Mortality is lowest just above the anaemia threshold, and rises each way

Design

Prospective population cohort, multivariable-adjusted, stratified by age and sex

Population

502,188 UK adults; median age 58, 54.4% women; median follow-up 13.6 years

Primary outcome

All-cause, cardiovascular and cancer mortality by haemoglobin

Effect

Lowest risk 1-3 g/dL above threshold; >2 g/dL below HR 2.87 (2.54-3.25); 0-1 above HR 1.18 (1.15-1.21)

A population cohort of 502,188 UK adults (haemoglobin available for 477,876) related a single baseline haemoglobin to mortality over a median 13.6 years, using the WHO thresholds of under 13.0 g/dL in men and under 12.0 g/dL in women.

The association was U-shaped, with lowest all-cause mortality 1 to 3 g/dL above the threshold. More than 2 g/dL below threshold, 10-year mortality was 12.5% versus 4.5% at the reference (adjusted hazard ratio 2.87). Even just below the line, at 0 to 1 g/dL above it, risk was modestly raised (hazard ratio 1.18). Patterns were less consistent in women under 60, and cardiovascular and cancer mortality mirrored the low and borderline findings.

The authors are explicit that these are interpretive, not a basis for new cut-offs. The practical value is that a borderline or mildly low haemoglobin is not a number to wave through, and a high-normal or raised haemoglobin is not reassuring either. It argues for looking at the value and the patient, not just whether a threshold is crossed.

• Mortality was lowest at haemoglobin 1 to 3 g/dL above the WHO anaemia threshold.
• More than 2 g/dL below threshold carried roughly a tripling of mortality (hazard ratio 2.87).
• Even 0 to 1 g/dL above the threshold showed a modest excess (hazard ratio 1.18).
• A raised haemoglobin was also associated with higher mortality.
• Treat a borderline haemoglobin as worth explaining, not automatically normal.

Why it matters

It challenges the habit of treating a value just inside the reference range as nothing to explain.

Don't overread it

This is observational and hypothesis-generating; the authors say it should not be used to set new haemoglobin thresholds, and a single baseline value was used.

The statistics, in plain English

A hazard ratio of 2.87 is a large association and 1.18 a small one; because this is observational, low haemoglobin here marks underlying disease rather than proving that correcting the number saves lives.

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.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.