BACKGROUND: Worldwide, more than half a billion women of reproductive age are anaemic. Each year, about 70 000 women who give birth die from postpartum haemorrhage. Almost all deaths are in low-income or middle-income countries. We examined the association between anaemia and the risk of postpartum haemorrhage. METHODS: We did a prospective cohort analysis of data from the World Maternal Antifibrinolytic-2 (WOMAN-2) trial. This trial enrols women with moderate or severe anaemia giving birth vaginally in hospitals in Pakistan, Nigeria, Tanzania, and Zambia. Moderate anaemia was defined as a haemoglobin concentration of 70-99 g/L and severe anaemia as less than 70 g/L. Hospitals in each country where anaemia in pregnancy is common were identified from a network established during previous obstetric trials. Women who were younger than 18 years without permission provided by a guardian, had a known tranexamic acid allergy, or developed postpartum haemorrhage before the umbilical cord was cut or clamped were excluded from the study. Prebirth haemoglobin, the exposure, was measured after hospital arrival and just before giving birth. Postpartum haemorrhage, the outcome, was defined in three ways: (1) clinical postpartum haemorrhage (estimated blood loss ≥500 mL or any blood loss sufficient to compromise haemodynamic stability); (2) WHO-defined postpartum haemorrhage (estimated blood loss of at least 500 mL); and (3) calculated postpartum haemorrhage (calculated estimated blood loss of ≥1000 mL). Calculated postpartum haemorrhage was estimated from the peripartum change in haemoglobin concentration and bodyweight. We used multivariable logistic regression to examine the association between haemoglobin and postpartum haemorrhage, adjusting for confounding factors. FINDINGS: Of the 10 620 women recruited to the WOMAN-2 trial between Aug 24, 2019, and Nov 1, 2022, 10 561 (99·4%) had complete outcome data. 8751 (82·9%) of 10 561 women were recruited from hospitals in Pakistan, 837 (7·9%) from hospitals in Nigeria, 525 (5·0%) from hospitals in Tanzania, and 448 (4·2%) from hospitals in Zambia. The mean age was 27·1 years (SD 5·5) and mean prebirth haemoglobin was 80·7 g/L (11·8). Mean estimated blood loss was 301 mL (SD 183) for the 8791 (83·2%) women with moderate anaemia and 340 mL (288) for the 1770 (16·8%) women with severe anaemia. 742 (7·0%) women had clinical postpartum haemorrhage. The risk of clinical postpartum haemorrhage was 6·2% in women with moderate anaemia and 11·2% in women with severe anaemia. A 10 g/L reduction in prebirth haemoglobin increased the odds of clinical postpartum haemorrhage (adjusted odds ratio [aOR] 1·29 [95% CI 1·21-1·38]), WHO-defined postpartum haemorrhage (aOR 1·25 [1·16-1·36]), and calculated postpartum haemorrhage (aOR 1·23 [1·14-1·32]). 14 women died and 68 either died or had a near miss. Severe anaemia was associated with seven times higher odds of death or near miss (OR 7·25 [95% CI 4·45-11·80]) than was moderate anaemia. INTERPRETATION: Anaemia is strongly associated with postpartum haemorrhage and the risk of death or near miss. Attention should be given to the prevention and treatment of anaemia in women of reproductive age. FUNDING: The WOMAN-2 trial is funded by Wellcome and the Bill & Melinda Gates Foundation.