Haemoglobin thresholds to define anaemia from age 6 months to 65 years: estimates from international data sources

Abstract

Background: Detection of anaemia is crucial for clinical medicine and public health. Current WHO anaemia definitions are based on statistical thresholds (fifth centiles) set more than 50 years ago. We sought to establish evidence for the statistical haemoglobin thresholds for anaemia that can be applied globally and inform WHO and clinical guidelines.

Methods: In this analysis we identified international data sources from populations in the USA, England, Australia, China, the Netherlands, Canada, Ecuador, and Bangladesh with sufficient clinical and laboratory information collected between 1998 and 2020 to obtain a healthy reference sample. Individuals with clinical or biochemical evidence of a condition that could reduce haemoglobin concentrations were excluded. We estimated haemoglobin thresholds (ie, 5th centiles) for children aged 6-23 months, 24-59 months, 5-11 years, and 12-17 years, and adults aged 18-65 years (including during pregnancy) for individual datasets and pooled across data sources. We also collated findings from three large-scale genetic studies to summarise genetic variants affecting haemoglobin concentrations in different ancestral populations.

Findings: We identified eight data sources comprising 18 individual datasets that were eligible for inclusion in the analysis. In pooled analyses, the haemoglobin fifth centile was 104·4 g/L (90% CI 103·5-105·3) in 924 children aged 6-23 months, 110·2 g/L (109·5-110·9) in 1874 children aged 24-59 months, and 114·4 g/L (113·6-115·2) in 1839 children aged 5-11 years. Values diverged by sex in adolescents and adults. In pooled analyses, the fifth centile was 122·2 g/L (90% CI 121·3-123·1) in 1741 female adolescents aged 12-17 years and 128·2 g/L (126·4-130·0) in 1103 male adolescents aged 12-17 years. In pooled analyses of adults aged 18-65 years, the fifth centile was 119·7 g/L (90% CI 119·1-120·3) in 3640 non-pregnant females and 134·9 g/L (134·2-135·6) in 2377 males. Fifth centiles in pregnancy were 110·3 g/L (90% CI 109·5-111·0) in the first trimester (n=772) and 105·9 g/L (104·0-107·7) in the second trimester (n=111), with insufficient data for analysis in the third trimester. There were insufficient data for adults older than 65 years. We did not identify ancestry-specific high prevalence of non-clinically relevant genetic variants that influence haemoglobin concentrations.

Interpretation: Our results enable global harmonisation of clinical and public health haemoglobin thresholds for diagnosis of anaemia. Haemoglobin thresholds are similar between sexes until adolescence, after which males have higher thresholds than females. We did not find any evidence that thresholds should differ between people of differering ancestries.

Funding: World Health Organization and the Bill & Melinda Gates Foundation.

Figure 1

Derivation of haemoglobin thresholds in males and non-pregnant females aged 18–65 years (A) Continuous thresholds in adult males aged 18–65 years and non-pregnant females aged 18–65 years pooled across the following datasets: NHANES, HSE, and CHNS. The pooled continuous centiles and 90% CIs across data sources were estimated without accounting for the complex survey design and weighting. (B) Individual study and pooled (fixed effect, black line) estimates of the fifth centile of the haemoglobin distribution in the healthy reference population in male adults aged 18–65 years. (C) Individual study and pooled (fixed effect, black line) estimates of the fifth centile of the haemoglobin distribution in the healthy reference population in non-pregnant female adults aged 18–65 years. The dashed line in panels B and C represents the cutoff for anaemia according to the WHO guideline. CHNS=China Health and Nutrition Survey. HSE=Health Survey for England. NHANES=National Health and Nutrition Examination Survey (USA). NHS=National Health Survey (Australia). NNPAS=National Nutrition and Physical Activity Survey (Australia).

Figure 2

Derivation of haemoglobin thresholds in children aged 6–59 months (A) Continuous thresholds in male and female children aged 6–59 months pooled across the following datasets: TARGet Kids!, BRISC, ENSANUT, and NHANES. The pooled continuous centiles and 90% CIs across data sources were estimated without accounting for the complex survey design and weighting. (B) Individual study and pooled (fixed effect, black line) estimates of the fifth centile of the haemoglobin distribution in the healthy reference population in children aged 6–23 months. (C) Individual study and pooled (fixed effect, black line) estimates of the fifth centile of the haemoglobin distribution in the healthy reference population in children aged 24–59 months. The dashed line in panels B and C represents the cutoff for anaemia according to the WHO guideline. BRISC=Benefits and Risks of Iron interventions in Children (Bangladesh). ENSANUT=Encuesta Nacional de Salud y Nutrición (Ecuador). NHANES=National Health and Nutrition Examination Survey (USA). TARGet Kids!=Applied Research Group for Kids (Canada).

Figure 3

Derivation of haemoglobin thresholds in children aged 5–11 years (A) Continuous thresholds in male and female children aged 5–11 years pooled across the following datasets: TARGet Kids!, NHANES, and CHNS. The pooled continuous centiles and 90% CIs across data sources were estimated without accounting for the complex survey design and weighting. (B) Individual study and pooled (fixed effect, black line) estimates of the fifth centile of the haemoglobin distribution in the healthy reference population in children aged 5–11 years. The dashed line in panel B represents the cutoff of anaemia according to the WHO guideline. CHNS=China Health and Nutrition Survey. NHANES=National Health and Nutrition Examination Survey (USA). TARGet Kids!=Applied Research Group for Kids (Canada).

Figure 4

Derivation of haemoglobin thresholds in adolescents aged 12–17 years (A) Continuous thresholds in male and female adolescents aged 12–17 years pooled across the following datasets: NHANES and CHNS. The pooled continuous centiles and 90% CIs across data sources were estimated without accounting for the complex survey design and weighting. (B) Individual study and pooled (fixed effect, black line) estimates of the fifth centile of the haemoglobin distribution in the healthy reference population in male adolescents aged 12–17 years. (C) Individual study and pooled (fixed effect, black line) estimates of the fifth centile of the haemoglobin distribution in the healthy reference population in female adolescents aged 12–17 years. The dashed line in panels B and C represents the cutoff for anaemia according to the WHO guideline; there is a different cutoff for male adolescents aged 12–14 years (120 g/L) and those aged 15 years and older (130 g/L). CHNS=China Health and Nutrition Survey. NHANES=National Health and Nutrition Examination Survey (USA). NHS=National Health Survey (Australia). NNPAS=National Nutrition and Physical Activity Survey (Australia).