Severe Vitamin D Deficiency Linked to Higher Mortality and Aging-Related Disease in Two Large Cohorts

Vitamin D has long occupied a curious position in medicine: almost everyone agrees that severe deficiency is bad, yet decades of observational studies and randomized trials have produced a confusing patchwork of results about exactly how bad, and whether pills can fix it. A new study published in Aging Cell now offers one of the most comprehensive real-world assessments to date, drawing on the electronic health records of hundreds of thousands of people in Israel and the United States. The researchers report that severe vitamin D deficiency, defined as serum 25-hydroxyvitamin D below 10 nanograms per milliliter, is associated with markedly elevated risks of death, diabetes, heart attack, stroke, dementia, kidney failure requiring dialysis, diabetic retinopathy, and even foot or toe amputation. Crucially, the team also found that pharmacy-documented vitamin D supplementation was linked to dose-dependent reductions in mortality and cardiometabolic disease, even after accounting for a person’s starting vitamin D level.

The investigation, led by researchers at Leumit Health Services, a nationwide Israeli health maintenance organization, was designed to tackle the central weakness of vitamin D research: disentangling whether low vitamin D is a cause of poor health or merely a marker of it. People who are sick tend to stay indoors, eat less well, and have lower vitamin D as a consequence of illness rather than as its cause. To address this, the team first characterized vitamin D levels across the Israeli population between 2009 and 2020, fitting a sinusoidal regression model to capture the pronounced seasonal rhythm of the hormone. Levels peaked in late summer and bottomed out in mid-February, and the researchers used this model to convert every measurement into a seasonally adjusted, nadir-equivalent value, ensuring that a blood draw in July was not unfairly compared with one taken in the depths of winter.

With this seasonal normalization in hand, the team constructed a tightly matched cohort of 12,352 severely deficient individuals and 12,352 controls with sufficient levels above 30 nanograms per milliliter. Matching was exact on age category, sex, smoking status, body mass index category, blood pressure category, ethnic sector, socioeconomic status, and year of enrollment, yielding groups with standardized mean differences below 0.01 on the key matching variables. Even so, at baseline the deficient group carried a heavier burden of disease: odds of foot or toe amputation were roughly fourfold higher, dementia odds were 1.89 times higher, cerebrovascular accident 1.79 times higher, diabetic retinopathy 1.69 times higher, and myocardial infarction 1.40 times higher. Intriguingly, skin malignancy was markedly less common in the deficient group, with an odds ratio of 0.44, a pattern the authors interpret as a signature of lower cumulative ultraviolet exposure rather than a protective effect of deficiency itself.

To guard against the possibility that these associations were an artifact of Israeli demographics or health system quirks, the team turned to TriNetX, a federated US research network aggregating de-identified records from more than 70 healthcare organizations covering over 120 million patients. After propensity score matching on age, sex, race, BMI, nicotine dependence, diabetes, hypertension, and hemoglobin A1c, they compared 223,175 severely deficient individuals with an equal number of sufficient controls, with more than half of the matched cohort identifying as Black or African American, reflecting the demographic distribution of severe deficiency. The baseline pattern replicated almost exactly: higher odds of amputation, retinopathy, dementia, stroke, myocardial infarction, and hemodialysis in the deficient group, and again lower odds of skin malignancy.

The longitudinal results were more striking still. Over follow-up of up to ten years, with a six-month lag excluding anyone who developed an outcome shortly after their index measurement to reduce reverse causation, severe deficiency predicted incident disease across every major domain examined. In the Israeli cohort, the adjusted hazard ratio for all-cause mortality was 2.57, with cumulative eight-year mortality of roughly 14 percent in the deficient group versus about 8 percent in controls. Hazard ratios reached 3.59 for foot or toe amputation, 1.84 for dialysis initiation, 1.73 for diabetic retinopathy, 1.54 for stroke, 1.35 for dementia, and 1.26 for myocardial infarction. In the US cohort the estimates were more modest but directionally identical, with a mortality hazard ratio of 1.59 and a hemodialysis hazard ratio of 2.07. Extending the lag to twelve and twenty-four months left the findings materially unchanged.

The most ambitious part of the study concerned supplementation. Rather than relying on self-reported supplement use, the researchers mined pharmacy dispensing records, assigning every vitamin D-containing product a total content and modeling how purchases translated into predicted serum 25-hydroxyvitamin D increases over monthly, quarterly, and yearly intervals. Supplementation explained an additional 9.3 percent of the variance in serum levels beyond demographics and season, confirming that the pharmacy data captured genuine biological exposure. The team then built time-dependent Cox models in a cohort of 247,848 adults, jointly including baseline vitamin D category and dynamically updated supplementation category, adjusted for age, sex, smoking, socioeconomic status, BMI, kidney function, albuminuria, blood pressure, and baseline diabetes.

The results showed a graded, dose-dependent pattern on both sides of the equation. Baseline deficiency below 5 nanograms per milliliter carried a mortality hazard ratio of 1.44 relative to levels above 20, with risk falling steadily as baseline status improved. Supplementation, modeled independently, was associated with lower mortality in a dose-dependent fashion: a hazard ratio of 0.72 for a predicted serum increase above 10 nanograms per milliliter, 0.78 for 5 to 10, and 0.81 for 0.2 to 5. Similar protective associations emerged for incident diabetes and myocardial infarction. To strengthen causal interpretation, the researchers applied inverse-probability-of-treatment-weighted marginal structural Cox models, a framework designed to handle time-varying confounding in which a person’s health status influences both their decision to supplement and their subsequent risk. Under this analysis the protective signal for mortality actually strengthened, with a hazard ratio of 0.65 for the highest supplementation category.

Perhaps the most headline-grabbing numbers came from the number-needed-to-treat calculations. Among adults aged 70 and older with baseline levels below 5 nanograms per milliliter, the estimated ten-year number needed to treat to prevent one death ranged from roughly 9.7 to 10.5, figures comparable to well-established preventive interventions. Even individuals with baseline levels of 15 to 20 nanograms per milliliter, who would not qualify as severely deficient, showed meaningful benefit in older age groups, with numbers needed to treat of roughly 18 to 28. The authors argue this targeting logic may explain why large randomized trials of vitamin D have often disappointed: most enrolled participants regardless of baseline status, diluting any benefit confined to the severely deficient, while placebo groups frequently took supplements anyway and fixed low doses failed to achieve meaningful repletion.

The study is not without limitations, which the authors acknowledge candidly. As an observational analysis, residual confounding cannot be excluded, pharmacy dispensing does not guarantee ingestion, and over-the-counter purchases outside the health system went untracked. Yet the convergence of evidence, exact matching in one cohort, propensity matching in another, seasonal normalization, longitudinal exposure modeling, dose-response gradients, landmark sensitivity analyses, and causal-framework weighting, makes a purely artifactual explanation increasingly difficult to sustain. The skin cancer findings add a further layer of reassurance: a generalized healthy-user bias would have lowered risk across the board, and ultraviolet confounding would have pushed skin cancer risk in the opposite direction, but neither coherent pattern appeared in the supplementation analyses. From a geroscience perspective, the breadth of affected outcomes spanning metabolic, vascular, renal, neurodegenerative, and mortality domains suggests vitamin D status may function as a broad determinant of systemic resilience during aging rather than a disease-specific biomarker, and the authors propose that immune-mediated mechanisms, including better containment of latent intracellular infections, may underlie the pattern. Whether randomized trials targeted at severely deficient older adults will confirm these benefits remains the critical next step, but this study makes a compelling case that the question deserves to be asked at scale.

Subject of Research: Vitamin D deficiency, supplementation, and aging-related morbidity and mortality in longitudinal population cohorts

Article Title: Vitamin D Deficiency, Supplementation, and Aging‐Related Morbidity and Mortality Across Two Longitudinal Population Cohorts

Article References: Israel, A., Weizman, A., Israel, S., Ashkenazi, S., Ruppin, E., Magen, E., Merzon, E., & Vinker, S. (2026). Vitamin D Deficiency, Supplementation, and Aging‐Related Morbidity and Mortality Across Two Longitudinal Population Cohorts. Aging Cell, 25(10), Article e70752. https://doi.org/10.1111/acel.70752

Image Credits: AI Generated

DOI: 10.1111/acel.70752

Keywords: vitamin D, 25-hydroxyvitamin D, mortality, aging, diabetes, cardiovascular disease, dementia, supplementation, electronic health records, cohort study, causal inference, geroscience