Screening for presence or absence of diabetic retinopathy: A meta-analysis

Methodological quality of the review: Medium confidence

Author: Bragge P, Gruen, RL, Chau M, Forbes A, Taylor HR.

Region: United Kingdom (UK), United States of America (USA), Spain, Australia, Canada, Germany, Hong Kong, India and the Netherlands.

Sector: Diabetic retinopathy, screening

Sub-sector: None specified

Equity focus: None specified

Review type: Effectiveness review

Quantitative synthesis method: Meta-analysis

Qualitative synthesis methods: Not applicable

Background

Studies have demonstrated that early diagnosis and laser treatment of diabetic retinopathy (DR) can prevent severe vision loss. Because there are few symptoms other than vision loss for DR, regular screening is very important. Two gold standard DR screening techniques include: an ophthalmological examination by a trained health professional and stereoscopic seven-field fundus photography by a trained photographer with image interpretation by an experienced grader.

Outreach clinics have been shown to improve access to DR screening for hard-to-reach populations or perhaps those living in low- to middle-income countries. Authors noted the emergence of a ‘non-specialist community based outreach model’ where photography (taken by a photographer, technician or health worker) is used instead or a direct examination by a specialist and where the image is sent to a specialist centre for interpretation.

Outreach DR screening may offer better coverage of at-risk populations but also carries the potential risk of reductions in screening sensitivity (where cases of DR are missed) or specificity (with the increase of unnecessary referrals).

Research objectives

‘To examine how mydriasis and the medical qualifications of photographers who take retinal photographs influence the accuracy of screening for diabetic retinopathy (DR)’.

Main findings

The authors identified 20 studies consisting of diagnostic test accuracy for inclusion in the review. Studies were conducted in the UK, USA, Spain, Australia, Canada, Germany, Hong Kong, India and the Netherlands. All studies included measured the sensitivity and specificity of tests designed to detect any DR.

Variations in mydriatic status did not significantly influence sensitivity or specificity. The 20 studies from nine countries assessed that variations in medical qualifications of photographers did not significantly influence sensitivity. This meant that retinal photography by a photographer with no specialist medical or eye qualifications appeared unlikely to miss cases of DR that screening methods using mydriasis or a photographer with specialist medical qualifications would detect.

However, specificity of detection of any DR was significantly higher for screening methods that used a photographer with specialist medical or eye qualifications. This meant that photographers without specialist medical training yielded significantly lower specificity (that is, a greater false positive rate).

‘Only the category of “any DR” had sufficient consistency in definition, number of studies (n=20), and number of assessments (n=40) for meta-analysis. Variations in mydriatic status did not significantly influence sensitivity (odds ratio [OR], 0.89; 95% confidence interval [CI], 0.56- 1.41; P=.61) or specificity (OR, 0.94; 95% CI, 0.57-1.54; P=.80). Variations in medical qualifications of photographers did not significantly influence sensitivity (OR, 1.25; 95% CI, 0.31-5.12; P=.75). Specificity of detection of any DR was significantly higher for screening methods that use a photographer with specialist medical or eye qualifications (OR, 3.86; 95% CI, 1.78-8.37; P=.001).’

Authors noted that future studies should use consistent DR classification schemes to enable further research and analysis.

Methodology

For further information in regards to the methodology used to conduct this review, the authors referred to a large companion report which accompanied this article.

Inclusion criteria consisted of all primary studies that aimed to evaluate the accuracy of a DR screening method among patients with diabetes. Primary outcome measures included sensitivity, specificity, Kappa of the method(s) under investigation compared to the reference standard (chosen reference standards are widely accepted in the field of ophthalmology). Based on the information provided in the referred large companion report and as confirmed by the authors it was clear that diagnostic test accuracy was the study design sought for this review.

Authors searched main databases including MEDLINE, EMBASE and the Cochrane Library from inception to June 2009. Reference lists of clinical guidelines and literature reviews were also searched for further relevant citations. Relevant non-English studies were identified through the systematic search; however, the authors noted that these were not included due to feasibility and resource constraints associated with translation. Two authors independently screened citations and full-text articles against the inclusion criteria. Data extraction was performed by two authors and piloted using 10 included studies; approximately 10% of the data were audited (conducted independently) and appropriate corrections were made to the data and protocols. Nevertheless, it should be noted that quality/risk assessment of bias in included studies was not conducted.

Authors conducted a meta-analysis based on the screening method rather than on the methodology of each included study, which seemed appropriate to answer the research question.

Applicability/external validity

The authors did not discuss the applicability/external validity of the results.

Geographic focus

The review included studies conducted in UK, USA, Spain, Australia, Canada, Germany, Hong Kong, India and the Netherlands. The review did not specifically focus on low- and middle-income countries, but did make reference to how the findings (that is, that outreach screening is an effective alternative to on-site specialist examination) were useful for patients living with DR who live in remote, resource-poor settings.

Quality assessment

Authors conducted a partially comprehensive search for published and unpublished literature; language bias was not avoided due to feasibility and resource constraints associated with translation. Methods used to include studies and extract data were appropriately conducted avoiding risk of bias, nevertheless it should be noted that quality/risk assessment of bias in the included studies was not conducted. The corresponding author, when contacted, noted that risk of bias tools were not in commonly use at that point in time and they were not well-regarded for RCT assessment until 2011.

Although authors did not report the methodology of each included study, it was clear that diagnostic test accuracy study designs were sought for this review. Authors conducted a meta-analysis based on the screening method rather than on the methodology of each included study, which seemed appropriate in order to answer the research question. Authors noted that the comparison of a screening method under investigation with an appropriate reference standard was consistent with accepted principles of screening studies; and that the chosen reference standards are widely accepted in the field of ophthalmology.

Taking into account the information provided on the companion report and by the authors when contacted, there was a medium confidence in conclusions about the effects of this study.