Nutritional status in young children prior to the malaria transmission season in Burkina Faso and Mali, and its impact on the incidence of clinical malaria

This study was conducted in Houndé district, Burkina Faso, and in Bougouni district, Mali as part of a large trial which studied the effect of adding azithromycin (AZ) to SMC (with sulfadoxine/pyrimethamine (SP) and amodiaquine (AQ)) on deaths and hospital admissions in young African children, as described in detail elsewhere [20]. A separate analysis indicated that the addition of AZ had no impact on nutritional indicators [21]. In brief, an initial household census was conducted prior to the rainy season in 2014, to recruit eligible children who would be aged between 3 and 59 months on 1 August, 2014. All children were given a long-lasting insecticide-treated bed net. Children were randomized by household to receive four 3-day courses of AZ or placebo alongside SMC with SP + AQ at monthly intervals during the peak malaria transmission season (August to November). All doses were based on age and administered by trial staff. Newly eligible children aged 3–59 months were recruited at an additional census conducted in May 2015 and May 2016. Children with a chronic disease, a known allergy to SP, AQ or AZ, or who were taking cotrimoxazole were excluded.

Nutritional status of the cohort was assessed in two ways: by assessment of MUAC in all study children in 2015 (Nutritional Cohort 1), intended to maximize the sample size available, and by a more detailed nutritional survey in 2016 (Nutritional Cohort 2), intended to explore nutritional status in more detail among a smaller number of children. No nutritional status measurements were taken prior to the 2014 rainy season. In 2015, MUAC was measured prior to the rainy season at the time of annual census in all study children within the age range 3–59 months, apart from 1253 children who were in follow-up, but who did not attend the screening visit. In 2016, a random sub-sample of approximately 2000 study children per country, stratified on randomization group and age, were selected for a more detailed anthropometric survey by an independent statistician. These children had MUAC, weight and height measured at the start and end of the 2016 transmission season.

Infants 3–11 months of age received a combined 250 mg of sulfadoxine and 12.5 mg of pyrimethamine plus 75 mg of AQ on day 1 and received 75 mg of AQ on days 2 and 3 (Guilin Pharmaceutical, Shanghai, China). In addition, they were randomly assigned to receive either 100 mg of AZ or matching placebo on days 1, 2, and 3 (Cipla, Mumbai, India). Children 1–4 years of age received double these doses. All doses were based on age and administration was directly observed by trial staff.

Participants were weighed using a mechanical scale from SECA (Chino, CA, USA) with or without a tray, according to size, and weight was recorded to the nearest 0.1 kg. Recumbent length measurements were taken for children under 2 years of age and standing height measurements for children above 2 years using a stadiometer, recording the measurement to the nearest 0.1 cm. Length and height measurements were taken twice and the average was recorded. If the difference between the two measures was larger than 0.5 cm a third measure was taken and the two values with a difference smaller than 0.5 cm were used. MUAC was measured on the left arm to the nearest 0.1 cm with a MUAC tape.

Morbidity episodes were recorded passively at study health centres and at the hospital in each district. In Mali, to avoid missing malaria episodes in villages located further from the health centres, malaria morbidity was also recorded by trained community health workers using the same procedures as described below. Children who presented to a community health worker or outpatient clinic at a health centre when they were feeling unwell, had a rapid diagnostic test (RDT) for malaria performed. Malaria diagnosis was confirmed when the child had a fever (temperature ≥ 37.5 °C, and/or a history of fever within the previous 24 h) and a positive RDT for malaria or a positive blood smear (as described below). Free first-line treatment (artemether-lumefantrine) was offered on the basis of the RDT result. Blood smears were also taken from a systematic sample of study children (1 day per week in Burkina Faso and 1 week per month in Mali) for quality control of the RDT, and read by two readers. Slides which were discordant for either positivity or parasite density were read by a third reader, and the discrepancy resolved following the algorithm developed by Swysen et al. [22].

Data on morbidity episodes were recorded on case report forms and entered into an electronic database by two separate data entry officers using the DataFax system. To avoid double counting of multiple attendances for the same clinical episode, malaria episodes within 7 days of a previous episode were discounted. Person-time at risk and malaria episodes were included up to the date of censoring for children who exited the study or who died during the study period.

The number of children available for the 2015 MUAC survey was determined by the number recruited to the main trial, which was powered to detect a 25% difference with 90% power between the two study groups in the incidence of the primary trial end-point, death or hospital admission [20]. The required sample size for cross-sectional surveys (pre-season and post-season) in 2016 was driven by an outcome not analysed in the present study: the prevalence of SP and AQ resistance genotypes among study children. However, given the high incidence rate of malaria in the study area, a sample size of approximately 2000 children in each country from whom anthropometric measurements were taken in 2016 provided a high level of statistical power to detect even small differences in incidence rates according to pre-season nutritional status.

Different forms of malnutrition were evaluated in this study. Acute malnutrition was measured using weight-for-height (wasting), MUAC, and MUAC-for-age. Chronic malnutrition was measured using height-for-age (stunting). Being underweight (low weight-for-age) is a mixture of both chronic and acute malnutrition [23]. Nutritional indicators were calculated using standardized reference charts according to WHO guidelines [24, 25]. Weight-for-height was calculated for children above 65 cm, weight-for-length was calculated for children under 65 cm and this was combined into one weight-for-height/length variable.

A height-for-age Z-score below − 2 indicates stunting, a weight-for-age score below − 2 indicates underweight and a weight-for-height/length score below − 2 indicates wasting. Severity of nutritional indicators (stunting, wasting, underweight, low MUAC-for-age) was categorized into three groups: normal (z > − 2), moderately affected (− 2 < z < − 3), and severely affected (z < − 3).

Children with incomplete anthropometric data were included for the indicators that could be calculated, e.g., if weight, MUAC and age were collected, but height was missing, then weight-for-age and MUAC-for-age were calculated, but height-for-age and weight-for-height/length were not. Distributions of demographic and nutritional status variables were described as present prior to the transmission season. Distributions of z-scores for nutritional measures were compared between countries using the Wilcoxon rank-sum test. Prevalence of malnutrition according to each nutritional indicator was calculated by age and country. Associations between nutritional indicators and variables of interest were studied by creating logistic regression models with normally distributed random effects to adjust for clustering at the household level. Reliability of the estimates was checked by performing the quadrature check. Associations between nutritional status and malaria incidence were estimated using separate random effects Poisson models for each nutritional indicator adjusting for clustering in households and potential confounders. Variables which showed an association with both nutritional status and malaria incidence were considered to be potential confounders and were therefore included in the models. To be able to adjust for dosage as a covariate, the received dose was calculated by dividing the amount of drug administered by the child’s weight. For SP, less than 25 mg/kg was defined as below target and more than 70 mg/kg was defined as above target [26]. For AQ, less than 10 mg/kg was defined as below target and more than 15 mg/kg was defined as above target [27]. The dose of AZ was not investigated as a confounder because this was found to have no effect on malaria incidence in the main trial [20]. Intervention arm and SMC dose were included a priori because of their hypothesized effect on malaria risk. ‘Adjusted’ models included adjustment for age, gender, calendar month, intervention arm, and distance to health facility. ‘Fully adjusted’ models also included adjustment for SMC dose to be able to separately study its effect. Gender was a priori considered as potential effect modifier. Therefore, interaction between gender and nutritional status was examined by adding interaction terms to the multivariable model. All analyses were done in Stata/IC 16.

A total of 20,185 children were included in the analysis in 2015 (9,889 in Burkina Faso and 10,296 in Mali). The sub-sample for which anthropometric measurements were obtained in 2016 comprised 4,149 children (2098 in Burkina Faso and 2051 in Mali). The distribution of baseline and nutritional variables is presented in Table 1. Participants were equally distributed by gender and intervention arm. In Mali, 45.7% of children were aged between 13 and 36 months in both years. This percentage was similar in Burkina Faso: 44.3% in 2015 and 44.7% in 2016. Distance to nearest health facility was smaller in Mali with 43.9–44.6% living within 1 km of the nearest health facility compared to 17.8–20.6% in Burkina Faso.

Poor nutritional status was common prior to the rainy season in 2016. Figure 1 shows histograms of z-scores for stunting, wasting, underweight and low MUAC-for-age for both countries in 2016. The dashed lines shows the cut-off values for the severity categories used for malnutrition (normal is z-score above − 2; mild malnutrition is a z-score − 2 to − 3; severe malnutrition is a z-score > − 3). This shows that nutritional z-scores were approximately normally distributed around a negative z-score. The median z-scores were below 0 for all nutritional measures in both countries and lowest for height-for-age (median z-score − 1.2 in both countries). Apart from for height-for-age (p = 0.2), there was very strong evidence that nutritional status z-scores were lower in Burkina Faso than in Mali in 2016 for all other measures (p < 0.001).

The most prevalent form of malnutrition in Burkina Faso was being underweight, which affected 30.5% of children (95% CI 28.6–32.6). However, stunting was the most prevalent form of malnutrition in Mali affecting 27.5% of children (95%CI 25.6–29.5). Prevalence of malnutrition by age is shown in Fig. 2. The prevalence of low MUAC-for-age is lower than the other three nutritional measures in most age groups.

Using MUAC without adjusting for age and gender greatly overestimated the prevalence of malnutrition in younger age groups compared to the adjusted MUAC measure (see Additional file 1). MUAC-for-age was therefore used for all subsequent analyses. The correlation between the two main measure of acute malnutrition, MUAC-for-age and weight-for-height/length showed an r-squared of 0.10. When calculated for each year of age, results varied between 0.10 and 0.14.

Univariate analyses (comparing children with poor nutritional status, z-score < − 2, to well-nourished children, z-score > − 2)) showed that drug doses were consistently associated with all nutritional indicators, see Table 2 and Additional file 2. Country-specific results can be found in Additional files 3 and 4. Information on dose of anti-malarials given was not available in 2015 because the weight of children was not measured. Children with a low weight-for-age prior to the transmission season were more likely to subsequently receive a dose of SMC drugs above the target of 70 mg/kg for SP and 15 mg/kg for AQ, a consequence of the fact that SP + AQ was dosed by age and not by body weight in this study. This was also observed for other measures of nutritional status. The only variable without evidence for an association with any nutritional indicator was intervention arm (i.e., receipt of AZ compared to placebo alongside the SMC regimen) (0.25 < p < 0.84). Age was associated with all nutritional indicators, apart from low MUAC-for-age in 2016. There is strong evidence that children in Mali have lower odds of wasting (OR = 0.43; 95% CI 0.35–0.54; p < 0.0001) and underweight (OR = 0.58; 95% CI 0.49–0.70; p < 0.0001) compared to children in Burkina Faso. The odds for low MUAC-for-age were similar between the two countries in 2015, but in 2016 odds were lower for children in Mali compared to Burkina Faso (OR = 0.77; 95% CI 0.59–0.99; p = 0.05). Children living further away from the health facility tended to have worse nutritional status compared to children living within 1 km of the nearest health facility.

In Burkina Faso, the rate of malaria episodes in 2015 (1454 per 1000 person-years; 95% CI 1407–1501) was higher than in 2016 (675 per 1000 person-years; 95% CI 613–744). In Mali, the rate was lower in 2015 (1042 per 1000 person-years; 95% CI 1002–1083) compared to 2016 (1245 per 1000 person-years; 95% CI 1152–1347).

After full adjustment, there is little evidence of an association with any of the nutritional indicators (Table 3). Moderately underweight children in Burkina Faso seemed to be at a 1.3 times increased odds of clinic malaria (OR 1.27; 95% CI 0.98–1.64; p = 0.07). In Mali, moderately wasted children seemed to be at a 1.3 times increased odds of clinic malaria (OR 1.27; 95% CI 0.98–1.64; p = 0.07). Both of these effects were however not observed in severely affected children nor in the other country. The strongest protective association found was for severe stunting in Mali (RR = 0.81; 95% CI 0.64–1.01; p = 0.08).

The association between MUAC-for-age and malaria was similar in 2015 and in 2016 in Mali, suggestive of an increased risk in those with very low MUAC-for-age (fully adjusted RR = 1.46 (95% CI 0.86–2.47) in 2016 and adjusted RR = 1.20 (95% CI 0.96–1.51) in 2015, although the confidence intervals overlapped unity in both years. No interaction between gender and nutritional status was found.