Refractive Errors and Concomitant Strabismus: A Systematic Review and Meta-analysis

This systematic review and meta-analysis is to evaluate the risk of development of concomitant strabismus due to refractive errors. Eligible studies published from 1946 to April 1, 2016 were identified from MEDLINE and EMBASE that evaluated any kinds of refractive errors (myopia, hyperopia, astigmatism and anisometropia) as an independent factor for concomitant exotropia and concomitant esotropia. Totally 5065 published records were retrieved for screening, 157 of them eligible for detailed evaluation. Finally 7 population-based studies involving 23,541 study subjects met our criteria for meta-analysis. The combined OR showed that myopia was a risk factor for exotropia (OR: 5.23, P = 0.0001). We found hyperopia had a dose-related effect for esotropia (OR for a spherical equivalent [SE] of 2–3 diopters [D]: 10.16, P = 0.01; OR for an SE of 3-4D: 17.83, P < 0.0001; OR for an SE of 4-5D: 41.01, P < 0.0001; OR for an SE of ≥5D: 162.68, P < 0.0001). Sensitivity analysis indicated our results were robust. Results of this study confirmed myopia as a risk for concomitant exotropia and identified a dose-related effect for hyperopia as a risk of concomitant esotropia.

Data Analysis. We evaluated the association between concomitant strabismus and refractive errors by synthesizing the outcomes using meta-analysis. Among the eligible studies, four studies have classified concomitant strabismus into esotropia and exotropia 11,12,17,18 . We assessed the association of exotropia and esotropia with different kinds of refractive errors, i.e., myopia, hyperopia, astigmatism and anisometropia. The combined odds ratios (ORs) with 95% confidence intervals (CIs) of refractive errors as associated factors for exotropia and esotropia were analyzed. The Cochran Q statistic testing for heterogeneity across studies and the I 2 statistic quantifying the proportion of total variation attributable to between-study heterogeneity were calculated 19 . The P value of Q statistics lower than 0.1 and I 2 above 50% indicated high heterogeneity. If significant heterogeneity was detected, result from the random-effect model was adopted 20 , otherwise, the fixed-effect model was used 21 . Two articles have stratified hyperopia into different severities (SE of 2-3D, SE of 3-4D, SE of 4-5D and SE of > 5D), therefore ORs and 95% CIs of were combined to compare the risk of esotropia. Sensitivity analysis was performed to confirm the association by removing studies of higher risk of introducing bias. We also assessed the contribution of each study to the heterogeneity by sequentially omitting each study and recalculating the combined ORs. The Modified Egger's regression test was used to assess the potential publication bias, where a P value less than 0.05 was considered statistically significant 22 . The Review Manager software (RevMan, version 5.2; the Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen; 2012) was used for data analysis. The Stata software (version 12; StataCorp LP, College Station, TX) was used to validate the results and perform the Egger's test. P values less than 0.05 were considered statistical significant.

Results
From 1946 to April 1 of 2016, a total of 6962 publications were identified from the EMBASE and MEDLINE databases. After detailed screening and evaluation, 157 reports were eligible for detailed evaluation. Among them, we found 7 articles 1,11,12,17,18,23,24 meeting our criteria for meta-analysis ( Fig. 1). All these articles were population-based cross-sectional studies with the strabismus status and refractive error status of all included cases clearly documented. The studies spread across different ethnic groups, including Caucasian 11,12,17,23 and East Asian 1,18,24 . Strabismus of all studies was evaluated using the cover-uncover test to define as any heterotropia at distant or near distance with or without spectacles. Cases of heterophoria were not eligible for the meta-analysis, and were excluded. Refractive errors, including myopia, hyperopia, astigmatism and anisometropia, were measured under cycloplegic condition in all included studies (Table 1). Overall, 23,541 subjects with age ranging from 6 months to 12 years were recruited for the meta-analysis in these 7 studies. Four articles 11,12,17,18 have studied the association between refractive errors and different types of strabismus (including exotropia and esotropia). Three articles 12,17,18 reported adjusted ORs and 95% CI for esotropia and exotropia, respectively. Factors that were usually adjusted included age, gender and refraction 12,17,18 (Table 1).

Risk of bias assessment and sensitivity analysis.
All studies were of high quality indicating low risk of bias when being included in this meta-analysis (Supplementary Table 2). Egger's tests did not show significant findings in all of the analyses (Tables 2 and 3). Furthermore, we performed sensitivity analysis by omitting each study at a time subsequently to confirm the results. The association of hyperopia with exotropia became significant after removing Cotter's study (OR: 9.59, 95% CI: 6.73-13.65, I 2 = 31%, P < 0.0001) 12 . None of the other results was altered in the sensitivity analysis.   Among the 7 studies, three have studied the association of myopia with exotropia and esotropia separately. Our results demonstrated no significant association between myopia and esotropia. On the other hand, children with myopia had 5.23-fold increase in risk to develop exotropia than those without significant ametropia. This may explain the high prevalence of exotropia in Asia, in which the prevalence of myopia is much higher 11,15,25 . The exact mechanism of how myopia may lead to exotropia is not certain. We postulated that the fusional control at distant of myopes is weakened due to the blurred distant vision. For near vision, less accommodative effort is required for clear image in myopes due to a larger accommodation lag 26 , which resulted in less accommodative convergence stimulated 27 . This prolonged suboptimal convergence may lead to breakdown of the fusional control and may predispose to exotropia development. In fact, previous studies have demonstrated that myopes without exotropia had a higher accommodative convergence to accommodation ratio (AC/A) than emmetropes, which also support our postulation 28,29 . The authors believed that myopes may require more convergence per accommodation in order to maintain good fusion and normal alignment, owing to the higher accommodation lag and thus less accommodative convergence stimulated. However, further prospective studies were warranted to confirm the postulation.   Our results revealed a very strong association between hyperopia and esotropia in a dose-related effect manner. We have therefore identified increasing risk of developing concomitant esotropia with the severity of hyperopia. Based on the meta-analysis of two papers 12,18 involving around 12000 children, our results showed that hyperopia starting at the 2.00D to less than 3.00D imposes more than a 10-fold increase in risk of developing concomitant esotropia, and even up to 40-fold increase in risk for hyperopia up to 5.00D hyperopia. Strikingly, children with hyperopia of 5.00D or more had 218 times of risk of developing esotropia compared to children with 0.00D to less than 1.00D. This dose-related effect is highly relevant to public health. Health care providers should be cautioned in that children with moderate to high hyperopia should be closely monitored for the risk of developing into esotropia.
When pooling up the unadjusted ORs of three articles, which have separated into esotropia and exotropia, astigmatism was also found to be a risk for both esotropia and exotropia. However, the results need to be interpreted with cautions, because the confounding effect of spherical myopia and hyperopia has not been adjusted in the analysis. In the seven studies for this meta-analysis, only two reports have provided the adjusted ORs for the effects of astigmatism to exotropia. No association was found between astigmatism and exotropia based on the pooled adjusted ORs. Therefore, further studies were needed to confirm the association.
Anisometropia was found to be associated with strabismus and both esotropia and exotropia based on pooled unadjusted ORs. This can be attributed to the much reduced binocularity in children with anisometropia 30 . However, the pooled adjusted OR did not support the association of anisometropia with exotropia and esotropia. Therefore, there is still no definite conclusion on their relationship and further studies are required to confirm the association.
In this meta-analysis, all included studies were population-based studies. The results obtained are of epidemiological relevance. Furthermore, we used risk of bias assessment tools for observational study recommended by MOOSE guidelines and Cochrane handbook for Systematic Reviews, which showed all included studies were of good quality. Sensitivity analysis has been conducted to further confirm our findings and no significant publication bias has been found. However, all the 7 reports were cross-sectional studies. Further longitudinal studies are warranted to establish a causal relationship between the two conditions. Moreover, we found the definition of refractive errors varied among studies, which could be a major source of heterogeneities. In our meta-analysis, we accounted for heterogeneities by using random-effect model and sensitivity analysis.
In summary, we confirmed that myopia increased the risk of exotropia. Hyperopia was associated with increased risk of esotropia in a dose-dependent manner. We also reported a suggestive association of astigmatism and anisometropia with concomitant strabismus, which should be further confirmed in follow-up studies. Refractive errors are extremely common especially in Asian populations. Strabismus is a difficult ophthalmic disorder that disrupts vision and depreciates quality of life. Investigations are warranted to understand the pathophysiology for the associations between refractive errors and concomitant strabismus.