The health records of nearly half a million Israelis say people very low in vitamin D died sooner, and that buying the pills went with up to 28% fewer deaths. Two large trials that handed out the pills found no difference at all

Few questions in nutrition have been asked as often, or answered as unhelpfully, as this one: does vitamin D do anything for people beyond their bones? Studies that measure people's blood levels keep finding that those with the least vitamin D are sicker and die sooner. Trials that give people vitamin D or a dummy pill and wait mostly find nothing. The usual explanation is that low vitamin D is a sign of poor health, not a cause of it. People who are ill, overweight, housebound or rarely outdoors make less of it from sunlight.

The new study, published in Aging Cell on 8 October 2026, tries to get past that with sheer size and some careful modelling. Researchers at Tel Aviv University, Leumit Health Services and colleagues in Israel and the US used the electronic records of Leumit, one of Israel's four national health funds: more than 1.67 million vitamin D blood tests from over 468,500 adults between 2009 and 2020. They took 12,352 people whose level was severely low (below 10 ng/mL, or 25 nmol/L, after correcting for the season) and paired each with someone of the same age group, sex, smoking status, weight band, blood pressure, community and income level whose level was above 30 ng/mL (75 nmol/L). Then they followed both groups for up to ten years. They repeated the comparison in TriNetX, a US network of records from more than 70 healthcare organisations, with 223,175 matched pairs.

The results look dramatic. In the Israeli records, about 14% of the severely deficient group had died within eight years against about 8% of their matched partners, a risk 2.57 times higher after adjustment. The US records gave 1.59 times. Amputation of a foot or toe was 3.59 times as likely in Israel, and heart attack, stroke, dementia, diabetic eye disease and dialysis were all more common in both countries. The one outcome that went the other way was skin cancer, which was rarer in the deficient group, most likely because they spent less time in the sun.

The second half of the paper is the part the authors lean on. Because Leumit runs its own pharmacies, the team could see who bought vitamin D and how much, and estimate how much each purchase would lift that person's blood level. Over 247,848 members aged 30 and over, people whose purchases worked out to more than 10 ng/mL of extra vitamin D had a 28% lower risk of dying over time (hazard ratio 0.72), and the result held up in a second, more elaborate statistical method. Diabetes and heart attacks were less common too. From these models the authors estimate that treating only 10 or so people over 70 with very low levels for ten years would prevent one death, and they conclude that severe deficiency is a 'potentially modifiable determinant' of ageing.

This is where we would slow down. Even in the matched pairs, the deficient group was already far sicker on the day of their first test: four times as likely to have had a foot or toe amputated, nearly twice as likely to have dementia and 1.8 times as likely to have had a stroke. Matching on age, weight and smoking did not make the two groups alike in health, and the follow-up differences may partly be the same illness carrying on. The supplement figures have a pattern of their own. The smallest category of purchases, enough to raise blood levels by under 5 ng/mL, went with 19% fewer deaths, nearly as much as the largest. A nudge that small doing most of the work looks less like a vitamin effect and more like a portrait of the people who pick up a prescription and keep taking it. The paper uses skin cancer as a check: supplements didn't lower it, so, the authors argue, sun and healthy habits can't explain the rest. But the one skin result they report for supplement buyers is a modest rise in non-melanoma skin cancer, which is what you would expect from people who see doctors more often and have more moles looked at. The authors themselves list the main weakness: supplement use 'may also co-occur with other unmeasured health-promoting behaviors'.

Then there are the trials. VITAL gave 25,871 older Americans (men from 50, women from 55) 2,000 IU of vitamin D a day or a placebo for a median of 5.3 years: no fewer cancers, no fewer heart attacks or strokes, and deaths from any cause level (hazard ratio 0.99). The D-Health trial gave 21,315 Australians aged 60 and over a 60,000 IU capsule or a dummy every month for five years: 562 deaths in the vitamin D group, 538 with placebo. The authors' answer is fair as far as it goes. Those trials took everyone, and most of their participants weren't deficient (D-Health's placebo group averaged 77 nmol/L, about 31 ng/mL), so a benefit confined to the severely deficient could have been missed. That is a reason to run a trial in people who are severely low, not a reason to treat the records as if they were one. Our reading: if a blood test shows a severely low level, correcting it is already standard care, for bones if nothing else. This study does not show that vitamin D adds years for people whose levels are normal, and its numbers needed to treat come from a model, not from anyone who was randomly given the pills. The work was funded internally by Leumit Health Services, whose records it uses, and the authors declare no conflicts of interest.