Hemoglobin just over the anemia cutoff linked to more deaths in half a million UK adults

A man whose hemoglobin comes back at 12.9 has anemia under the World Health Organization’s definition. At 13.1 he does not. The line sits at 13 grams of hemoglobin per deciliter of blood for men and 12 for women, and those WHO sex-specific thresholds are still widely used. The chance of dying does not flip at that line.

In nearly half a million UK adults followed for more than a decade, deaths from any cause were least common when hemoglobin sat 1 to 3 points above the WHO threshold. The research, led by the University of Glasgow, is published in the Annals of Internal Medicine.

Those only just above the line, and so officially normal, already died at a slightly higher rate: their estimated 10-year risk of death was 5.3%, against 4.5% in the band where death rates were lowest. That works out to roughly 8 extra deaths per 1,000 over a decade. The authors frame the results as a lead for future research and say they should not define new thresholds, though they may inform how doctors interpret borderline values.

Where the WHO’s 1959 anemia definition came from

Hemoglobin is the iron-rich protein in red blood cells that carries oxygen from the lungs to the rest of the body, and it is usually measured as part of a complete blood count. The test is often used to check for anemia, a condition in which the body does not make enough healthy red cells.

The Glasgow team sorted its participants using what it calls the 1959 WHO definition of anemia. The WHO first proposed hemoglobin cutoffs that year, but the current cutoffs for men and women were first proposed in 1968. They came out of technical meetings of clinical and public health experts working with data from five studies of populations in Europe and North America with predominantly European ancestry. Data on elderly people were not available at the time.

The numbers describe how hemoglobin varies in healthy adults. When the WHO revisited them in 2024, it said basing cutoffs on outcomes such as symptoms or impaired function was preferred but not possible for all groups, so it used a statistical approach. The cutoffs are based on the 5th percentile of hemoglobin in healthy people, which means 95% of healthy individuals would have a higher value. The other 5% would be falsely diagnosed as anemic.

Nothing in that method asks who dies. The WHO’s own list of research gaps includes the hemoglobin levels below which an increased risk of mortality starts to show. For the general adult population, the link between the 1959 definition and death was uncertain, and that is the gap the researchers set out to fill.

How death risk climbed as hemoglobin approached the anemia cutoff

Across the whole range, deaths traced a U-shaped association with hemoglobin, higher at both ends. The bottom of the U, 1 to 3 points above each threshold, works out to roughly 14 to 16 for men and 13 to 15 for women.

“Our data show that the lowest risk of death is associated with a ‘definitely normal’ haemoglobin,” said Malek Ahmad, a clinician-scientist at the University of Glasgow and the study’s first author.

The steepest rise came well below the line. Among participants more than 2 points under the threshold, the estimated 10-year risk of death was 12.5%, nearly triple the lowest band’s 4.5%, or about 8 extra deaths for every 100. The same ordering held over the entire follow-up as well as at ten years.

The associations were less consistent in women under 60. One possible reason is that they are at higher risk for anemia during their menstrual periods, so in that age group a low reading may more often have an ordinary explanation than a hidden illness behind it. Split by cause, cardiovascular deaths and cancer deaths followed a similar pattern at low and borderline levels, with a weaker upturn at the high end.

Half a million UK Biobank adults, but only one hemoglobin reading each

UK Biobank, a population-based study in the United Kingdom, recruited 502,188 adults from 2006 to 2010. Hemoglobin was measured in 477,876 of them, and they were followed for a median of 13.6 years. Paul Welsh, a professor of molecular epidemiology at Glasgow and one of the authors, points to the very large study population, long follow-up and consistency across several statistical approaches, then adds that this remains an observational study. It can show which groups died more often; it cannot show that their hemoglobin was the reason.

The authors flag two limitations: a single baseline hemoglobin concentration measurement and a predominantly White population. The first means each person’s hemoglobin was captured once, at the start, with no record of how it moved over the years that followed. The second repeats a gap in the cutoffs themselves, which came from populations with predominantly European ancestry.

Korean data show the same kind of curve. In about 290,000 adults over 40 without cardiovascular disease, hemoglobin had a U- or J-shaped association with cardiovascular and all-cause deaths, according to a study published in 2018. The trend was similar in women but less clear-cut, the Seoul National University Hospital team reported.

Illness, oxygen or thicker blood: why hemoglobin might be linked to earlier death

Neither the British nor the Korean data can say why. One explanation treats low hemoglobin as a signal of a problem that is already there. Anemia may be a sign of a more serious condition, such as bleeding in the stomach, inflammation from an infection, kidney disease, cancer or autoimmune disease. The Seoul team suggested anemia may be a marker for an underlying inflammatory process. If that is most of the story, a single reading at the start would catch some people whose condition had not yet been diagnosed, and the same illness would explain both their low hemoglobin and their earlier death. Researchers call this reverse causation, because the underlying problem comes first and the low reading follows.

Hemoglobin could also matter in itself. When it is low, the cells in the body don’t get all the oxygen they need. The Seoul team noted that chronic anemia below 10 is known to result in increased cardiac output, meaning the heart pumps harder to compensate, which may lead to a thickened heart muscle. That is far deeper than the borderline band, where any direct effect would presumably be smaller.

High hemoglobin has its own suspects. It can be a sign of lung disease, heart disease, sleep apnea or polycythemia vera, a disorder in which the bone marrow overproduces red cells, and smoking and dehydration can affect results too. Extra red cells thicken the blood, slowing its flow, the Seoul researchers added.

Which explanation dominates decides whether raising hemoglobin would lower the risk at all, and neither cohort can tell them apart.

What a borderline hemoglobin result means while the WHO cutoffs stand

The WHO has already considered raising the line for men. In its 2024 review, it noted that data in adult men may suggest a higher cutoff of 13.5. It kept 13 anyway, citing uncertain evidence and the potential unintended consequences of a change, such as iron supplementation programs in men.

This study cannot say if investigating or treating people just above the threshold would improve their outcomes, and the researchers say that needs to be tested directly. Pierpaolo Pellicori, the study’s corresponding author and a consultant cardiologist in Glasgow, put it plainly: “Whether this should influence clinical practice is an important question for future research.”

For anyone reading their own lab report, reference ranges may vary based on the laboratory used and a person’s age, race and sex, so a borderline reading may or may not be flagged. The test itself can’t tell you what’s causing a low number, and a provider also weighs symptoms, medical history and the results of other blood tests. The study did not test iron or any other way of raising hemoglobin, so it offers no evidence that doing so would help.

The man at 13.1 is still not anemic, but he sits in a band where slightly more people died, for reasons one blood test cannot reveal.