Methodological quality of the review: Low confidence

Author: Bonovas S, Peponis V, Filioussi K

Geographical coverage: United States of America (USA), Congo, Australia, Netherlands, United Kingdom (UK), France, Korea

Sector: Glaucoma

Sub-sector: Open-angle glaucoma, diabetes mellitus

Equity focus: Patients with open-angle glaucoma

Review type: Effectiveness review

Quantitative synthesis method: Meta-analysis

Qualitative synthesis method: Not applicable

Background: The association of diabetes mellitus (DM) with primary open angle glaucoma (POAG) has been controversial. Diabetes mellitus has been suggested as one of the risk factors for POAG, along with other risk factors. Although numerous studies have tried to investigate this association, the presence of a link between the two diseases has not yet been confirmed.

Objectives: This study aims to examine the strength of the association between DM and POAG.

Main findings: Overall, 12 studies were included in the review. Five were case-control studies and seven were cross-sectional studies. 11 studies reported positive association and one study reported negative association. Only five studies had estimated odds ratios (ORs) that were statistically significant. The one study reporting negative association was not significant. Five studies were conducted in the USA, two in Congo and one each in the UK, France, Korea, Australia and Netherlands.

Significant heterogeneity between the studies was detected (P = 0.023). No evidence of publication bias was found (P = 0.37). The association of DM with POAG was statistically significant assuming either a random-effects model [OR = 1.50, 95% confidence interval (CI) 1.16, 1.93] or fixed-effects model (OR = 1.27, 95% CI 1.10, 1.45).

Based on the pooled analysis, authors concluded that diabetic patients are at significant risk of developing POAG and that clinicians should be aware of this possibility.

Methodology: Studies considered for inclusion were case-control, cohort or cross-sectional clinic- or population-based studies that evaluated DM and POAG. Articles were excluded for any of the following reasons: they were in a language other than English; they did not include DM as a risk factor for POAG; there was insufficient published data for determining an estimator of relative risk (odds ratio) or a confidence interval; and the disease was not specifically designed as POAG.

Authors conducted a literature search in MEDLINE for papers published between 1996 2002. The MESH headings, keywords and text words searched for included: glaucoma combined with diabetes. Authors also reviewed references of included studies to identify further potentially relevant studies. The titles and abstracts, and full text articles, were scanned to exclude any that were clearly irrelevant. The full text of the remaining articles was read to determine its relevance. However, it is not clear if this was conducted by two reviewers independently.

Data extraction was conducted by two reviewers independently. Authors extracted data on each study’s reference details, geographical coverage, study design, number of subjects, odds ratio and 95% CI, case definition for POAG, diabetes ascertainment, and which confounders were adjusted for. To assess the methodological quality of the studies, authors performed an open review of all studies considered for the meta-analysis.

Significant weight was given to: adequate control of potential confounders, case definition of POAG, diabetes ascertainment, random selection and response rates for cross-sectional studies; and non-biased selection of cases and controls for case-control studies. Authors noted that to avoid selection bias, no study was rejected due to the quality criteria. Instead, authors investigated the impact of the poor-quality studies on the overall effect size by removing these studies in the sensitivity analysis.

Authors performed a pooled analysis of data using two different techniques to estimate the pooled relative risk estimates of dichotomous factors: the Mantel-Haenszel method (assuming a fixed-effects model) and the DerSiminian-Laird methods (assuming a random-effects model). Publication bias was evaluated using the funnel graph and the Begg and Mazumdar adjusted rank correlation test. The absence of significant correlation suggests that the studies have been selected in an unbiased manner. To evaluate whether the results of the studies were homogeneous, authors used the Cochrane’s Q-test and the Galbraith’s plot. Additionally, to evaluate the stability of the results of the pooled analysis and explore heterogeneity, authors conducted a sensitivity analysis.

Applicability/external validity: Authors did not discuss the generalizability of the results.

Geographic focus: Review authors included one study conducted in a lower middle-income country and 11 remaining studies conducted in high-income settings. Nevertheless, the authors did not describe how applicable the results may be to low- and middle-income settings.

Summary of quality assessment: Overall, there is low confidence in the conclusions about the effects of this study as important limitations were identified. Authors used appropriate methods to analyse findings of included studies, however major limitations were identified in the methods used to identify, include and critically appraise studies. The search strategy was not comprehensive enough that we can be confident that relevant studies were not omitted in the review. Authors excluded the possibility of publication bias within the review as the Begg and Mazumdar test result showed that studies were selected in an unbiased manner, and did not take into account the possibility of the publication bias due to the non-robust search of the literature. Authors performed an open review of all included studies to quality-appraise studies, but it is not clear if this was based on a valid tool/checklist. Additionally, authors attributed an overall quality assessment to each study, but quality assessment was only provided for two studies.