Thresholds of socio-economic and environmental conditions necessary to escape from childhood malnutrition: A natural experiment in rural Gambia

Background: Childhood malnutrition remains highly prevalent in low-income countries, and a 40% reduction in under-5 year stunting is WHO’s top Global Target 2025. Disappointingly, meta-analyses of intensive nutrition interventions reveal that they generally have low efficacy at improving growth. Unhygienic environments also contribute to growth failure, but large WASH Benefits and SHINE trials of improved water, sanitation and hygiene (WASH) recently reported no benefits to child growth. Methods: To explore the thresholds of socio-economic status (SES) and living standards associated with malnutrition, we exploited a natural experiment in which the location of our research centre within a remote rural village created a wide diversity of wealth, education and housing conditions within the same ecological setting and with free health services to all. A composite SES score was generated by grading occupation, education, income, water and sanitation, and housing and families were allocated to 5 groups (SES1 = highest). SES ranged from very poor subsistence-farming villagers to post graduate staff with overseas training. Nutritional status at 24 m was obtained from clinic records for 230 children and expressed relative to WHO Growth Standards. Results: Height-for-age (HAZ) and weight-for-age (WAZ) Z-scores were strongly predicted by SES group. HAZ varied from – 0.67 to – 2.23 (P < 0.001) and WAZ varied from - 0.90 to - 1.64 (P < 0.001), from SES1 to SES5, respectively. Weight-for-height (WHZ) showed no gradient. Children in SES1 showed greater dispersion so were further divided in a post hoc analysis. Children resident in Western housing on the research compound (SES1A) had HAZ = + 0.68 and WAZ = + 0.36. The residual gradient between those in SES1B and SES5 spanned only 0.65 Z-score for HAZ (- 1.58 to - 2.23) and was not significant for WAZ or WHZ. Conclusions: The large difference in growth between children in SES1A living in Western-type housing and SES1B children living in the village, and the very shallow gradient between SES1B and SES5, implies a very high SES threshold before stunting and underweight will be eliminated. This may help to explain the lack of efficacy of the recent WASH interventions and points to the need for what is termed 'Transformative WASH'. Good quality housing, with piped water into the home, may be key to eliminating malnutrition. The UK Medical Research Council has maintained a research presence in the poor rural West Kiang region of The Gambia, since 1948. The field station comprises a self-sufficient 4-acre site with housing, laboratories, clinic and administrative buildings contained within a perimeter fence. All housing is of ‘Western’ standard with tiled floors, metal roofs and suspended ceilings, fully plumbed kitchens and bathrooms with flushing toilets and hot and cold water to basins and showers, screened and louvered windows, gas cooking, fans and air conditioning. Water and electricity are available constantly and are not charged to staff, thus allowing unlimited usage. The keeping of livestock is strongly discouraged in the compound. Originally sited on the edge of the village, it has now become engulfed by new village compounds as the population has expanded. Most villagers rely for their livelihoods on subsistence farming, occasionally supplemented by some petty trading, and increasingly by remittances from family members who have migrated to urban areas or abroad. The village has a primary and middle school and an Islamic school. Uptake of education has increased in the past two decades, initially male dominated but now more equitable due to universal free education for girls. The MRC compound and the village are both supplied by clean tube-well water. Villagers obtain their water from a series of 30+ standpipes distributed as equitably as possible around the village. Water is potable as it emerges from the taps. Subsequent contamination during collection and storage is possible. For over 35 years, we have provided an exceptional level of free healthcare for villagers and staff (summarised in Box 1). Thus, availability of healthcare is constant across all socio-economic and educational levels, and for staff and villagers alike, and uptake is only determined by parental health-seeking behaviours. All women receive regularly scheduled ante-natal care, and their infants have a post-natal check-up and are called to well-baby clinics at 6 week, 3 m and then at 3 m intervals until 2 years of age. All children receive their full schedule of Extended Programme of Immunisation (EPI) vaccines and vitamin A supplementation according to WHO guidelines. The greatest variance in SES was associated with MRC employment. Using the Kiang West Demographic Surveillance System (DSS) [14], a list of all children with anthropometric data (length and weight measurements) taken at 24 ± 4 m of age and born between January 1993 and December 2009 was created. The sample size was determined by the number of MRC employees who had children at 24 m within the study window. All MRC employees on the list were approached to participate. This yielded 51 families and 133 children. This list was supplemented with data from 129 village children selected at random from our DSS database and matched by year of measurement. Children with any condition known to affect growth were excluded. A total of 98 parents were interviewed during the data collection period. Parental SES was evaluated using a modified version of a 43-item questionnaire originally developed to study the relationship between parental SES and child mortality elsewhere in The Gambia [15]. The questions were grouped according to occupation, income and possessions, education, access to water and sanitation, and housing. In the absence of indications to the contrary, each of these factors was assumed to have an equal impact on child growth; thus, each category was given the same weight of a maximum of 10 points. Each question within each category was then assigned an appropriately weighted score based on information from senior staff at the MRC about the perceived socio-economic value of items in the questionnaire. The aggregate SES score for each set of parents was then determined by adding up the 5 separate scores obtained for each category, giving a total maximum of 50 points. Initially the total parental scores were assigned a priori to 5 SES groups as follows: SES1 = > 25; SES2 = 21–25; SES3 = 17–21; SES4 = 14–17; and SES5 = < 14. It was subsequently noted that SES1 showed a wide dispersion of anthropometric values so this group was further divided as follows: SES1A = > 25 points and resident in Western-style housing within the MRC compound; and SES1B = > 25 points and resident in the village. The headline descriptors of each SES category are listed in Table 1. Typical SES characteristics of the five pre hoc determined groups and subsequent post hoc division of group 1 In this population, anthropometric status declines rapidly in the first and second years of life and then remains somewhat stable before recovering slightly in later childhood [16]. We therefore obtained clinic records for heights and weights at 24 ± 4 m. Weight (in minimal clothing) and recumbent length (height) were recorded by trained anthropometrists using standard techniques and regularly calibrated apparatus. Appropriate weight data was found for 262 children, but there were 32 missing heights. The data were then restricted to those that had full anthropometric values available; yielding a final sample size of 230. HAZ, WAZ and WHZ were calculated using WHO Anthro software (version 3.2.2) based on the 2006 WHO Child Growth Standards [8]. Within the narrow age range selected for this study, there was no influence of age on any of the anthropometric measures. There was an influence of sex, so this has been included in the analysis model. Stunting was defined as HAZ < − 2, underweight as WAZ < − 2 and wasting as WHZ < − 2. Parental heights were obtained from our DHSS database. There were 2 missing maternal heights. Multi-level linear analysis was performed on anthropometric attainment at 24 m according to the 5 (later 6) SES groups with adjustment for parental heights and with intra-household clustering accounted for by including mother’s ID on the basis that children of this age live with their mothers. Date of measurement was included in the model to test for any secular drift but was not significant so was not included in the final analysis. All analyses were performed using the ‘Linear Models’ function in DataDesk version 7.0.2 (Data description Inc., Ithaca, NY).