Methodological quality of the review: Medium confidence
Author: Zhu W, Meng YF, Wu Y, Xu M, Lu J
Region: Asia, Europe and America.
Sector: Diabetic retinopathy
Sub-sector: Alcohol intake, risk
Equity focus: None specified
Review type: Other review
Quantitative synthesis method: Meta-analysis
Qualitative synthesis method: Not applicable
Diabetic retinopathy (DR) is a common microvascular complication of diabetes mellitus. Alcohol intake is reported to be a risk factor for several kinds of diseases, including cancers, gastrointestinal diseases, respiratory disorders and infections. However, the associations of alcohol intake with DR risk have demonstrated contradictory results.
To investigate the association between alcohol intake and risk of DR.
In total, 15 observational studies were included in the meta-analysis, of which five were cohort studies, four were case control studies and six cross-sectional studies. The locations where the studies were conducted were as follows: six in Europe, two in America, six in Asia and one in Australia. The range of methodological scores was 3 to 8 points, and the average score was 6.07 points. Relatively high quality (6 points or more) was detected in 12 of the 15 included studies.
The pooled estimation of the 15 included observational studies that reported the association between alcohol consumption and DR risk was 0.91 (95% confidence interval (CI), 0.79 to 1.06), with significant heterogeneity detected (I2=62.3%). When authors stratified analyses by design, they found no significant association between alcohol consumption and DR incidence. In the subgroup analyses, authors report that intake of neither beer (odds ratio (OR) = 0.90, 95% CI = 0.76 to 1.08) nor spirits (OR = 1.00, 95% CI = 0.81 to 1.24) was associated with DR risk. Furthermore, authors found that protective effects were detected in the wine (OR = 0.77, 95% CI = 0.64 to 0.92) and sherry (OR = 0.22, 95% CI = 0.05 to 0.95) groups. Further subgroup analyses on the adjusting status, age adjustment, gender adjustment and BMI adjustment showed no significant association in each subgroup.
Authors note that, based on the sensitivity analyses, the outcome was not significantly changed when any study was removed from the meta-analysis. The funnel plot analysis revealed symmetry for the funnel plot (P = 0.692). However, Egger’s test demonstrated a significant publication bias (P = 0.049).
Authors conducted a search on three electronic databases including PubMed, EMBASE and Web of Science from inception to 2016. The search strategy combined the keywords and corresponding MeSH terms regarding alcohol intake (alcohol, drink, wine, beer and spirits) and diabetic retinopathy in observational epidemiological studies such as cohort, case control and cross-sectional studies. No restrictions were set in the literature search. Additionally, possible studies were detected by reviewing the references lists of the reviews or articles. When the data authors needed was missing in the article, they attempted to contact the corresponding authors for additional information. All the literature searches were conducted by one author and then checked by another author. The studies were included when they met the inclusion criteria as follows: (1) consisted of cohort, case control or cross-sectional design; (2) reported the association of alcohol intake and DR risk; (3) presented relative risk (RR), odds ratio (OR) or original data that could lead to OR values.
Two authors independently extracted data from included studies including name of the authors, publication year, age and number of participants, etc. The methodological quality of each included study was assessed by a checklist: (1) Defined study design (case control or cohort study, 1 point; cross-sectional study, 0 points), (2) List inclusion and exclusion criteria for all participants (Yes, 1 point; No, 0 points); (3) Indicate study period and follow-up duration (Yes, 1 point; No, 0 points); (4) Diagnosis of DR based on fundal examination or fundal photography (Yes, 1 point; No, 0 points); (5) Provided enrollment duration for all participants (Yes, 1 point; No, 0 points); (6) Described the general characteristics, such as age and sex, of the full participant population (Yes, 1 point; No, 0 points); (7) Adjusted for confounding factors, such as age, gender, DM duration or cardiovascular factors (Yes, 1 point; No, 0 points); (8) Stratified alcohol intake into more than three stratifications (Yes, 1 point; No, 0 points); (9) Common influence factors, including age, gender, DM duration, were matched among all the groups (Yes, 1 point; No, 0 points). Studies with over 6 points were considered to have relatively high methodological quality.
Authors evaluated the association between alcohol consumption and risk of DR by pooling the results from all the included studies. Adjusted OR or RR with 95% CI were used for the meta-analysis whenever possible. Meta-analysis was conducted using the random-effects model in the presence of significant heterogeneity. In this study, both the I2 method and the χ2 test were used to detect heterogeneity. Authors conducted subgroup analyses by study design types (cohort studies, case control studies and cross-sectional studies). Sensitivity analyses were also conducted to evaluate the influence of individual studies on the final conclusion.
Authors do not discuss the applicability or external validity of the results.
Authors reported that in advanced subgroup studies, no significant associations were detected in Europe (OR = 0.94, 95% CI = 0.75 to 1.17), America (OR = 0.97, 95% CI = 0.70 to 1.34) or Asia (OR = 0.94, 95% CI = 0.66 to 1.33). However, only one study in Australia was included in this study, and an inverse association between alcohol intake and DR risk was detected (OR = 0.52, 95% CI = 0.31 to 0.88).
Summary of quality assessment:
Overall, there is medium confidence in the conclusions about the effects of this study. Authors used appropriate methods to pool the data of included studies and assess heterogeneity across the studies. However, it is not clear from the review, if authors used appropriate methods to screen studies for inclusion avoiding selection bias. In addition, authors did not conduct a thorough search of the literature to ensure unpublished studies were included in the review and studies written in languages other than English were included.