Relationship between vitamin D deficiency and diabetic retinopathy: a meta-analysis

Methodological quality of the review: Medium confidence

Author: Zhang J, Upala S, Sanguankeo A

Region: Worldwide

Sector: Diabetic retinopathy

Sub-sector: Vitamin D

Equity focus:  None specified

Review type: Other review

Quantitative synthesis method: Meta-analysis

Qualitative synthesis method: Not applicable


DR is one of the leading causes of blindness worldwide. It is thought that at 20 years after diabetes onset, nearly all patients with diabetes Type 1 and more than 60% of those with diabetes Type 2 will have evidence of DR on examination. DR, characterized as a neurovascular disease entity, results from hyperglycemia-induced changes to the blood–retinal barrier and retinal vasculature.


The purpose of this review was to evaluate the evidence for an association between diabetic retinopathy (DR) and vitamin D deficiency.

Main findings:

Authors included a total of 12 observational studies in the review. These were from Turkey, United States of America (USA), Japan, Italy, India, Iran, Lebanon, Spain, The Netherlands, China, England and Korea. Participants of included studies had both diabetes Type 1 and 2. Authors note that all studies measured 25(OH) D level at the time of examination or seasonally. The quality score of included studies attributed by the authors ranged from 3 to 8.

Nine out of the 12 studies were included in the meta-analysis comparing DR in optimal vitamin D and VDD groups. Authors reported statistical significant association between DR and vitamin D deficiency (VDD): 1.39(95% confidence interval (CI), 1.08–1.79) for studies reporting data associated with any type of DR; 1.21(95% CI, 1.08–1.35) for studies reporting data associated with non-proliferative diabetic retinopathy (NPDR); and 1.32 (95% CI, 1.16–1.50) for studies reporting data associated with proliferative diabetic retinopathy (PDR). Authors reported the overall pooled odds ratio (OR) for the nine studies as 1.27 (95% CI, 1.17–1.37; P = 0.001; I2 = 80%, P (heterogeneity) = 0.01.

In the meta-analysis which included a total of seven studies of serum vitamin D levels in patients with DR and the control group, authors reported statistically significant lower serum vitamin D level in patients with DR than in the control group, with an overall MD of -1.32 ng/mL (95% CI, –2.50 to –0.15; P = 0.027; I2 = 89%; P (heterogeneity) = 0.01).

Authors conducted a separate subgroup analyses for patients with NPDR (MD = –0.58 ng/mL; 95% CI, –2.09 to0.92ng/mL) and patients with PDR (MD = –2.21 ng/mL; 95%CI, –4.30 to –0.11 ng/mL. Authors noted a higher degree of VDD was seen in groups with higher grades of DR (i.e.,PDR) than those with lower grades (i.e.,NPDR).

Based on the sensitivity analysis of the meta-analysis comparing groups of groups of patients with optimal vitamin D and VDD, authors reported an OR of 1.20-1.38. In the sensitivity analysis of serum 25(OH) D levels comparing DR groups and control patients, after excluding each study one at a time, the mean difference ranged from –2.07 to –1.13ng/mL. Authors reported that none of the results were significantly altered in the sensitivity analysis.

Authors reported no presence of publication bias based on the funnel plot and Egger’s test.


Articles were considered eligible for inclusion if the following criteria were met: (i) published randomized controlled trials or observational studies including cross-sectional, cohort, and case-control studies; (ii) clear diagnostic criteria for VDD and DR were reported; and (iii) association of VDD and DR was reported as either adjusted or unadjusted hazard ratios (HRs), relative risks (RRs), or odds ratios (ORs) with associated 95% CIs, or the numbers of events. Authors included adult participants aged 18 or over with diabetes Type 1 or 2 who had serum vitamin D levels [25 (OH) D or 1,25 (OH) 2D] measured by standardized techniques. Authors defined Vitamin D deficiency as a serum level <20 ng/mL. Optimal vitamin D level was defined as a serum level 430 ng/mL. The main outcome of this study was DR as assessed by standardized method in patients with optimal vitamin D and VDD. The secondary outcome was vitamin D level [25(OH)D or 1,25(OH)2D] in patients with DR and control group patients.

Two authors independently conducted a search of published studies in the MEDLINE and EMBASE databases. References of all included studies were also reviewed. No date restrictions were applied. Two authors independently reviewed studies for inclusion in the review and extracted data of included studies. Methodological quality assessment of observational studies was conducted by two reviewers independently using the Newcastle-Ottawa Scale (NOS). A total score of 3 of less was considered poor, 4 to 6 was considered moderate, and 7 to 9 was deemed high quality. Authors noted excluding poor quality studies in the meta-analysis.

Meta-analysis used a random-effects model and authors calculated the pooled effect estimates with 95% CIs comparing optimal vitamin D and VDD groups. Authors also conducted a separate pooled analysis of subgroups of data for each DR severity. They also noted excluding studies from meta-analysis and presenting the results narratively when there were no sufficient comparable data available for the outcome of interest. Heterogeneity was quantified using the Q statistics, its P value, and I2. In addition, subgroup analysis by study design and meta-regression was performed by the authors to identify the source of heterogeneity. Reviewers assessed publication bias using funnel plot, Egger’s regression test and its implications with the trim and fill method.

Applicability/external validity:

Authors report that their analysis demonstrates significant association between CDD and DR and a statistically significant difference in mean serum vitamin D levels between DR and non-DR patients.

Geographic focus:

Authors noted that studies conducted in Lebanon, Japan, Italy and China found significantly increased odds of vitamin D deficiency and lower serum 25(OH) D in patients with DR. Authors noted that latitudes for these countries are from 33o85’N (Lebanon) to 41o87’N (Italy), whereas latitudes in countries that did not have significant results are from 32o43’N (Iran) to 52o36’N (England). In addition, authors reported that difference in latitudes form the included studies did not seem to affect the association between VDD and DR.

Summary of quality assessment:

Overall, medium confidence was attributed in the conclusions about the effects of this review. Although authors used appropriate methods to pool data of included studies, the literature search was not robust to ensure that all relevant studies were included in the review.

Publication Source:

Zhang J, Upala S, Sanguankeo A. Relationship between vitamin D deficiency and diabetic retinopathy: a meta-analysis. Can J Ophthalmol. 2017 Nov;52 Suppl 1:S39-S44

Downloadable link: source