemic in East and Southeast Asia [3] with the fastest growing tendency pointing to the countries of Singapore and China [4]. Demographic comparisons among 12-year-olds show higher prevalence of myopia in Asians (Singapore: 62%; China: 49.7%) than Caucasians (USA: 20%; Australia: 11.9%) [5,3]. In a study of Chinese schoolchildren, myopia was reported to be present among 33.6% of first graders and 54% of seventh graders [6]. Yet, these percentages pale when compared with older Chinese teenagers and young adults where myopia is present in 90% of this demography [7]. Even more serious prevalence data was shown by a Shanghai study where myopia was present among 94.9% of undergraduate students and 96.9% of postgraduate students, and 19.5% of all myopic students had high myopia [8]. Data from the Ministry of Education of the People’s Republic of China in 2017 show that there are about 27.5 million university or college students affected with myopia [9]. Age-wise, myopia currently affects people from 10 to 39 years, but this demographic age range is predicted to widen from 10 to 79 years by the year 2050 [3]. Etiology of myopia is complex with genetic and environmental factors playing a role. Associations of genetic polymorphisms with myopia have been investigated [10-12], so has the impact of family history on the incidence of myopia among children [13,14]. Even high intelligence was shown to be associated with myopia [15]. Consequently, populations with a higher level of education are reported to have higher proportions of myopia [16]. Thus, we look at the intertwining of high educational attainment with genetics as influencing myopia prevalence [17]. The complication of environmental factors affecting myopia is that they are tied to human behavior, which in the modern context, is also tethered to the influence of technology (smartphones and computers). Still, these behavioral factors and technological influence afford greater human direct control compared to genetics in addressing the myopia epidemic. A major contributory factor in the environmental context point to the ubiquitous use of smartphones and computers [7], which dictates the behaviors of those at most risk for myopia. Risk behaviors for myopia include the following: One, high educational pressure in East and Southeast Asia does not permit children to spend extended time outdoors [18,19]. Thus, less outdoor activities promotes sedentary behavior. Two, long hours of exposure to digital screens [20]. Three, continuous reading without rests, has been shown to lead to myopia [21,22]. Myopia not only impacts upon physical health, but also increases risk of eye complications, including myopic retinopathy, myopic glaucoma, retinal detachments, and blindness [23-25]. High myopia (≥ −6.0 D) can result in loss of vision due to retinal detachment, neovascularization, cataract, glaucoma, or macular atrophy [1,26]. It can also cause impaired vision and blindness [27]. Interventions have been developed to prevent and control myopia [28]. Increased time for outdoor activities could effectively prevent myopia [29-32]. In its presence, however, myopia progression is hindered with low-dose atropine and orthokeratology [33,34]. Even then, these two therapeutic approaches may only be used for myopic control, rather than prevention. Therefore, there is a need to introduce a simple and easy-to-use intervention for myopia control. Chinese eye exercises (CEE), an intervention for visual protection and myopia prevention and progression, originated from the theories of Traditional Chinese Medicine [35]. Yet, reports of CEE effects in preventing myopia have been inconsistent where studies have suggested reduced risk and others have shown risk increases. This prompted a meta-analysis to obtain a more precise estimate of association. We aim to foster better understanding of the role of CEE in myopia onset and progression.
Materials and Methods
Selection of studies
We searched MEDLINE using PubMed, Google Scholar and Science Direct for relevant publications as of October 01, 2019. Terms used were “Chinese eye exercises “, and “myopia”, as medical subject heading and text, unrestricted by language. References cited in the retrieved articles were also screened manually to identify additional eligible studies. In cases of duplicate articles, we selected the one with a later date of publication. Inclusion criteria were (1) case–control design evaluating the association between CEE and myopia; (2) eye exercise in the presence and absence of myopia and (3) presence of raw frequency data or presence of odds ratios (ORs) and 95% confidence intervals (CIs). Exclusion criteria were (1) those not involving myopia and CEE; (2) reviews; and (3) studies whose raw data were unusable/absent. Primary study authors were contacted in order to obtain more information on incomplete data.
Data extraction
Two investigators (PS and NP) independently extracted data. Disagrements were adjudicated by a third investigator (PT) until arrival at a consensus. The following information were obtained from each publication: first author’s name, published year, country of origin, age of participants, details of the myopia condition, and study design.
Data distribution
Data distribution was assessed with the Shapiro-Wilks (SW) test using SPSS 20.0 (IBM Corp., Armonk, NY, USA). Normal distribution (P > 0.05) warranted descriptive and inferential expressions of mean ± standard deviation (SD) and the parametric approach, respectively. Otherwise, the median with interquartile range (IQR) and non-parametric tests were used, respectively.
Methodological quality of the studies
We assessed the methodological quality of the included studies using two scales that were appropriate to the study design. The Jadad scale [36], for randomized control trials (RCTs) has five questions that examines randomization, double-blind and attrition. It provides a total score that ranges from zero to five, where zero indicates low quality and five, high quality [36]. We used the Newcastle-Ottawa Scale (NOS) for cross-sectional (xs) and case-control (cc) studies [37]. The ⋆- based NOS contains eight items with three dimensions (number of ⋆in parentheses): selection (4), comparability (1), and outcome (3). The NOS scores range from zero up to nine ⋆for cc studies and zero to 10 ⋆for xs studies. Low, moderate and high have scores of < 4, 4-5 and ≥ 6-7, respectively.
Risk of bias
Risk of bias assessment followed the Cochrane handbook [38], which was applied to cc and RCT study designs [39]. This method evaluates biases originating from sequence generation (selection bias), allocation sequence concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective outcome reporting (reporting bias). Every item was judged as either yes, no, or unclear. More items judged as “yes” indicated a low likelihood of bias. In contrast, high likelihood of bias garnered more “no” judgements. Insufficient descriptions merited “unclear” judgment. Judgments were assigned by two of the authors (NP and PT) working independently, and discrepancies were remedied through discussions with a third investigator (PS) to obtain a consensus.
Meta-analysis
Because we addressed the issues of myopia onset and progression, we hypothesized that performing CEE would afford greater or lesser control of myopia. Greater and lesser control entailed less and more risk, respectively. In order to quantify this control, we assessed the types of data provided by each article, which were ORs and 95% CIs or raw frequency data (number of participants who performed CEE versus those that did not). In the presence of both, we opted to extract the former. Presence of raw data prompted their extraction from each study and were used to calculate study-specific ORs with 95% CIs. The study-specific ORs were transformed logarithmically (log OR). We also derived the standard error (SE) from the 95% CI [40]. Using the generic inverse variant (GIV) method [41], log OR and SE from each paper were entered as the operating data in Review Manager (RevMan) 5.3 (Cochrane Collaboration, Oxford, England). Subgroup analysis consisted of high quality (HQ)/serious, low quality (LQ) and frequency of performing CEE. Heterogeneity between studies was estimated with the χ2-based Q test [42], with threshold of significance set at Pb < 0.10. Heterogeneity was also quantified with the I2 statistic which measures variability between studies [43]. I2 values of > 75% indicate substantial variability and 0%, zero heterogeneity (homogeneity). Evidence of functional similarities in population features of the studies warranted using the fixed-effects model [44], otherwise the random-effects model [41] was used. Sources of heterogeneity were detected with the Galbraith plot [45] followed by re-analysis (outlier treatment). Of note, outlier treatment dichotomized the comparisons into pre-outlier (PRO) and post-outlier (PSO). Sensitivity analysis, which involves omitting one study at a time and recalculating the pooled OR, was used to test for robustness of the summary effects. Publication bias was considered for assessment if the comparisons met two conditions: (i) ≥ 10 studies only [46] and (ii) significant outcomes. Multiple comparisons were Bonferroni-corrected. Except for heterogeneity estimation [42], two-sided P-values of ≤ 0.05 were considered significant. Other data were analyzed using SIGMASTAT 2.03, SIGMAPLOT 11.0 (Systat Software, San Jose, CA, USA) and SPSS 20.0 (IBM Co., Armonk, NY, USA).
Results
Search results and study features
Figure 1 outlines the study selection process in a PRISMA-sanctioned flowchart (Preferred Reporting Items for Systematic Reviews and Meta-Analyses). Initial search resulted in 60 citations, followed by a series of omissions that eventually yielded eight articles for inclusion [47-54].
Characteristics of the included studies
Table 1 shows the year span of the included articles (2011-2019) and ethnic composition, Chinese (n = 6) and Indian (n = 2). Non-normal (SW: P = 0.022) age distribution of the participants in the collection of studies showed a young demography (median: 12.7 years, IQR: 11.3-20.7). Methodological quality of the component studies was generally high for xs but probably not for RCT studies. S1 and S2 Tables detail the scores for each study based on the Jadad (RCT) and NOS (cc/xs) scales. This meta-analysis followed the PRISMA guidelines (S3 Table).
Characteristics of the included studies
Risk of bias outcomes
Although Fig 2 shows a general low risk of bias based on the six criteria, Fig 3 delineates the poor quality of the Rathod et al study [47] which contrasted with the high quality of Li et al [51] and Kang et al [48].
Risk of bias analysis of CEE influence on myopia using the Cochrane Collaboration tool
1: random sequence generation; 2: allocation concealment; 3: blinding of participants and personnel; 4: blinding of outcome assessment; 5: incomplete data outcome; 6: selective reporting
Risk of bias summary among and RCT/cc studies that examined the influence of CEE on myopia
CEE: Chinese eye exercise; RCT: randomized control trial; cc: case-control; 1: random sequence generation; 2: allocation concealment; 3: blinding of participants and personnel; 4: blinding of outcome assessment; 5: incomplete data outcome; 6: selective reporting
Meta-analysis outcomes
Table 2 shows six significant outcomes, four of which are attributed to outlier treatment. One finding indicating less myopia control was found in the LQ subgroup (OR 1.57, 95% CI 1.24-2.01, Pa = 0.0002). Otherwise, most of the findings indicate greater control of myopia. In PRO, OR magnitude was greater (ORs 0.38-0.49, 95% CIs 0.22-0.82, Pa = 0.0007-0.007) than PSO, whose findings are modulated (ORs 0.55-0.75, 95% CIs 0.45-0.86, Pa < 10−5-0.0002). Significance, however, was greater in PSO (up to Pa < 10−5) than in PRO (up to Pa = 0.0007).
Summary associations between CEE and myopia
Mechanism of outlier treatment
The mechanism of outlier treatment for the CEE analysis is visualized in Figs 4-6. Fig 4 shows the PRO forest plot, non-significant (OR 0.84, 95% CI 0.48-1.49, Pa = 0.55) and heterogeneous (Pb < 10−5, I2 = 94%). The Galbraith plot identifies five studies in three articles [50-52] as the sources of heterogeneity (outliers), located above and below the +2 and −2 confidence limits (Fig 5). In Fig 6, the PSO outcome (outliers omitted) shows eliminated heterogeneity (Pb = 0.95, I2 = 0%); greater myopia control (OR 0.72, 95% CI 0.61-0.86) and escalated significance (Pa = 0.0002). This operation is numerically summarized in Table 2.
Forest plot outcome in the CEE analysis
O: onset; P: progression; Diamond denotes the pooled odds ratio (OR), here indicating more myopia control (0.84). Squares indicate the OR in each study. Horizontal lines on either side of each square represent the 95% confidence intervals (CI). The Z test for overall effect is non-significant (Pa = 0.55). The chi-square test shows the presence of heterogeneity (Pb < 10−5, I2 = 94%); I2: a measure of variability expressed in %
Galbraith plot in the CEE analysis
log OR: logarithm of standardized odds ratio; SE: standard error. The two studies above the +2 and the one study below the −2 confidence limits are the outliers.
Forest plot outcome of outlier treatment in the CEE analysis
O: onset; P: progression; The diamond denotes the pooled odds ratio (OR) indicating even more myopia control (0.72) compared with the forest plot in Fig 4. Squares indicate the OR in each study. Horizontal lines on either side of each square represent the 95% confidence intervals (CI). The Z test for overall effect was significant (Pa = 0.0002).The chi-square test indicates zero heterogeneity (Pb = 0.95, I2 = 0%); I2: a measure of variability expressed in %
Sensitivity analysis and publication bias
Table 3 shows that five of the six significant outcomes were robust. Only the LQ comparison was non-robust on account of one study contributing to its stability [50]. We found no evidence of publication bias (Table 4).
Outcomes of sensitivity treatment on the significant findings
Outcomes of publication bias tests on the significant findings
Discussion
Summary of associations
Our principal findings point to the consistency of significant control of myopia which is supported by homogeneity, robustness and stability. Outlier treatment has unraveled significant and homogeneous associations which were not present in the component single-study outcomes. Conflicting outcomes between primary studies may be attributed to small sample sizes. Outcomes based on small sample sizes appear to be common in clinical studies [55] and are prone to the risk of Type 1 error. Outlier treatment and subgroup analysis underpinned the importance of doing CEE properly and frequently (> 5 time/week) for better myopia control and highlighted the adverse consequence of improper CEE performance. This meta-analysis showed that outcomes were more dependent on xs than cc or RCTs. The xs studies had high sample sizes but not RCTs in our meta-analysis. This is unsurprising since RCTs require greater logistical organization than xs studies, which are mainly questionnaire-based. In fact, two xs studies [52,49] included in our meta-analysis recommended RCTs for better understanding of CEE influence on myopia.
Comparison with a 2019 meta-analysis
We compare our findings with a recent meta-analysis (Lu et al. 2019) in terms of methodology, findings and conclusion [56]. The initial difference is that Lu et al examined myopia onset only, while we addressed both myopia onset and progression. Despite this difference, both meta-analyses used the GIV approach, lending comparability to the outcomes. Lu et al overall findings indicated increased risk for myopia onset (1.2-fold, n = 4) which contrasted with ours (28% more control of myopia, n = 6). Lu et al stated the harmful effects of low quality CEE on myopia which prompted us meta-analyze the data on low quality CEE from our included studies. Our initial low quality subgroup finding showed a null effect (OR 0.97, 95% CI 0.50-1.86, P = 0.92) until it was subjected to outlier treatment, the outcome of which indicated significantly less myopia control (OR 1.57, 95% CI 1.24-2.01, P = 0.0002). Thus, our PSO low quality finding concurred with the previous meta-analysis [56].
CEE and myopia
CEE is an eye-care program that aims to protect against or prevent myopia and is mandated in Chinese schools [51]. Massaging the acupoints around the eyes accelerates the blood circulation, improves metabolism, rests the eye muscles, and relieves eye fatigue [57]. From the perspective of traditional Chinese Medicine, Qi is stimulated in meridians by massaging the acupoints, thus relieving eye strain [48]. By far, performing CEE was found to have a modest protective effect against myopia [52]. In contrast, non-performance of CEE was found to increase the risk for myopia [53]. A third variation in CEE study findings was the non-significant association between performing CEE and risk of myopia onset and myopia progression [48]. These associative variations may be due to the limited sample sizes and lack of rigor in measuring the quality of CEE performed. In schools, most Chinese children were required to perform CEE only once or twice a day, once in the morning and once in the afternoon, with 5 minutes for each time. Therefore, the total time of performing CEE was at most 10 minutes a day and in some places the children only performed the exercise once a day [48]. Studies have reported that most children could not perform the CEE with standard manipulation. Previous studies showed that about 90% of Chinese children did not perform CEE correctly, most of them could not find the exact periocular acupoints and did not have accurate pressure and manipulation skills for the exercises although they did them every day [58,59]. Our findings pointed to a significant 57% reduced control of myopia, which may exacerbate the coonidtion when performed improperly. On the other hand, we also showed significantly better control (up to 51%) of myopia (which may retard its progression or reduce the onset) when CEE is performed seriously. About 15% of total children and about a third of those who performed them were found to achieve high quality CEE [48].
Strengths and limitations
Interpreting our findings should consider its limitations and strengths. Limitations include: (i) Given the greater reliance on RCTs over xs studies [52,49] limits our meta-analysis findings because xs data preclude direct evidence on the association between CEE and myopia [52]. (ii) All the component studies were underpowered; (iii) the low quality subgroup finding was non-robust; (iii) although less prevalent, myopia is not limited to the populations studied here, thus rendering the results inapplicable to other ethnic groups. (v) Clinically, the long-term effect of myopia prevention or slowing its progression remains an open question [51]. (vi) The likelihood of recall bias regarding responses of the participants to the questionnaires may have affected the accuracy of our study. However, impact of these limitations may have been mitigated with our assessments of the risk of bias and methodological quality of the studies. On the other hand, the strengths comprise of the following: (i) our use of the GIV allowed comparison with a previous meta-analysis. (ii) Outlier treatment was key to generating significance and eliminating heterogeneity which underpinned our core findings. This demonstrates the utility of this meta-analysis tool in elevating the level of evidence for associations; (iii) subgroup analysis validated the overall findings, allowed comparison and confirmed the adverse effect of LQ performance of CEE; (iv) all significant PSO ORs survived the Bonferroni correction, thus minimizing the possibility of a Type 1 error; (v) sensitivity treatment conferred robustness to most of the significant findings; (vi) no evidence of publication bias.
Conclusions
We have shown the significant influence of CEE in controlling myopia, provided the performance was serious and frequent. These findings are buttressed with layers of evidential strength which include high statistical significance, stability and homogeneity.The substantial amount of evidence presented here may render CEE as a useful preventative and therapeutic option in myopia control. In spite of the evidence for associations, the complexity of CEE performance involves interactions between genetic and non-genetic factors that may not have been covered in this meta-analysis because of the logistical limitations. Additional well-designed studies exploring other parameters would confirm or modify our results and add to the extant knowledge about the influence of of CEE in controlling myopia
Data Availability
Available upon request
Data availability
The data [frequency, odds ratios and 95% confidence intervals] supporting this meta-analysis are from previously reported studies and datasets, which have been cited. The processed data are available from the corresponding author upon request.
Conflicts of interest
The authors declare that there is no conflict of interest regarding the publication of this paper
Funding statement
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for- profit sectors
Supplementary materials
S1 Table Jadad DOCX
S2 Table NOS DOCX
S3 Table PRISMA DOCX
Author contributions
Conceptualization: PS, JT
Formal analysis: NP, PT, PS
Investigation: NP, PT, PS
Methodology: NP, PT
Resources: PS, JT, AT
Supervision: AT
Validation: NP, PT, KP
Writing – original draft: NP, PT, PS
Writing – review & editing: NP, PT, PS, AT
List of abbreviations
- AM
- analysis model
- cc
- case-control
- CEE
- Chinese eye exercises
- CI
- confidence interval
- D
- diopter
- EH
- eliminated heterogeneity
- F
- fixed-effects
- GS
- gained significance
- GIV
- generic inverse variance
- I2
- measure of variability
- IQR
- interquartile range
- Log
- logarithm
- LS
- lost significance
- maf
- minor allele frequency
- MC
- myopia control
- MQJ
- methodological quality judgment
- n
- number of studies
- NOS
- Newcastle-Ottawa Scale
- OR
- odds ratio
- Pa
- P-value for association
- Pb
- P-value for heterogeneity
- PRISMA
- Preferred Reporting Items for Systematic Reviews and Meta-Analyses
- PRO
- pre-outlier
- PSO
- post-outlier
- R
- random-effects
- [R]
- Reference
- RCT
- randomized control trial
- RNS
- retained non-significance
- RH
- reduced heterogeneity
- SD
- standard deviation
- SE
- standard error
- SW
- Shapiro-Wilk
- xs
- cross-section
References
