The Association Between Serum Vitamin D Levels and Age-Related Macular Degeneration: A Systematic Meta-Analytic Review

Abstract

Purpose: We conducted a meta-analysis of individual studies reporting an association between serum vitamin D levels and AMD.

Methods: Relevant studies evaluating the association between serum vitamin D levels and AMD risk were identified by systematically searching four electronic literature databases (Ovid Medline, PubMed, EMBASE, and ISI Web of Science) censored by June 2015. Due to the heterogeneity of studies in categorizing serum vitamin D levels, all individual odds ratios (ORs) were recalculated and transferred for an increase of serum vitamin D levels by 10 ng/ml. Summary ORs and 95% confidence intervals (CIs) of AMD risk per 10-unit increase of serum vitamin D were obtained using standard meta-analysis. Publication bias was evaluated using funnel plots and Kendall's rank correlation tests.

Results: Ten individual studies were included and pooled in this meta-analysis. Meta-analysis of studies on AMD risk led to a pooled OR (95% CI) of 0.91 (0.69–1.22) for an increase of 25-hydroxy vitamin D by 10 ng/mL (P = 0.12). No indication for publication bias was found, but substantial heterogeneity was obtained (I² = 79.7%, P < 0.01). Estimates from subgroup analyses also did not show statistically significant associations of serum vitamin D levels with different stages (early AMD, late AMD, and advanced AMD) and subtypes of AMD (neovascular AMD and nonneovascular AMD; P > 0.05).

Conclusions: There is no evidence to indicate an inverse association between serum vitamin D levels and any stages and subtypes of AMD risk, but opposite results from the United States and Korea resulted in this nonsignificance. Potential difference across various study designs might exist, based on few studies reporting in heterogeneous manners so far. More studies are needed to further confirm the causality of vitamin D and AMD, especially longitudinal studies.

Age-related macular degeneration (AMD) is known as one leading cause of blindness among old adults aged 60 years or older.^1,2 Various risk factors have been investigated to evaluate the associations with development and progression of AMD, including aging, smoking, sunlight exposure decreased, family history of AMD, cardiovascular diseases, increased dietary fat intake, and abdominal obesity.^3–7 Several studies have shown an inverse association between 25-hydroxyvitamin D [25(OH)D] and some chronic diseases.^8–12 In 2007, Parekh and his colleagues¹³ were the first to examine the potential relationship between serum vitamin D and prevalent AMD using the third National Health and Nutrition Examination Survey (NHANES III), 1988 through 1994 and reported protective impacts of serum vitamin D on AMD. Afterward, researchers from several countries also reported their results. However, the number of epidemiologic studies on the associations between serum vitamin D levels and AMD was limited so far, and the results were reported in heterogeneous manner.^13–22 To our knowledge, no systematic review and meta-analysis was published to evaluate this association. Our team aims to perform a systematic review and meta-analysis by combining individual studies and summarizing an overall effect size for the association between serum vitamin D and AMD. Because of the different categorization of 25(OH)D levels from individual studies, were used to recalculate effect sizes by an increase 10 ng/mL of serum vitamin D levels for both within studies and across studies. 

Methods

Literature Search and Study Selection

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement for reporting systematic reviews and meta-analyses of studies was used for our study.²³ A systematic literature review was performed to search relevant studies reporting the relationship between AMD risk and circulating vitamin D levels up to June 10, 2015 in the following four common literature databases: (1) Ovid Medline (Ovid Technologies, Inc., New York, NY, USA), (2) PubMed (National Center for Biotechnology Information [NCBI] at the US National Library of Medicine [NLM], and the National Institutes of Health [NIH]; Bethesda, MD, USA), (3) EMBASE (Elsevier, Amsterdam, the Netherlands), (4) ISI Web of Science (Thomson Scientific Technical Support, New York, NY, USA). The following combinations of relevant key words in the article were searched: (25-OH-D OR vitamin D OR calcitriol OR cholecalciferol OR calcidiol OR 25-hydroxyvitamin D OR hydroxycholecalciferols OR 25-Hydroxyvitamin D3 1-alpha-Hydroxylase OR 1,25-dihydroxyvitamin D) AND (macular degeneration). All relevant articles were imported into Endnote X7.4 (Thomson Reuters, New York, NY, USA) to check for relevance, and duplicate articles were deleted. The full text was reviewed further if the abstract or title indicated possible relevance. Any potential included literatures were evaluated further their eligibility, and odds ratios (OR) and participant numbers across various serum vitamin D level subgroups had to be provided in original articles or obtained via personal contacts with authors (if possible) for our meta-analysis calculation and transformation. Cross-referencing was performed to further search possibly missed articles in above-mentioned study identification process. 

Data Extraction

If eligible, at least two researchers extracted the following data independently from each study: author(s), publication year, country, study period, study design, the numbers of AMD patients and the total study population, characteristics of the study population, respective measures of relevant risk for AMD according to serum vitamin D status, as well as covariates adjusted for. If such data were not explicitly reported or were presented in articles, especially measures used for the calculation of meta-analysis, they were derived from data provided in the articles or requested from the authors through email contacts, wherever possible. Any initial disagreement was resolved by consensus after additional review of the articles. 

Quality Score.

We developed a quality score for each included study using a 10-point scale as follows: (1) Was study design clearly described? (2) Were participant selection procedures reported clearly? (3) Was participant enrollment duration provided? (4) Were included participants compared with excluded ones? (5) Were the cutoffs of various vitamin D categories provided? (6) Were the age and sex of eligible participants clearly described? (7) Were total and patient numbers in each serum vitamin D category provided? (8) Were the associations between serum vitamin D levels and AMD were adjusted for both age and sex? (9) Were serum vitamin D measurement methods clearly reported? (10) Did authors considered the sunlight-related or season difference in their analyses? One point is awarded if a study met one of 10-item criteria; therefore, the total quality score of each study may range from 0 to 10, with a larger value representing higher study quality. If one certain item was not applicable, this item was scored as 0.5. 

Meta-Analyses.

Because no articles reported data of 1,25(OH)₂D₃ and AMD risk so far, our review were restricted to serum 25(OH)D levels, representing an integrated measure for vitamin D from diet, dietary supplements, and skin production. Main outcome variables were adjusted ORs for the association between serum 25(OH)D levels and AMD risk. For consistency, serum 25(OH)D levels provided in nanomoles per liter were converted to nanograms per milliliter, dividing by 2.5. Due to the different categorization of 25(OH)D levels based on various study population, ORs of AMD risk were recalculated for an increase of serum 25(OH)D levels by 10 ng/mL both within studies and across studies.²⁴ Median or midpoint for each category of serum vitamin D levels was used for ORs transformation and recalculation. Summary ORs from random effects models were calculated using standard meta-analysis methods.²⁵ The DerSimonian-Laird method was used to calculate random effect sizes in our study,^26,27 because the random effects estimates allow for variation of true effects across studies²⁸ (i.e., substantial heterogeneity was observed across studies). Heterogeneity was assessed by the I² statistic. If some included studies were identified as outliers in standardized deleted residuals analyses (i.e., absolute value of standardized deleted residual > 2), which may be considered to delete in further sensitivity analyses to evaluate the stabilities of effect sizes. Subgroup analyses were carried out to investigate the associations of serum vitamin D levels with each type or stage of AMD risk, including early AMD, late AMD, advanced AMD, neovascular AMD, and nonneovascular AMD. Only three individual studies^18,20,22 reported the associations of serum vitamin D levels with any AMD, no matter what AMD classification and stages, so the results of any AMD combining three studies were also presented in our subgroup analyses. Further subgroup analyses regarding different countries (United States, Korea, and France), study design (cross-sectional studies, case-control studies, and retrospective cohort studies), quality score (≤5 vs. >5), and vitamin D measurement approaches (radioimmunoassay, liquid chromatography-tandem mass spectrometry, chemiluminescence immunoassay, and not reported) were used to evaluate the source of heterogeneity. The funnel plot and Kendall's rank correlation tests were employed to evaluate publication bias across individual studies.²⁹ The R software, version 3.2.0 (in the public domain at https://www.r-project.org) was used for the analyses. 

Results

Literature Search Procedures and Characterization of Individual Studies

A flow diagram of the search process was illustrated in Figure 1. Total searches from four above-mentioned literature databases yielded 235 relevant articles. After excluding 95 duplicates, 140 titles and abstracts were reviewed and 26 studies showed to be possibly relevant. Sixteen articles were deleted due to the following reasons: no original studies but editorials, comments, reviews (n = 8), only data regarding vitamin D intake and receptors were provided (n = 6), repeated studies with included studies in our meta-analyses (n = 2).^30,31 Finally, the remaining 10 studies were included in our review, involving 57,643 old adults,^13–22 eight of which were eligible for our meta-analyses.^{13,15–18,20–22} Two studies were not used for our meta-analysis,^14,19 because only means and standard deviances among AMD groups and control groups were provided in original articles,^14,19 which could not be used for our meta-analysis calculation and transformation. The references of excluded studies are provided in Supplementary Material Appendix SI. 

Figure 1

 

Flow diagram of the literature search process.

Details on the respective study design, the study populations and the study results are shown in Table 1. Five cross-sectional studies,^{13,14,19,20,22} three case-control studies,^16,18,21 and two respective cohort studies^15,17 reported the associations of serum vitamin D with macular degeneration, all of which were published during 2007 through 2014. Five studies were conducted in the United States,^{13,15–17,21} two in Korea,^20,22 one in Israel,¹⁴ one in France,¹⁸ and one in Denmark.¹⁹ Each study reported the association of vitamin D levels with different AMD subtypes. For example, three individual studies described the association of serum vitamin D levels with early AMD,^13,15,22 three with late AMD,^18,20,22 two with advanced AMD,^13,15 three with neovascular AMD,^16,17,21 two with nonneovascular AMD,^17,21 and three with any AMD.^18,20,22 The quality score in these studies ranged from 3 to 9, with a mean of 6.3. Three studies had a score of 8^13,15,21 and one with 9 points.²² Detailed scoring results were presented in Table 2. 

Table 1

 

Individual Studies Reporting on the Association of Serum 25(OH)D Levels With Macular Degeneration

 

Table 2

 

Quality Scores of Individual Studies

 

Results of Meta-Analyses

Eight individual studies reporting the associations between serum vitamin D levels and at least one AMD type, were included in our meta-analysis. Six of eight studies showed that increasing 25(OH)D levels were associated with decreasing risk of AMD,^{13,15–18,21} but statistical significance was observed only in two studies.^13,21 A nonsignificant association was obtained per 10-ng/mL increase of serum vitamin D levels (OR, 0.91; 95% confidence interval [CI], 0.69−1.22; P = 0.12) in the meta-analysis. A substantial heterogeneity among these eight included studies was observed (I² = 79.7%; P < 0.01). Corresponding forest plot was presented in Figure 2. The funnel plot did not show evidence of publication bias (Fig. 3, Kendall's tau = −0.14; P = 0.72). In standardized deleted residuals analysis, no study was identified as an outlier (absolute value of standardized deleted residual ≤ 2). 

Figure 2

 

Meta-analysis: ORs of AMD per 10-ng/mL increase in serum vitamin D levels.

Figure 3

 

Funnel plot: natural logarithm of ORs for serum vitamin D levels by standard errors for eight individual studies.

Different classification of AMD were evaluated their associations with serum vitamin D levels in different included studies, such as early AMD, late AMD, advanced AMD, neovascular AMD, and nonneovascular AMD. Hence, subgroup analyses by various classification of AMD were performed and shown in Figure 4. No significance was observed for nonneovascular AMD (OR, 1.01; 95% CI, 0.90–1.15; P = 0.75), neovascular AMD (OR, 0.85; 95% CI, 0.59–1.22; P = 0.12), advanced AMD (OR, 1.06; 95% CI, 0.81–1.38; P = 0.69), late AMD (OR, 0.82; 95% CI, 0.66–1.01; P = 0.06), early AMD (OR, 0.96; 95% CI, 0.66–1.39; P = 0.65), and any AMD (OR, 1.02; 95% CI, 0.77–1.36; P = 0.83). Substantial heterogeneity was observed among individual studies reporting neovascular AMD (I² = 75.3%; P = 0.02), early AMD (I² = 89.1%; P < 0.01), and any AMD (I² = 67.3%; P = 0.05). 

Figure 4

 

Subgroup meta-analysis: ORs of AMD by different classifications per 10-ng/mL increase in serum vitamin D levels

Detailed results from further subgroup analyses and sensitivity were presented in Table 3, evaluating the stability of overall effect sizes by countries, study design, quality score, and vitamin D measurement approaches. American studies reported decreasing AMD risk (OR, 0.87; 95% CI, 0.78–0.96; P = 0.01),^{13,15–17,21} but Korean studies showed increasing AMD risk (OR, 1.09; 95% CI, 1.01, 1.18; P = 0.03) with the increase of serum vitamin D.^20,22 Substantial heterogeneity came primarily from studies in the United States (I² = 59.8; P = 0.04). Inverse association of serum vitamin D levels with AMD risk was observed only in three case-control studies (OR, 0.75; 95% CI, 0.66–0.85; P < 0.01), but not for cross-sectional studies and retrospective cohort studies. There was substantial heterogeneity across three cross-sectional studies (I² = 91.0%; P < 0.01). The effect sizes differed neither among low-quality (quality score ≤ 5) and high-quality (quality score > 5) studies, nor among different vitamin D measurement approaches (radioimmunoassay, liquid chromatography-tandem mass spectrometry, chemiluminescence, not reported). Results from sensitivity analyses found that inverse associations of serum vitamin D levels with AMD risk only after excluding two Korea studies.^20,22 

Table 3

 

Subgroup and Sensitivity Analyses for the Association of Serum Vitamin D Levels With Age-Related Macular Degeneration

 

Discussion

There is no evidence to confirm the inverse association of serum vitamin D levels with AMD risk in the meta-analyses based on substantial heterogeneity across individual studies. Further subgroup analyses were performed to examine the associations between serum vitamin D levels and various AMD types, such as nonneovascular AMD, neovascular AMD, advanced AMD, early AMD, and late AMD, but these associations did not reach statistical significance. Opposite results from the United States and Korea might centralize potential associations. However, available data on association of serum vitamin D levels with AMD risk are too limited to draw a conclusion so far. Further longitudinal studies, ideally taken both additional genetic and environmental factors into account, are needed to further elucidate the role of serum vitamin D on AMD risk and potential prevention. 

Various methods are available for the assessment of serum 25(OH)D levels. Current methods are radioimmunoassay, high-pressure liquid chromatography, chemiluminescence immunoassay, and liquid chromatography-tandem mass spectrometry.³² However, these methods have not been standardized, so between-study comparisons can be delicate.^33,34 Differences of between the highest and the lowest values were reported at 38% in a cross-calibration study of the 25-hydroxyvitamin D assays among five laboratories.³³ Another study also addressed that the average 25(OH)D concentrations as measured by enzyme immunoassay were on average 6.28 ng/mL lower compared with Diasorin radioimmunoassay.³⁵ However, no significant difference of effect size was detected across various vitamin D measurement approaches in our subgroup analyses, based on few studies. More numbers of individual studies might be needed to evaluate this possible influence on associations of vitamin D and AMD risk. 

The sex discrepancy on association of serum vitamin D levels only with late AMD risk was reported in one Korean study by Kim et al.,²² who conducted subgroup analyses by both sexes and found significant inverse association of serum vitamin D levels with late AMD in men, but not in women. In addition, sex discrepancy for early AMD was not confirmed in this study. However, other included studies did not report the association between serum vitamin D levels and AMD risk among females and males, respectively.^13–21 Possible inverse association of serum vitamin D levels and AMD risk by sex might be ignored by researches in these previous studies, especially different subtypes of AMD. More studies from various study population are needed to confirm this potential sex discrepancy. 

Adjusted effect sizes from individual studies were used for our meta-analyses if available, because the association between exposure and outcome is affected by other factors and adjusted effect sizes are regarded as more accurate results. However, three included studies in our meta-analyses only presented their results without any adjustments.^14,16,19 Additionally, smoking is documented a major modifiable behavioral factors associated with the development of AMD,^4,36–39 but only four included studies adjusted for current smoking status or history of smoking when they conducted their models for vitamin D and AMD risk.^15,17,21,22 Calcium intake may be an important covariate when assessing the association of serum vitamin D with AMD risk given the independent effect of calcium intake on AMD and given that calcium absorption is strongly influenced by vitamin D status.^40,41 However, no study adjusted for calcium intake among 10 included studies. It should be suggested to take relevant covariates into account when evaluating the association between vitamin D and AMD risk in further studies, especially smoking and calcium intake. 

To our knowledge, only one study among monozygotic twin pairs with discordant AMD⁴² was the first to explore a possible association between higher dietary vitamin D and earlier stages of AMD risk, but measuring dietary vitamin D intake in this case series study is only an approximate estimation through various food intake, which may lead to misinterpretations.^39,42,43 A randomized trial with vitamin D supplementation might be more valuable to evaluate the temporal and causal relationship between vitamin D and AMD risk. 

Our meta-analyses have own advantages and disadvantages. A major advantage is the use of advanced statistical analysis techniques summarizing adjusted associations across studies and over the entire range of serum vitamin D values, despite the very heterogeneous categorization of serum vitamin D levels from individual studies. Additionally, to the best of our knowledge, our systematic review and meta-analyses is the first to summarize systematically individual studies of the association between serum vitamin D levels and AMD so far, involving 57,643 older adults and 5983 AMD cases. Major disadvantages are related to limitations of the data provided by the individual studies. Firstly, median and midpoints of serum vitamin D categories had to be used for pooling. Hence, risk estimates might be less accurate than original data if available. Secondly, the meta-analyses presented here only include cross-sectional studies, cohort studies, and case-control studies, which have potential methodological limitations to detect the causality between exposure and outcome. Thirdly, heterogeneous AMD classifications resulted in fewer numbers of included studies for each AMD type, which also limited more in-depth analyses. Finally, despite the lack of indication of major publication bias in the formal evaluations employed, potential publication bias is impossible to be excluded completely, especially in the light of the low number of studies. Finally, although four databases were searched for the reviews (i.e., Ovid Medline, PubMed, EMBASE, and ISI Web of Knowledge), and extensive checks for completeness by cross-referencing was employed, we cannot exclude having missed relevant studies. 

Conclusions

Despite these limitations, this might be the first systematic and comprehensive meta-analytic review assessing the association of serum vitamin D levels with AMD risk, based on observational studies. No inverse association was confirmed between serum vitamin D levels and each subtype of AMD risk in our meta-analyses. However, it seemed a little early to draw a conclusion based on limited numbers of available studies so far, especially only retrospective study designs were conducted and substantial heterogeneity was detected across included studies. Longitudinal studies are highly desirable to enable more precise estimates and a better understanding of the role of vitamin D in different stages and types of AMD. 

Acknowledgments

Supported by grants from the Public Welfare Technology Project of Science Technology Department of Zhejiang Province (2015C33115; Hangzhou, Zhejiang, China), Zhejiang Committee of Science and Technology (2015F10013; Hangzhou, Zhejiang, China), Zhejiang Provincial Program for the Cultivation of High-level Innovative talents (2014; Hangzhou, Zhejiang, China). 

Disclosure: W. Wu, None; Y. Weng, None; X. Guo, None; L. Feng, None; H. Xia, None; Z. Jiang, None; J. Lou, None