Anaemia in hospitalised preschool children from a rural area in Mozambique: a case control study in search for aetiological agents

The study was conducted at the Centro de Investigação em Saúde de Manhiça (CISM) and the Manhiça District Hospital (MDH), Southern Mozambique. The CISM runs a continuous Demographic Surveillance System (DSS) and round-the-clock morbidity surveillance at the MDH. The area has been described in detail elsewhere [10]. Briefly, malaria was main cause of death in the area (21.8% of cases) followed by pneumonia (9.8%), HIV/AIDS (8.3%) and diarrhoea (8%) [11]. Malaria transmission is perennial with marked seasonality and mainly due to Plasmodium falciparum [12]. The HIV prevalence in adults (18–47 years of age) was 37.5% [13] and 1.4% in children ≤11 years old [14]. The HIV-mother-to-child-transmission rates in the first month and at 12 months of age were 9% and 27%, respectively [15]. Almost half (47%) of children visited in the out-patient clinic presented some degree of undernutrition being 6% of them severely malnourished [16].

A case–control study was designed to recruit 450 hospital anaemia cases and 450 community controls. All children with clinical criteria for hospital admission with ages of 1–59 months, no history of blood transfusion in the previous month and residence in the CISM DSS area were offered participation in the study. Hb levels were determined with the Hemocue HB 201⁺ system (Änghelom, Sweden) in children whose guardians signed the informed consent and those with Hb <11 g/dl were recruited as cases. Recruitment was consecutive, from October 2008 to August 2010, running continuously during working hours (8:00–16:00). In order to avoid the confounding effect of malaria seasonality, a maximum of ten cases per week were recruited until reaching 450 cases.

Community controls were randomly selected from the DSS among children between 1 and 59 months of age, and visited at home by project personnel. Hb levels were determined by Hemocue in children with no history of blood transfusion in the previous month whose guardians had signed the informed consent. Those with Hb ≥11 g/dl were invited to a study visit at the MDH. Due to the high prevalence of anaemic children in the community (91%) (unpublished data), only 289 community controls were recruited. Enrolment of cases and controls occurred concurrently.

The recruitment of hospital controls, which would have allowed ruling out possible confounding factors associated with being severely sick, was not possible due to the extremely high prevalence of anaemia among admitted children in this setting. Similarly, due to the high prevalence of anaemia in the community, it was not possible to enrol the planned sample size of controls. The implications of this limitation on the results would depend on the assumptions made as to the proportion of controls exposed to the factor.

The results of the clinical examination and socio-demographic information were entered into standardised questionnaires. Four mL of venous blood was drawn, blood smears and filter paper blood smears were made. A maximum of 1 ml/kg was drawn from children. A stool and a urine sample were collected whenever possible.

Anaemia was categorised as severe (Hb <5 g/dl) or moderate (Hb ≥5 g/dl to <11 g/dl). Fever was defined as axillary temperature ≥37.5 °C. Bacteraemia was defined as at least one bacterium isolated in blood (excluding common contaminants listed in Table 3). Pf infection was defined as the presence of asexual parasites in blood by microscopy or qPCR. Clinical malaria was defined as Pf infection plus fever or a history of fever in the preceding 24 h. Detection of parasites by qPCR and a negative blood smear defined submicroscopic Pf infections. Hyperparasitaemic Pf infection was defined as >100,000 parasites/μl of blood. EBV and PV-B19 infections were defined as detection of the virus in blood by qPCR.

Wasting was defined as a weight-for-height/length Z-score (WHZ) of < −2 standard deviations (SD), stunting as height-for-age Z-score (HAZ) < −2 SD and underweight as weight-for-age Z score (WAZ) < −2 SD [22]. Albumin deficiency was defined as <34 g/l (laboratory reference value), and prealbumin deficiency as <0.142 g/l in children <6 months of age, <0.120 g/l in children 6–12 months of age and <0.108 g/l in children 13–59 months of age [23]. Folate deficiency was defined as <3 ng/ml (laboratory reference value), vitamin A deficiency as <20 μg/dl [24], and vitamin B12 deficiency as <200 pg/ml [25]. ID was defined as the ratio of soluble transferrin receptor to log ferritin (TfR-F index) >1.5 if CRP <1 mg/dl, and >0.8 if CRP ≥1 mg/dl. This biomarker was identified as the best predictor of ID with an accuracy of 71%, using the bone marrow iron content as the gold standard [9]. In order to compare ID prevalence between cases and controls, this ID definition was used instead of bone marrow iron content, as bone marrow examination was undertaken in anaemic children only [9]. Inflammation was defined as a CRP level ≥1 mg/dl. Ineffective erythropoiesis was a Reticulocyte Production Index <2 [26].

Data were double entered using Microsoft Visual FoxPro 5.0 (Microsoft Corp., USA) and analysed using STATA 12 (STATA Corporation, College Station, USA). Crude comparisons were performed by Chi-square or Fisher’s exact tests for proportions, and the Student’s t test or Wilcoxon rank sum test for means or medians, respectively. Multivariable logistic regression analysis was performed by stepwise selection including all variables with p-value <0.05 in the crude models, using significance levels of 0.05 and 0.10 for addition to and for removal from the model, respectively. Among the variables with no significant association with anaemia in the univariable model only vitamin B12 and G6PD deficiencies were included in the multivariable model based on biologic plausibility. Intestinal parasitic infections, schistosoma infection, haemoglobinopathy type E, sickle cell trait and β-thalassaemia trait were not included because of their low prevalence. No children in the control group had clinical malaria or hyperparasitemia, so these variables were not included in the regression models. Due to the difficulty in obtaining sufficient volume of blood samples in some children, α-thalassaemia was analysed in only 258 subjects and it was also excluded from the regression analysis. Missing observations were excluded from the multivariate model. Attributable fraction (AF) were estimated using Bruzzi’s approach [27]. The variance of adjusted AF (AAF) was approximated by Taylor series to first derivative terms (delta method) [28].