The study was conducted between May 2002 and January 2005 at the Kenya Medical Research Institute, Kilifi District Hospital, Kenya. Patients attend this hospital from a well mapped Demographic Surveillance Site (DSS) in which a census is conducted three times a year. Malaria is endemic in Kilifi District, where falciparum parasitaemia at the time of the study was present in over 30% of children in the community and 46% of hospital admissions428. Although many children are admitted with a primary diagnosis of severe malaria complicated by some degree of anaemia, the etiology of anaemia in this population is complex429. The study area is one of the poorest in Kenya with 40% of children <5 years having anthropometric measures of undernutrition, half with biochemical markers of iron deficiency, 14% of children within the DSS have sickle cell and almost 60% have α⁺ thalassemia deletion genotypes13.

Demographic and clinical data were recorded on admission using a standard proforma. Children were eligible for inclusion in this study if they manifested all of the following features: P. falciparum parasitaemia (at any density), a clinical feature of severe malaria (prostration, coma or respiratory distress) and metabolic acidosis (base deficit of >8). Such children are at the greatest risk of death, including the subgroup of children with symptomatic severe malaria anaemia (Hb<5 g/dl in association with signs of cardiorespiratory compromise).

In accordance with WHO guidelines whole blood transfusion (20 mls/kg of whole blood) was administered (without diuretics) for children admitted with or subsequently developing a Hb <4 g/dL or 4–5 g/dL if associated with respiratory distress. Those with these features at admission were transfused immediately (early transfusion group) while those without these features at admission were transfused later if they developed these features (delayed transfusion group). Children with features of decompensated shock meeting above transfusion criteria (early transfusion group) were administered pre-transfusion management with 10 mls/kg fluid while awaiting transfusion26. Children with a Hb>5 g/dl and respiratory distress were enrolled into intervention trials and received volume expansion with saline, albumin or gelofusine 24252627. Following resolution of features of severe malaria and once children were able to take and retain oral medication, anti-malarial treatment was completed with a full course of oral anti-malarial therapy. In accordance with local national guidelines, the children received a single dose of sulphadoxine-pyrimethamine (SP) up to December 2003, and SP plus amodiaquine thereafter when the national anti-malarial policy had changed.

Discharge from hospital was consequent upon establishing both clinical and parasitological recovery (malaria slide negative). Children were discharged with any level of haemoglobin ≥ 4 g/dL as long as they did have respiratory distress. At discharge all survivors received a course of iron sulphate (20 mg/kg/day in three divided doses for 28 days which was continued to complete three months), folate (2.5/5 mg once day for body weight <10/>10 kg respectively for 28 days and continued to complete three months), and mebendazole (100 mg twice daily for three days). Children reviewed before 21 days or after 35 days were excluded from the primary analysis of haematological recovery due to temporal effects on haemoglobin recovery, but included in survival analysis. Follow-up review included a full clinical assessment, anthropometric measurement, full blood count, and a malaria slide. Investigation of parasite genotyping was not performed (to distinguish re-infection from recrudescence). Survival of study children was also obtained by linking the clinical data to the census data routinely done in the DSS reference to find out the long term progress of the study participants. In this study, information from census done in May 2005 was used.

Data was analysed using Stata version 8.2. Patients were retrospectively classified as transfused or not transfused and by compliance with follow-up. In children attending follow-up within one month of post discharge admission, discharge and follow-up characteristics of children who were transfused according to WHO guidelines were compared to those of their relatively less sick counterparts who did not received blood transfusion, because they did not fulfil the WHO criteria, by comparing means and standard deviations for continuous variables and Chi² or Fisher's exact test for proportions. The haemoglobin at follow-up was further categorized into <9.3 g/dL (WHO criteria for anaemia) and >9.3 g/dL with those who failed to attain a Hb of 9.3 g/dL classified as anaemic according to WHO criteria19. Mantel-Haenszel methods were then used to calculated crude odds ratios for association of various factors with anaemia at follow-up (i.e. Hb <9.3 g/dL) and a logistic regression analysis done to determine independent associations. All variables with P-value < 0.25 at univariate analysis were included in a logistic regression analysis to identify independent determinants of haemoglobin recovery and only those with a P value < 0.05 were retained in the model. An a priori decision was made to include age in each model due possible confounding effect of age on incidence of anaemia in children with malaria. Formal survival analysis was not performed, instead survival was described as the proportion of cases that were ascertained at discharge, follow-up and long-term follow-up by DSS.

Of the 213 survivors, 166 (78%) patients returned for follow-up assessment: 158 (74%) children were reviewed between 21–35 days post discharge and eight (4%) were reviewed outside this interval. Out of these eight children, three attended earlier due to malarial fever (2) or malnutrition (1) and five were well, but attended on the wrong date. Haematological recovery was assessed in 158 survivors who subsequently attended follow-up within 22 to 35 days post-admission

Baseline data are summarized in Table 1 with respect to transfusion status. The mean admission Hb was 5.0 g/dL [SD 1.9] in those receiving a blood transfusion while in those not transfused, the mean Hb was 8.3 g/dL [SD 1.9]. Those transfused were younger, median age 25 months (SD 14–35) versus 36 months (SD 27–47); P < 0.001) and had poorer nutritional status, mean mid-upper arm circumference (MUAC) 13.6 cm [SD: 1.3] versus 14.5 cm [SD: 1.5]; (P < 0.001). Geometric mean parasitaemia was similar in the transfused group 2.6 × 10⁵ (95% Reference Range (RR) 1.7, 3.5) and non-transfused group 2.3 × 10⁵ (95% RR 1.7, 2.9) (p = 0.62) as were sex, prevalence of hypoglycaemia or mean corpuscular volume (MCV) (Table 1). However, presence of respiratory distress, base deficit and mean lactate levels were all significantly higher in the group of patients who received a blood transfusion (Table 1).

Significant changes were observed in mean Hb between admission and discharge which increased on average by 1.4 g/dl (SE 0.2) in those receiving a blood transfusion and fell by a mean of 1.5 g/dl (SE 0.1) of Hb in the non-transfused group. Although there were clear differences in Hb at admission there were no differences in discharge Hb in the two groups (F = 1.4; P = 0.23) (Table 1). At discharge from hospital, 37/213 (17%) remained severely anaemic (Hb 4–5 g/dL) and 100 (47%) children had moderately severe anaemia (Hb 5–7 g/dl). All survivors were aparasitaemic at time of discharge.

Haematological recovery was identical in both groups with no significant differences in the mean follow-up haemoglobin between those transfused and those not transfused (p = 0.25) (Table 1). Overall, there was an 46% mean increase of post-admission Hb at follow up, with 74% attaining an effective haematological recovery (Hb >9.3 g/dL) and 89/158 (56%) achieving values greater than 10 g/dl at follow up (Figure 1). For the 47 (22%) non-attendees, there were no differences in the mean haemoglobins at admission (p = 0.16), discharge (p = 0.34), and transfusion status or other marker of severity from those that attended follow-up.

The presence of malaria parasitaemia at follow-up and young age were the only factors independently associated with follow-up haemoglobin. Nutritional status (determined by mid-upper arm circumference (MUAC) and weight), mean corpuscular volume (MCV) and review interval did not have a significant influence on the follow-up haemoglobin.

Presence of malaria parasitaemia was found to be the most significant determinant of haemoglobin recovery in patients following admission with severe malaria (Figure 2). For those with parasitaemia, mean haemoglobin was 8.8 g/dL (SD: 1.5) compared to a mean of 10.5 g/dL (SD: 1.3) in aparasitaemic children; P < 0.001. This was not influenced by mean haemoglobin at admission or discharge, age, sex, review interval, MCV and MUAC as they were not significantly different in the two groups. Of the patients with parasitaemia at follow-up, 15/35 (43%) were transfused while 20/35(57%) were not transfused (P = 0.48). The odds ratio of a low haemoglobin (Hb <9.3 g/dL) in cases with parasitaemia compared to no malaria was 4.00 (1.61–8.42) (p = 0.002) and this remained significant even after controlling for age, treatment received, MUAC, MCV and sex (p = 0.001). Anti-malarial treatment used at discharge had a significant influence on the follow-up mean haemoglobin. Of the 158 patents included, 27/158 (17%) had sulfadoxine-pyrimethamine(SP)-amodiaquine combination while 131/158 (83%) had SP alone. Children who had a combination therapy with SP-amodiaquine had significantly higher mean haemoglobin (10.9 g/dL SD 1.4) at the time of follow-up compared to those who had SP alone (9.9 g/dL SD1.5) (p = 0.005).

Age <24 months was found to be a significant determinant of haemoglobin at admission (p < 0.001), discharge (p = 0.008), and at follow-up (p = 0.01). Younger children (age <24 months) had a lower Hb at admission (F = 15; p < 0.001), discharge (F = 5; p = 0.008) and at review (F = 4; p = 0.03). Children under 24 months old accounted for 31/61 (51%) of all transfusions; 14 given for SSMA and 17 delayed transfusions.

Medium term survival was ascertained in the 166/213 (78%) patients who returned to follow-up, and the demographic surveillance was used to trace survival in those who failed to attend follow-up review. Of the 47 non-attendees 32 (68%) came from within the study area, of whom we could confirm longer term survival on 30. Twenty one were reported to be still alive and living in the census area, 9/32 were still alive when they migrated out of the study area and two were not traceable. This increases the overall medium term survival to 196 (92%) children with severe and life threatening malaria who were managed conservatively with respect to blood transfusion in accordance with WHO transfusion guidelines.