Vitamin D and Age-Related Macular Degeneration

Vitamin D and Age-Related Macular Degeneration: A Definitive Synthesis of Observational, Interventional, and Causal-Genetic Evidence

Executive Summary and the Central Conflict

The relationship between vitamin D and age-related macular degeneration (AMD) represents a prominent case study in scientific contradiction, where different tiers of evidence provide conflicting and increasingly complex conclusions. For years, a compelling hypothesis suggested a protective role for vitamin D, a notion supported by a large body of observational research and strong biological plausibility. This report synthesizes "everything we now know" by systematically analyzing three distinct and discordant streams of evidence.

1. Observational Data: A substantial collection of cross-sectional and case-control studies consistently demonstrates an association between low serum vitamin D levels and a higher prevalence of AMD, particularly its advanced, neovascular (wet) forms.
2. Interventional Data: The "gold-standard" large-scale randomized controlled trial (RCT), the Vitamin D and Omega-3 Trial (VITAL), found no overall effect of vitamin D supplementation on the incidence or progression of AMD, directly challenging the observational data.
3. Causal-Genetic Data: Mendelian Randomization (MR) studies, which use genetic proxies to mitigate confounding, are themselves in profound conflict. Some major MR analyses conclude there is no causal linkwhatsoever. Another pivotal MR study suggests the observed observational link is a confounding artifact, proposing that calcium (whose homeostasis vitamin D regulates) is the true protective causal agent. A third, recent MR study suggests a causal link for harm, finding that genetically high vitamin D levels increase the risk of wet AMD.

This report will critically deconstruct each evidence stream, analyzing their methodological strengths and flaws to reconcile these disparate findings. The primary conclusion is that the simple, protective hypothesis—that vitamin D deficiency causes AMD—is not supported by high-level evidence. The interventional data, while widely cited as "null," is determined to be inconclusive due to critical design flaws. The current frontier of the science lies in resolving the profound contradictions within the genetic data, which has shifted the research focus from vitamin D itself to the new, compelling hypotheses of calcium's protective role and the potentially confounding role of sunlight exposure.

I. The Vitamin D Hypothesis: From Population Association to Biological Plausibility

A. The Observational Evidence: A Consistent Signal of Association

The foundation for the vitamin D hypothesis rests on a large and consistent body of observational research linking low vitamin D status to a higher prevalence of AMD.

A 2016 systematic review and meta-analysis by Annweiler et al. provided the strongest codification of this link.¹ This review, which included 11 observational studies, found that individuals with high circulating 25-hydroxyvitamin D concentrations (in the highest quintile) had a 17% lower prevalence of AMD compared to those in the lowest quintile (summary odds ratio: 0.83).

Crucially, this association was found to be strongest for the most severe forms of the disease. The meta-analysis concluded that 25(OH)D concentrations below 50 nmol/L (approximately 20 ng/mL) were significantly associated with late-stage AMD, with an OR of 2.18, indicating a more than two-fold increase in risk.¹ This finding established the primary hypothesis: the most important role of vitamin D may not be in preventing the incidence of early AMD, but in slowing its progression to advanced, vision-threatening stages.

This finding is strongly supported by subsequent research. A 2022 study in the Turkish population by Kabataş et al. specifically evaluated vitamin D levels according to AMD stage.¹ The analysis found that serum vitamin D levels were significantly lower in patients with advanced-stage AMD compared to those with early- and intermediate-stage disease ($p=0.01$).¹

Further analysis has refined this association, specifying a link to the neovascular (wet) form of advanced AMD.

• The Kabataş et al. study noted that the risk for developing a choroid neovascular membrane (CNV) and subretinal fibrosis—the hallmarks of wet AMD—was found to increase as vitamin D levels decreased.¹
• A separate study by Kan et al., which focused exclusively on exudative (wet) AMD, found that the median 25(OH)D3 level in wet AMD patients was significantly lower than in healthy, matched controls (11.7 ng/mL vs. 17.0 ng/mL, $p=0.042$).¹ The frequency of wet AMD was significantly higher in the "deficient" (<20.0 ng/mL) group (55.0% vs. 36.0%, $p=0.043$).¹

While the majority of studies support this link, the observational evidence is not perfectly uniform. A 2022 study by Pérez Serena et al. also found that AMD patients had statistically significant lower median 25(OH)D levels compared to healthy controls (15 ng/mL vs. 21 ng/mL, $p=0.004$).¹ However, when stratifying by stage, this study did not find a significant difference in vitamin D levels between early, intermediate, advanced atrophic, and advanced neovascular AMD ($p=0.442$).¹

Despite this outlier, the overwhelming signal from observational data points to a dose-dependent association between vitamin D deficiency and a higher prevalence of advanced, neovascular AMD. However, this entire body of evidence is subject to critical methodologic limitations. Cross-sectional and case-control studies cannot prove causation. They are notoriously susceptible to confounding (e.g., individuals with AMD may have poorer diets or less healthy lifestyles, which also lead to low vitamin D) and reverse causality (e.g., individuals with vision-impairing AMD may be less mobile and spend less time outdoors, leading to reduced sunlight-induced vitamin D production).¹

B. The Mechanistic Rationale: A Biologically Primed Hypothesis

The observational hypothesis was widely embraced because it is supported by exceptionally strong biological plausibility. Vitamin D is not just a nutrient; it functions as a potent steroid hormone that modulates fundamental cellular processes known to be dysfunctional in AMD. A 2019 review by Kaarniranta et al. outlines the precise mechanisms by which vitamin D could theoretically protect the retina.¹

1. Immune System and Complement Modulation: AMD is strongly linked to chronic, localized inflammation and dysregulation of the complement cascade. Vitamin D is a known immune system modulator. Its most compelling mechanistic link is its synergistic action with Complement Factor H (CFH) and Complement Factor I (CFI).¹ This is highly significant, as specific genetic variants in the CFH and CFI genes are the most important and well-established genetic risk factors for developing AMD.
2. Anti-Angiogenesis: The observational data ¹ pointed to a specific link with wet AMD, a disease defined by choroidal neovascularization (CNV), or the growth of abnormal, leaky blood vessels. Vitamin D is "reputed to possess antiangiogenic properties".¹ This provides a direct, testable biological mechanism that perfectly aligns with the specific observational finding.
3. Retinal Pigment Epithelium (RPE) Health: At its core, AMD is a disease of RPE dysfunction. The RPE is a critical cell layer that supports photoreceptors. Vitamin D is known to act on RPE cells in several protective ways ¹:
• Autophagy: Impaired autophagy—the RPE's cellular waste-disposal system for clearing toxic byproducts—is a key driver of AMD. Emerging evidence suggests vitamin D can "influence autophagy," potentially helping RPE cells clear this damaging buildup.¹
• Inflammation: Vitamin D can modulate key inflammatory pathways in RPE cells, including the NLRP3 inflammasome.
• DNA Damage Response (DDR): The cellular response to DNA damage is weakened in AMD patients. Vitamin D has been shown to modulate this DDR, potentially protecting RPE cells from oxidative stress.¹

The biological case for vitamin D is, therefore, not vague. It aligns precisely with the most advanced molecular understanding of AMD pathogenesis, including its genetic (CFH), angiogenic (CNV), and cellular (RPE autophagy) components. This powerful, combined hypothesis demanded testing by higher-level evidence, which would ultimately produce profoundly challenging results.

II. Testing the Hypothesis: The Null Verdict from Randomized Controlled Trials

Given the strong observational and mechanistic evidence, the VITAL-AMD ancillary study was initiated to provide definitive, "gold-standard" interventional proof. Its findings, published in 2020, directly challenged the prevailing hypothesis.¹

A. The VITAL-AMD Trial: Design and Primary Outcomes

The VITAL (Vitamin D and Omega-3 Trial) was a massive, nationwide, $2 \times 2$ factorial randomized clinical trial involving 25,871 US men and women.¹ The relevant intervention arm for this analysis tested 2000 IU per day of vitamin $D_3$ (cholecalciferol) against a placebo.¹ The primary endpoint for the AMD ancillary study was a composite of total AMD events, which included both new incident cases of AMD and cases of progression to advanced AMD among participants who had AMD at baseline.¹

The top-line results, followed for a median of 5.3 years, were negative:

• Primary Endpoint (Total AMD Events): Supplementation with vitamin $D_3$ had no significant overall effect. There were 163 events in the vitamin D group and 161 in the placebo group, yielding a hazard ratio (HR) of 1.02 (95% CI: 0.82-1.27).¹
• Secondary Endpoint (Incident AMD): The intervention also had no effect on preventing new cases of AMD, with an HR of 1.09 (95% CI: 0.86-1.37).¹

As summarized in an accompanying commentary, these negative results from a large, well-designed, and well-conducted trial were "discouraging" and did not support the optimism that AMD progression could be slowed by this inexpensive supplement.¹

B. Critical Deconstruction of the "Null" Finding: A Trial Unfit for the Hypothesis?

A surface-level interpretation of the VITAL trial ¹ is that it definitively "disproves" the vitamin D hypothesis. However, a deeper, critical analysis of the trial's design reveals that it was methodologically unfit to test the specific hypothesis generated by the observational data. The VITAL trial's "null" finding is, in fact, inconclusive.

1. The "Progression" Signal: An Underpowered Endpoint

The strongest observational evidence 1 suggested the primary benefit of vitamin D was not in preventing incidence in healthy people, but in slowing progression to advanced disease in at-risk patients.

The VITAL trial did analyze this progression endpoint separately. The result was an HR of 0.63, suggesting a 37% risk reduction for AMD progression in the vitamin D group.¹ This trend supports the observational hypothesis. However, the finding was not statistically significant (95% CI: 0.33-1.21). The reason for the lack of significance was not necessarily a lack of effect, but a lack of statistical power. Across the entire cohort of 25,871 participants, only 39 progression events occurred.¹ The trial was, therefore, severely underpowered to detect a statistically significant effect on this specific, clinically crucial endpoint.

This is reinforced by prospective cohort data from an ancillary study of the AREDS2 trial, which followed a high-risk AMD population. That study found that higher dietary intake of vitamin D was significantly associated with a reduced risk of progression to neovascular AMD (HR: 0.59).² The remarkable similarity between this 0.59 HR in an at-risk cohort and the non-significant 0.63 HR in VITAL ¹ strongly suggests the VITAL trial's "null" result may be a Type II error (a false negative) due to insufficient power for that specific outcome.

2. The "Healthy Population" Problem: A Flawed Design for a Nutrient Trial

The VITAL trial was a primary prevention trial in a generally healthy population.1 The mean baseline 25(OH)D level of the participants was 30.8 ng/mL.1

This is a fatal flaw for testing a deficiency hypothesis. A level of 30.8 ng/mL is already considered "sufficient" by most clinical guidelines (which often define deficiency as <20 ng/mL).¹ Therefore, VITAL did not test the relevant clinical question: "Does correcting a deficiency in at-risk patients slow progression?" Instead, it tested the question: "Does extravitamin D (2000 IU/day) provide additional benefit to an already-sufficient population?" The negative answer to this question is not surprising and does not invalidate the deficiency hypothesis.

As noted in reviews of nutrient trial design, RCTs for vitamins often "fail" because they are designed like drug trials, neglecting the unique metabolism of nutrients. Key limitations include enrolling participants with high baseline nutrient levels (few participants having low baseline 25(OH)D) and using relatively small doses, which cannot demonstrate a benefit that has already been achieved by the population's baseline status.³ The VITAL-AMD trial is a textbook example of these design limitations.

At this stage of the analysis, the observational hypothesis (protection from progression in deficient states) remains viable and has not been adequately disproven by the inconclusive VITAL trial.

III. The Search for Causality: Contradictory Revelations from Mendelian Randomization

Mendelian Randomization (MR) offers a higher level of evidence. This genetic epidemiological technique uses genetic variants (SNPs) as a proxy (an instrumental variable) for lifelong exposure (e.g., lifelong high or low vitamin D levels). Because genes are randomly assigned at conception, this method avoids the confounding (by lifestyle, diet, etc.) and reverse causality (illness causing low VitD) that plagues observational studies. This approach, however, has yielded the most complex and contradictory findings to date.

A. Genetic Evidence for a Null Causal Relationship

The first and most replicated MR finding directly refutes the observational hypothesis, suggesting the association is not causal.

• A large, bidirectional 2-sample MR study by Dong et al. utilized summary-level data from the UK Biobank and FinnGen datasets.¹ Its primary conclusion regarding vitamin D was unambiguous: "No significant causal relationships were detected between genetically predicted 25(OH)D concentrations and AMD and its subtypes" (all $P > 0.05$).¹
• This conclusion was independently replicated by Luo et al. in a 2025 MR study.¹ This analysis also found no evidence to support a causal relationship between 25(OH)D concentrations and AMD, dry AMD, or wet AMD in Europeans ($P > 0.05$).¹
• A third MR study confirmed this, finding no causal link between either vitamin D or calcium and AMD.⁵

This powerful, replicated genetic evidence suggests that the entire observational link ¹ and the supportive trend in the VITAL trial ¹ are likely statistical artifacts, not reflective of a true causal relationship.

B. A Novel Causal Hypothesis: The Protective Role of Calcium

The Dong et al. study ¹ provided a pivotal, elegant explanation for this discrepancy. While it found no causal link for vitamin D, it also tested the causal role of calcium, whose homeostasis is "regulated primarily by vitamin D".¹

The results for calcium were the opposite of those for vitamin D. The study provided evidence supporting a causal relationship between higher calcium concentrations and a reduced risk of AMD.¹

• A standard deviation increase in genetically predicted calcium was linked to a 14% reduction in overall AMD risk (95% CI: 0.77, 0.97).
• The protective effect was clear for both subtypes: a 17% reduction in wet AMD (95% CI: 0.73, 0.95) and a 13% reduction in dry AMD (95% CI: 0.75, 1.00).¹

This finding provides a powerful synthesis. The observational studies ¹ were correct that they were seeing a realbiological signal, but they misattributed it. Because low vitamin D status can lead to dysregulated calcium homeostasis, the association they measured between "low vitamin D" and "high AMD" was likely a non-causal shadow of the true, underlying causal relationship: "low calcium" and "high AMD." Vitamin D was merely a proxy, or confounding variable, for calcium status.¹ This hypothesis elegantly explains why observational studies (which cannot separate the two) see a link, while MR studies (which can genetically separate them) find the link is only with calcium.

C. A Paradigm-Shifting Contradiction: Genetic Evidence for Harm

Just as the "calcium hypothesis" provided a clean resolution, a 2024 MR study by Fan et al. introduced a profound, new contradiction.¹ This study, also using UK Biobank and FinnGen data, directly conflicts with the "no link" findings of Dong et al. ¹ and Luo et al..¹

The Fan et al. study reported "robust evidence of a causal relationship between genetically predicted 25(OH)D levels and an increased risk of wet age-related macular degeneration (WAMD)".¹

• The odds ratio was 1.35 (95% CI: 1.09-1.67), suggesting a 35% increase in wet AMD risk for those with genetically high vitamin D.¹
• This harmful association remained significant (OR 1.86) even after multivariable MR analysis adjusted for confounders like type 2 diabetes, hypertension, smoking, alcohol, and C-reactive protein (CRP).¹

This finding is the precise opposite of the original observational hypothesis ¹ and directly refutes the "no link" MR studies.¹

D. Reconciling the Reconcilers: Why Do the MR Studies Conflict?

The highest level of evidence is now in a state of turmoil. Three major MR studies ¹, often using the same data sources, have arrived at two opposite conclusions: "no causal link" and "causal harm." This discrepancy must stem from differences in their statistical methodology, specifically the selection of genetic instruments (SNPs) and the methods used to control for horizontal pleiotropy (where a single gene affects multiple, unrelated traits).

The Fan et al. study ¹ offers a plausible, if unproven, explanation for its "harm" finding: sunlight exposure. The paper notes that "sunlight exposure is the main source of vitamin D" and that sunlight (specifically UV and blue light) is a known risk factor for AMD.¹¹ It is, therefore, possible that the genetic instruments (SNPs) selected by Fan et al. to proxy for "high vitamin D" are also linked to a genetic predisposition for sun-seeking behavior or greater sun exposure. If so, the model would be capturing the harmful causal effect of sunlight and misattributing it to vitamin D.

This would imply that the "no link" findings from Dong et al. ¹ and Luo et al. ¹ are the more accurate conclusion, as their methods may have more effectively filtered out this specific pleiotropic pathway. However, this remains speculative. Furthermore, the "calcium" hypothesis, while elegant, is also contested by at least one MR study that found no causal link for calcium or vitamin D.⁵

The only unified consensus across all MR studies ¹ is that vitamin D itself is not a simple, direct, protective agent.The scientific question has shattered from "Is vitamin D protective?" into three new, unresolved debates:

1. Is the relationship truly null? ¹
2. Is the relationship null for vitamin D, but causally protective for calcium? ¹
3. Is the relationship causally harmful, or is this a pleiotropic artifact of sunlight? ¹

IV. Synthesis and Final Analysis: Reconciling a Triad of Conflicting Evidence

This report was tasked with synthesizing "everything we now know." The final analysis must weigh these three conflicting streams of evidence to arrive at a definitive conclusion.

• Hypothesis 1: The Observational Model.¹
• Finding: Low vitamin D is associated with a high prevalence of advanced/wet AMD.
• Verdict: Likely Incorrect. This association is consistent but fails to hold up to causal analysis. The "no link" MR findings ¹ are robust. The most likely explanation is that the association is an artifact of confounding, with the MR study by Dong et al. ¹ providing the most compelling candidate for the true causal agent: calcium.
• Hypothesis 2: The Interventional RCT Model.¹
• Finding: Supplementing a healthy, vitamin D-sufficient population with 2000 IU/day provides no overall benefit.¹
• Verdict: Inconclusive, not "Null." The VITAL trial was the wrong experiment to test the hypothesis. By enrolling an already-sufficient population (baseline 30.8 ng/mL) and being severely underpowered for the progression endpoint (n=39), it failed to address the core clinical question.¹ The non-significant 37% risk reduction (HR 0.63) for progression remains a tantalizing, untested signal.¹
• Hypothesis 3: The Causal-Genetic Models.¹
• Finding: A state of profound contradiction. The evidence points simultaneously to "No Link" ¹, a "Protective Link via Calcium" ¹, and a "Harmful Link".¹
• Verdict: Unresolved, but Definitive on One Point. The only unified consensus across all genetic studies is that the simple, protective, causal relationship hypothesized by observational studies is false.

We now know that the simple idea that "vitamin D deficiency causes AMD" is incorrect. We know that supplementing a healthy, non-deficient population is useless.¹ The frontier of the science is now focused on resolving two new, contradictory causal hypotheses: the protective role of calcium ¹ and the potentially harmful (or sunlight-confounded) role of high vitamin D.¹

Table 1: Summary of Key Evidence on Vitamin D and AMD by Study Methodology

Study Type	Key Studies (Source ID)	Population Studied	Key Finding on VitD & AMD Relationship	Expert Conclusion on Validity/Limitations
Observational(Meta-Analysis)	Annweiler et al. 1	11 observational studies (mostly cross-sectional)	Protective Association. High VitD (vs. low) linked to 17% lower AMD risk (OR 0.83). Deficiency (<50 nmol/L) linked to 118% higher late-AMD risk (OR 2.18).	Likely Not Causal. High risk of confounding (e.g., by diet, lifestyle, or calcium) and potential reverse causality.
Observational(Case-Control)	Kabataş, Kan, Pérez Serena 1	AMD patients (esp. wet AMD) vs. healthy controls	Protective Association. VitD levels significantly lower in AMD patients, particularly in advanced/wet stages.1	Likely Not Causal. Reinforces the observational signal but shares the same fundamental flaws (confounding) as the meta-analysis.
Interventional(RCT)	Christen et al. (VITAL) 1	25,871 healthy adults (mean baseline 30.8 ng/mL)	No Significant Effect. (HR 1.02 for total AMD). Non-significant trend for protection in progression (HR 0.63).	Inconclusive. Trial was flawed: tested a sufficient population, not a deficient one. Severely underpowered for the key progression endpoint (n=39).3
Causal-Genetic (MR)	Dong et al. 1	UK Biobank & FinnGen	No Causal Link for VitD. (But, a protective causal link for Calcium).	Strong Causal Evidence.Elegantly reconciles observational data, suggesting VitD is a non-causal proxy for Calcium, the true protective factor.
Causal-Genetic (MR)	Luo et al. 1	FinnGen	No Causal Link for VitD.	Strong Causal Evidence.Replicates the "no link" finding from Dong et al..1
Causal-Genetic (MR)	Fan et al. 1	UK Biobank & FinnGen	Causal Harm. High genetic VitD linked to a 35-86% increased risk of Wet AMD.	Strong but Contradictory.Directly refutes 1 &.1 Mechanism unclear, but may be an artifact of unmeasured pleiotropy (e.g., sunlight exposure).

V. Clinical and Research Recommendations

A. Current Clinical Implications

1. On Supplementation: Based on the highest-quality evidence—the inconclusive VITAL trial ¹ and the contradictory MR studies ¹—there is no current evidence to support recommending vitamin D supplementation for the specific prevention or treatment of age-related macular degeneration.
2. On Patient Counseling: Clinicians should advise patients that vitamin D's benefit for AMD is unproven. For patients taking vitamin D for other indications (e.g., osteoporosis), they should be counseled that it is unlikely to benefit their AMD. The Fan et al. finding of potential harm ¹ is concerning but isolated and contradicted; it does not warrant advising patients to stop supplementation for other valid medical reasons, but it strongly argues against starting it for eye health.
3. Standard of Care: The clinical focus for at-risk patients must remain on proven, evidence-based interventions: smoking cessation and the Age-Related Eye Disease Study (AREDS/AREDS2) formulations.¹

B. Future Research Imperatives

1. Resolve the Mendelian Randomization Conflict: This is the highest scientific priority. The direct contradiction between the Dong/Luo studies ¹ and the Fan study ¹ must be resolved. This requires new MR studies using refined genetic instruments to specifically test and control for the horizontal pleiotropic pathway of sunlight exposure.
2. Investigate the Calcium Hypothesis: The finding from Dong et al. ¹ that calcium has an independent, protective causal link to AMD is the single most compelling and actionable new hypothesis generated from this body of research. This finding demands independent replication by other research groups. If confirmed, this would pivot the entire nutritional research paradigm for AMD.
3. Design a Smarter RCT: If another interventional trial for vitamin D and AMD is ever funded, it must learn from the critical flaws of VITAL.¹ As advocated in reviews of nutrient trial design ³, such a trial must:
• Enroll only patients with intermediate AMD and documented vitamin D deficiency (e.g., <20 ng/mL).
• Be powered exclusively for the primary endpoint of progression to advanced AMD.
• Use a dose sufficient to achieve and maintain a target serum level in the treatment arm.