Menu
Menu
Menu
Menu
Objectives: With a fourth of all under-five children affected, stunting remains one of the biggest health challenges worldwide. Even though the main underlying factors are known, the exact pathways to stunting varying in affected regions, and interventions thus need to be tailored to the local contexts. This study aimed assessing and comparing factors associated with stunting in two understudied sub-Saharan urban contexts with some of the highest stunting prevalence globally: Bangui, Central African Republic (~ 36%) and Antananarivo, Madagascar (42%). Methods: We performed a case–control study on 175 + 194 stunted and 237 + 230 non-stunted control children aged 2–5 years and matched for age, gender and district of residency. Factors associated with stunting were identified using a standardized, paper questionnaire delivered by trained interviewers. Statistical analysis was done using logistic regression modelling. Results: In both sites, formal maternal education lowered the risk of being stunted and restricted access to soap, suffering of anaemia and low birth weight were associated with higher risk of stunting. Short maternal stature, household head different from parents, diarrhoea and coughing were associated with an increased risk and continuing breastfeeding was associated with a lower risk of stunting in Antananarivo. Previous severe undernutrition and dermatitis/ fungal skin infections were associated with higher and changes in diet during pregnancy with lower risk of stunting in Bangui. Conclusions: Our results suggest maternal education, antenatal care, iron supplementation and simple WASH interventions such as using soap and infection control as general and breastfeeding (Antananarivo) or better nutrition (Bangui) as area-specified interventions.
The AFRIBIOTA study (Vonaesch et al., 2018b) is a case–control study for stunting in children aged 2–5 years in Bangui, CAR and Antananarivo, Madagascar. The choice of the two study sites was based on the high stunting prevalence and the fact that the consortium worked previously together on a project on diarrhoeal diseases (Breurec et al., 2016; Randremanana et al., 2016). Inclusion criteria were HIV-negative children, neither suffering from acute malnutrition nor from any other severe disease, living in the 6st, 7st or 8th district of Bangui or in two neighbourhoods of Antananarivo (Ankasina or Andranomanalina Isotry). Included children were admitted to the hospital for sample collection and anthropometric measurement. Assuming an alpha-error of 0.05%, a power of 80% and an expected 20% exposure in the cases, we needed at least 169 children per group, hence 676 individuals in the two countries. The final sample size analysed in this study was of 836 children (Fig. 1). Stunted and control children were matched according to age in years, gender, neighbourhood and season of inclusion (dry or wet season). Recruitment took place between December 2016 and March 2018 in Antananarivo and between January 2017 and May 2018 in Bangui. Detailed recruitment procedures are given in the Supplementary methods and the study protocol (Vonaesch et al., 2018b). Flow-chart of the children included in the final analysis. Data is summarized for children included in A Antananarivo, Madagascar and B in Bangui, Central African Republic Height was measured by trained personnel to the nearest 0.1 cm in a standing position using collapsible height boards (Antananarivo: ShorrBoard Measuring Board, Maryland, USA; Bangui: height board provided by UNICEF); weight was measured to the nearest 100 g using a weighing scale (Antananarivo: KERN, ref. MGB 150K100 and EKS, Inter-équipement Madagascar; Bangui: weighting scale provided by UNICEF) and mid-upper arm circumference (MUAC) was measured using commercial MUAC tape (provided by UNICEF) to the nearest 0.1 cm. Cut-offs were based on the official cut-offs defined by WHO (Onis, 2006). A standardized, paper questionnaire in French that was translated ad hoc to the local languages Sangho and Malagasy was used in both study sites and data was entered in double in an Access database. The questionnaire included information about children’s age, gender, family structure, socioeconomic status indicators, sanitary indicators, data about the mother’s pregnancy and child’s and family nutrition and feeding practices. A wealth index was created based on the minimal set of assets, leading to a separation of subjects in three distinct groups in a principal component analysis (PCoA). Details of the wealth index are given in the extended methods. Each child was further examined for comorbidities and venous blood was collected. Complete blood count, C-reactive protein (CRP) and ferritin levels were measured. Ferritin levels were corrected for systemic inflammation (Thurnham et al., 2010), haemoglobin values were adjusted for altitude (Centers for Disease, 1989; Sullivan et al., 2008) and anaemia was defined as less than 110 g/l, according to WHO criteria (OMS, 2011; Onis, 2006). A dietary diversity score (DDS) was calculated based on a 24 h recall (World Health Organization, 2007a, 2007b). Mother’s nutritional status was based on the Body Mass Index (BMI). Non-pregnant mothers were classified by BMI categories as defined by the WHO and pregnant mothers according to the categories proposed by Ververs (Ververs et al., 2013). Clinical parameters such as cough (observed and reported by the mothers), dermatitis (as visible dermal affections of various origins diagnosed by a medical doctor), diarrhoea (> 3 loose stools/day), and tooth decay were assessed during a clinical examination. Previous episode of severe acute malnutrition and perceived low birth weight was based on mother’s recall. The statistical analysis was performed with Stata 13. Significance level was fixed for all analyses at 0.05 and all tests were performed bilaterally. Categorical variables were expressed as percentages; quantitative variables were expressed as a mean (± Standard Deviation) or median (interquartile range). The stunted vs. non-stunted groups were compared using Chi2 or Fisher Exact test for qualitative variables and the Student t test or the Mann–Whitney U test for quantitative variables. All variables were assessed in a bivariate analysis. Factors associated with stunting in bivariate analysis with a P value of < 0.2 were checked for potential confounding factors and interactions and then included in a backward logistic regression. As we did not get a perfect matching for age, gender and season of inclusion, these variables were forced in the multivariate model. Results are reported as adjusted OR with 95% CI, corrected for age in years, gender, season of inclusion and country of origin.
