Lowest mortality sits above the anaemia threshold, not at it

Design

prospective population-based cohort study, sex- and age-stratified, median follow-up 13.6 years

Population

502,188 UK adults recruited 2006-2010, median age 58 years, 54.4% women

Primary outcome

all-cause mortality by haemoglobin relative to WHO thresholds; cardiovascular and cancer mortality

Effect

more than 2 g/dL below threshold: 10-year mortality 12.5% vs 4.5% (absolute difference 8.1 points, 95% CI 6.7-9.4; HR 2.87, 2.54-3.25); 0-1 g/dL above threshold: HR 1.18 (1.15-1.21)

A UK population cohort of 502,188 adults, haemoglobin measured in 477,876, followed for a median 13.6 years, with mortality analysed against the 1959 WHO thresholds (men below 13.0 g/dL, women below 12.0 g/dL). The relationship was U-shaped, and the lowest mortality fell 1 to 3 g/dL above the threshold rather than at or just above it.

The numbers at the bottom end are large. Participants more than 2 g/dL below threshold had a 10-year standardised cumulative mortality of 12.5% against 4.5% in the reference group — an absolute difference of 8.1 percentage points (95% CI 6.7 to 9.4), adjusted HR 2.87 (2.54 to 3.25). The finding that will change reading habits is at the borderline: those sitting 0 to 1 g/dL above the threshold, the group most often waved through, had a 10-year mortality of 5.3% against 4.5%, HR 1.18 (1.15 to 1.21). Associations were less consistent in women under 60, and patterns at the high end attenuated for cardiovascular and cancer death.

The authors say plainly that this should not define new thresholds, and they are right — this is a single baseline measurement in a predominantly White cohort, and low haemoglobin marks illness as readily as it causes it. What it does justify is not dismissing a haemoglobin of 13.2 in a man simply because it clears the line. In Indian practice, where iron deficiency and haemoglobinopathy are both common and baseline values differ, the threshold question is even less settled — the practical lesson is to look at the trajectory rather than the binary.

• Compare against the patient's own previous values before calling a borderline haemoglobin normal
• A fall of 1-2 g/dL within the normal range is a finding, even if no threshold was crossed
• Investigate a borderline value in an older man or postmenopausal woman as you would a low one
• Remember that low haemoglobin may be marking the illness rather than causing the death
• Do not transfuse or supplement on the strength of a cohort association

Why it matters

The line between anaemic and not was drawn in 1959, and most clinicians act on it as though it marked where risk begins.

Don't overread it

This is observational, with one baseline measurement — it cannot show that correcting a borderline haemoglobin improves survival.

The statistics, in plain English

A hazard ratio of 1.18 at the borderline is small, but it applies to a very large group, which is why the absolute difference is only 0.8 percentage points over ten years. A U-shaped curve means risk rises at both ends, so the high-haemoglobin signal is real too — but observational data cannot tell you whether raising or lowering a value would change anything, only where the curve's bottom sits.

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.