The unadjusted RR and IRR values were obtained from log-binomial regression models (RR) and Poisson regression models (IRR) with the stratification factors of site and HIV status (known positivevsnewly diagnosed HIV infection) included as covariates. The adjusted analyses were done using several other prespecified baseline covariates in addition to HIV status and site, including gravidity, malaria status, socioeconomic status, season (average rainfall in the last 6 months before delivery), and malaria transmission intensity by study site (based on the prevalence of malaria at enrolment, continuous). The term malaria refers to malaria infection. Clinical malaria refers to malaria infection detected by malaria rapid diagnostic test (histidine-rich protein 2 or Plasmodium lactate dehydrogenase) or microscopy in conjunction with documented fever (>375°C) or reported fever in the past 48 h. IRR=incidence rate ratio. RR=risk ratio. *Haemoglobin at delivery, or otherwise in the third trimester if the measurement at delivery was unavailable.