- Design
- prospective population-based cohort study, United Kingdom, median follow-up 13.6 years
- Population
- 477,876 adults with haemoglobin measured, from 502,188 recruited 2006-2010; median age 58 years, 54.4% women
- Primary outcome
- all-cause, cardiovascular and cancer mortality by haemoglobin category relative to WHO sex-specific thresholds
- Effect
- 10-year standardised cumulative mortality 4.5% in the reference group vs 12.5% more than 2 g/dL below threshold (absolute difference 8.1 points, 95% CI 6.7-9.4; adjusted HR 2.87, 95% CI 2.54-3.25); 0-1 g/dL above threshold 5.3%, adjusted HR 1.18 (95% CI 1.15-1.21)
The WHO definition of anaemia dates from 1959 and sets thresholds of 13.0 g/dL in men and 12.0 g/dL in women. A haemoglobin just above those numbers is routinely reported as normal and routinely ignored. This cohort asked what mortality actually does across the range.
UK Biobank recruited 502,188 adults between 2006 and 2010, with haemoglobin measured in 477,876, median age 58, 54.4% women, and followed them for a median 13.6 years. All-cause mortality showed a U-shaped relationship with haemoglobin, with the lowest risk 1 to 3 g/dL above the WHO threshold rather than at it. In the primary adjusted model, 10-year standardised cumulative mortality was 4.5% in that reference group and 12.5% among those more than 2 g/dL below the threshold, an absolute difference of 8.1 percentage points (95% CI 6.7 to 9.4), with an adjusted hazard ratio over the whole follow-up of 2.87 (95% CI 2.54 to 3.25).
The finding that changes practice is the borderline group. Among those sitting 0 to 1 g/dL above the WHO threshold, technically not anaemic, 10-year mortality was 5.3%, an absolute excess of 0.8 percentage points (95% CI 0.7 to 1.0) and an adjusted hazard ratio of 1.18 (95% CI 1.15 to 1.21). Cardiovascular and cancer mortality followed the same pattern at low and borderline values. Associations were less consistent in women under 60.
The authors are explicit that these are hypothesis-generating findings and should not be used to redefine anaemia, and that caution is right: this is a single baseline measurement in a predominantly White, healthier-than-average volunteer cohort, and low haemoglobin is as likely to be a marker of undiagnosed illness as a cause of death. The practical change is smaller and safer than a new threshold. A haemoglobin that has drifted down to just above the cutoff, particularly in an older man, is worth a ferritin and a look at the trend rather than a glance and a normal. In India, where iron deficiency is common and often the whole explanation, that costs very little to act on.
- Look at the haemoglobin trend across previous results rather than at whether the latest one clears the threshold
- Check ferritin, transferrin saturation, B12 and folate in a borderline haemoglobin rather than filing it as normal
- Pay particular attention to a falling haemoglobin in an older man, where the association was most consistent
- Do not adopt a new anaemia threshold on this evidence; the authors explicitly advise against it
- Remember that low haemoglobin often signals another disease, so the useful question is what is causing it
The statistics, in plain English
The hazard ratio of 1.18 for the borderline group is small in relative terms and corresponds to an absolute excess of under one percentage point over ten years, so this is a reason to look rather than a reason to alarm anyone. By contrast the group more than 2 g/dL below the threshold carried an 8.1 percentage point absolute excess and a hazard ratio of 2.87, which is large by any standard. The intervals are narrow throughout because the cohort is enormous, but precision is not the same as validity: with a single baseline measurement, reverse causation is the main threat, since undiagnosed cancer, kidney disease and inflammation all lower haemoglobin and all raise mortality on their own account.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for top clinical updates, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free