Adults whose hemoglobin sat just above the official anemia cutoff, a result any lab report would call normal, died at a slightly higher rate over the following decade than adults sitting a bit higher, according to a University of Glasgow study of 502,188 UK Biobank participants published September 8, 2026 in Annals of Internal Medicine. Death from any cause was lowest 1 to 3 g/dL above the World Health Organization’s anemia thresholds, and rose on both sides of that band.
Here is the concrete version. Pull up your last routine blood count and find the hemoglobin number. If you are a man and it sits just over 13.0 g/dL, or a woman just over 12.0 g/dL, this is a reason to ask your doctor whether anything explains it rather than letting “normal” close the question. It is not a reason to start iron on your own. The same curve that rises below the cutoff also rises above it.
Key takeaways
- Death rates were lowest at hemoglobin 1 to 3 g/dL above the WHO anemia thresholds of 13.0 g/dL for men and 12.0 g/dL for women.
- People 0 to 1 g/dL above the threshold, currently called normal, had a 10-year death rate of 5.3% versus 4.5% in the reference group, about 8 extra deaths per 1,000 over a decade.
- People more than 2 g/dL below the threshold had a 10-year death rate of 12.5%, nearly triple the risk and a far larger signal.
What the study found
Hemoglobin was measured in 477,876 of the 502,188 adults recruited into UK Biobank from 2006 to 2010, who were then tracked for a median of 13.6 years. Median age at entry was 58 and 54.4% were women. The team sorted people by how far their hemoglobin sat from the WHO sex-specific thresholds. The relationship with death was U-shaped, lowest in a band 1 to 3 g/dL above the threshold and higher on both sides of it.
The deficit group carried by far the heavier burden. Among people more than 2 g/dL below the threshold, 12.5% had died at 10 years versus 4.5% in the reference group, a difference of 8.1 percentage points, or roughly 8 extra deaths per 100 people over a decade (very likely between 6.7 and 9.4 points). Over the full follow-up their risk of dying was close to three times as high (hazard ratio 2.87, very likely between 2.54 and 3.25 times).
The borderline group is the news, and its effect is much smaller. Among people 0 to 1 g/dL above the threshold, 5.3% had died at 10 years, an increase of 0.8 percentage points, about 8 extra deaths per 1,000 over a decade (very likely between 7 and 10 per 1,000). That is about 18% higher risk (hazard ratio 1.18, very likely between 15% and 21% higher).
Associations were less consistent in women younger than 60. Cardiovascular and cancer death followed the same pattern at low and borderline hemoglobin but weakened at higher concentrations.
Dr. Kumar’s take
The wire summary blurs together two results that deserve to be read separately.
The below-threshold finding is not really about a lab cutoff. A hemoglobin 2 g/dL under the line in a middle-aged adult is a sign something is wrong, and tripled mortality is what you would expect from a group carrying undiagnosed cancers, slow gastrointestinal bleeding, kidney disease, and chronic inflammation.
The borderline finding is the interesting one, and it is small. Under one extra death per hundred people over a decade is the kind of effect a single baseline blood draw can produce on its own, which is the study’s own stated limitation: hemoglobin was measured once, at enrollment, and never again. Some people in that band were heading down toward real anemia from a cause that had not declared itself yet, so their hemoglobin was a marker of the illness rather than a cause of the death. That trap shows up across large cohorts, including long daytime naps linked to higher death risk in older adults, where the exposure is partly a symptom.
Two things worth saying plainly. The curve turns back up at high hemoglobin, which makes this a poor argument for iron supplements, and iron in people who are not iron deficient is not a neutral act. And the association held up least well in women under 60, the group most often told a low-normal hemoglobin is nothing to think about. That neither confirms nor overturns the reassurance; it means the data are thinnest where the question comes up most.
The authors say plainly that these findings should not be used to redraw the anemia thresholds, and the population was predominantly White and enrolled in one country.
What it means for you
If your hemoglobin is comfortably in range, this study does not ask anything of you.
If it sits in the first gram above the cutoff, treat it as a number worth a sentence of explanation rather than a number worth worry. In a menstruating woman or an endurance athlete, the explanation may already be obvious. In someone whose result used to sit higher, the change is worth a question at the next visit. The study reports how death rates tracked with hemoglobin across a whole population, not what any one person should do about a single result.
Do not take iron based on this. Iron treats iron deficiency, which requires testing to establish, and this study measured no one’s response to any treatment. And if your hemoglobin is below the threshold, that is not a borderline result and should not be filed as one. That is the group where the risk here was real and large.