Photoscreening for amblyopia risk factors assessment in young children: a systematic review with meta-analysis

Authors: Ferreira A, Vieira R, Maia S, Miranda V, Parreira R, Menéres P.

Geographical coverage: Europe, North America and Asia

Sector: Biomedical

Sub-sector: Treatment

Equity focus: Children aged under 3 years.

Study population: Children aged under 3 years of age who had been photo-screened for amblyopia.

Review type: Other review

Quantitative synthesis method: Systematic review and meta-analysis

Qualitative synthesis method: Not applicable

Background: Amblyopia is a leading cause of preventable and treatable vision loss in the paediatric population. Instrument-based screening of amblyopia-risk factors is being widely adopted but the audit of its results is still lacking. The US Preventive Services Task Force (USPSTF) recommends vision screening at least once in all children aged 3 to 5 years to detect amblyopia or its risk factors. However, this entity affirms that the benefits and harms of the vision screening in children younger than three years cannot be determined with the available evidence. Thus, there is no consensus on the optimal age to begin vision screening, its frequency or referral criteria.

Objectives: To review the existing evidence regarding the outcomes of photo-screening applied to children under the age of three years.

Main findings:

A total of 13 studies met the inclusion criteria and were included in the qualitative analysis for this systematic review. One study enrolled children aged between 6 months and 5 years but did not specify the number of patients by age; thus, this study was excluded from the quantitative analysis but was appraised in the qualitative synthesis as the proportions of interest to this work were discriminated by age group. Six of the 13 studies presented a low risk of bias.

Thirteen primary cross-sectional studies published between 2001 and 2021 were included. The number of children enrolled spanned between 404 and 42,149. The quantitative analysis included 12 studies assessing 64,041 children. Of these, 13% (95%CI: 7-19%) were referred for further confirmation of the screening result. Astigmatism was the most common diagnosis both after screening and after ophthalmologic assessment of referred children. The pooled untestable rate and PPV were 8% (95%CI: 3-15%) and 56% (95%CI: 40-71%), respectively.

Clinician experience and judgement still plays a major role on amblyopia management, as there is still no consensus on what level magnitude of refractive error must be considered an amblyopia-risk factor at different ages, nor the optimum time frame to prescribe refractive correction and prevent amblyopia.

Methodology:

The review included full-text original articles on photo-screening outcomes for children aged three or under in community settings. Exclusions were made for systematic reviews, guidelines, abstracts, comments, editorials, studies involving older children, pre-commercial equipment development studies, studies comparing device performance, non-community settings, and non-English articles. Studies involving children above and below three were included if outcomes for the target group were specified. Primary outcomes were referral rate, proportion of inconclusive tests, and positive predictive value (PPV), which was calculated if not provided. Secondary objectives involved summarizing referral criteria, identifying common diagnoses among referred children, and determining the proportion receiving amblyopia treatment.

One author searched three electronic bibliographic databases (Medline/PubMed, Web of Science and Scopus) from inception to 17 March  2021. Searches on electronic bibliographic databases were complemented by reviewing the references of included primary studies and by contacting authors.

One author screened abstracts from the initial search and excluded publications not written in English or fulfilling the first exclusion criteria. The abstracts of the remaining papers were screened by two authors using Rayyan software. Full-texts were retrieved for studies meeting the selection criteria or when eligibility was doubtful from the title and abstract alone; discrepancies were solved by a third reviewer. One author carried out the data extraction for a pre-prepared sheet using criteria agreed by all.

Quantitative analysis was conducted to pool referral rate, untestable/inconclusive rate, and positive predictive value (PPV) using the “metaprop” package in Stata software. Data was grouped by device and region for analysis. The Freeman-Tukey transformation was applied to avoid negative proportions in the confidence interval. A random-effects model was used to account for study heterogeneity, which was assessed using the I² statistic. Two reviewers independently evaluated each study’s risk of bias using the QUADAS-2 tool, considering patient selection, index test interpretation, reference test execution, and uniformity of reference test application.

Applicability/external validity: The authors highlighted several limitations, which limits the applicability and external validity of review findings.

Geographic focus: Included studies were undertaken across range of countries in Asia, Europe and North America with a small number of LMICs (India and Iran).

Summary of quality assessment:

Although data extraction was independently confirmed by a second author, it’s unclear if this task was performed independently by two authors. The review lacks a detailed, study-by-study presentation of quality assessment and risk of bias. Despite the generally robust data analysis methods, there was no discussion about the risks of bias in individual studies or attempts to analyse based on this. The authors acknowledged significant heterogeneity among the included studies. Due to these factors, the review’s findings have low confidence.