Chronic disease outcomes after severe acute malnutrition in Malawian children (ChroSAM): a cohort study

Background Tackling severe acute malnutrition (SAM) is a global health priority. Heightened risk of non-communicable diseases (NCD) in children exposed to SAM at around 2 years of age is plausible in view of previously described consequences of other early nutritional insults. By applying developmental origins of health and disease (DOHaD) theory to this group, we aimed to explore the long-term effects of SAM. Methods We followed up 352 Malawian children (median age 9·3 years) who were still alive following SAM inpatient treatment between July 12, 2006, and March 7, 2007, (median age 24 months) and compared them with 217 sibling controls and 184 age-and-sex matched community controls. Our outcomes of interest were anthropometry, body composition, lung function, physical capacity (hand grip, step test, and physical activity), and blood markers of NCD risk. For comparisons of all outcomes, we used multivariable linear regression, adjusted for age, sex, HIV status, and socioeconomic status. We also adjusted for puberty in the body composition regression model. Findings Compared with controls, children who had survived SAM had lower height-for-age Z scores (adjusted difference vs community controls 0·4, 95% CI 0·6 to 0·2, p=0·001; adjusted difference vs sibling controls 0·2, 0·0 to 0·4, p=0·04), although they showed evidence of catch-up growth. These children also had shorter leg length (adjusted difference vs community controls 2·0 cm, 1·0 to 3·0, p<0·0001; adjusted difference vs sibling controls 1·4 cm, 0·5 to 2·3, p=0·002), smaller mid-upper arm circumference (adjusted difference vs community controls 5·6 mm, 1·9 to 9·4, p=0·001; adjusted difference vs sibling controls 5·7 mm, 2·3 to 9·1, p=0·02), calf circumference (adjusted difference vs community controls 0·49 cm, 0·1 to 0·9, p=0·01; adjusted difference vs sibling controls 0·62 cm, 0·2 to 1·0, p=0·001), and hip circumference (adjusted difference vs community controls 1·56 cm, 0·5 to 2·7, p=0·01; adjusted difference vs sibling controls 1·83 cm, 0·8 to 2·8, p<0·0001), and less lean mass (adjusted difference vs community controls −24·5, −43 to −5·5, p=0·01; adjusted difference vs sibling controls −11·5, −29 to −6, p=0·19) than did either sibling or community controls. Survivors of SAM had functional deficits consisting of weaker hand grip (adjusted difference vs community controls −1·7 kg, 95% CI −2·4 to −0·9, p<0·0001; adjusted difference vs sibling controls 1·01 kg, 0·3 to 1·7, p=0·005,)) and fewer minutes completed of an exercise test (sibling odds ratio [OR] 1·59, 95% CI 1·0 to 2·5, p=0·04; community OR 1·59, 95% CI 1·0 to 2·5, p=0·05). We did not detect significant differences between cases and controls in terms of lung function, lipid profile, glucose tolerance, glycated haemoglobin A1c, salivary cortisol, sitting height, and head circumference. Interpretation Our results suggest that SAM has long-term adverse effects. Survivors show patterns of so-called thrifty growth, which is associated with future cardiovascular and metabolic disease. The evidence of catch-up growth and largely preserved cardiometabolic and pulmonary functions suggest the potential for near-full rehabilitation. Future follow-up should try to establish the effects of puberty and later dietary or social transitions on these parameters, as well as explore how best to optimise recovery and quality of life for survivors. Funding The Wellcome Trust. The original prospective cohort consisted of 1024 patients admitted for treatment of SAM at the Moyo nutrition ward at Queen Elizabeth Central Hospital in Blantyre, Malawi, from July 12, 2006, to March 7, 2007. All patients were treated in accordance with the national guidelines at the time.9 This treatment involved admission based on National Center for Health Statistics (NCHS) references and initial inpatient stabilisation for all children by use of therapeutic milk followed by nutritional rehabilitation at home with ready-to-use therapeutic food. Detailed baseline data were collected as part of the PRONUT study.10 Median age at admission was 24 months (IQR 16–34). In a follow-up study at one year post-discharge (FuSAM),11 477 (47%) children from the original cohort remained alive; the surviving children from this study form the case group for the present follow-up, the ChroSAM study. For comparison, we aimed to recruit one sibling control and one community control per child in the case group. The sibling control was the sibling closest in age to the case child, limited to children aged between 4 and 15·9 years. The community control was defined as a child living in the same community, of the same sex, and aged within 12 months of the case child. We selected community controls randomly by spinning a bottle at the case child's home and enquiring door to door, starting from the nearest house to where the bottle pointed. Written informed consent was obtained from the children's parent or guardian. Any children who had ever been treated for SAM were not eligible as controls. Additional assent was sought from the child if they were older than 13 years. Ethical approval for the study was granted by the Malawi College of Medicine Research and Ethics Committee (reference P.02/13/1342), and the University College London Research Ethics Committee (reference 4683/001). Our outcomes of interest were anthropometry, body composition, lung function, physical capacity (ie, hand grip strength, step test, and physical activity), school achievement, and blood markers of NCD risk. Exposures included previous admission for SAM, socioeconomic status, HIV, and maternal education. Anthropometric assessments were done in accordance with the guidelines by Lohman and colleagues12 and WHO13 and were subject to quality control, which involved two members of the trained study team taking independent readings.13 Body composition was measured with skinfold thickness measured at the biceps, triceps, subscapular, and suprailiac sites, and by use of bioelectrical impedance analysis (BIA) with a Quadscan 4000 device (Bodystat, Douglas, Isle of Man). Usually, BIA outputs are converted to total body water and fat-free mass via population-specific empirical equations.14 In the absence of a population-specific equation, it is possible to assess relative hydration and lean mass by use of raw BIA values adjusted for height.15 Results are presented as resistance index (R/height) and reactance index (Xc/height). Physical activity was measured in a subset (n=78) of children with Actilife GT3X accelerometers (ActiGraph, Pensacola, FL, USA). Muscle strength was measured with a Takei Grip-D device (Takei, Niigata, Japan). Physical capacity was measured with the iSTEP (incremental step) test.16 Lung function was measured with spirometry on an Easy-On PC device (ndd Medical Technologies, Zürich, Switzerland); quality grades were applied in a blinded manner by a senior respiratory physiologist at the Institute for Child Health of University College London (London, UK).17 Spirometry outcomes were forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC), and the FEV1-to-FVC ratio, all expressed as Z scores.18 More detailed methods are available in the appendix (pp 9–11). Venous blood samples to measure glycated haemoglobin (HbA1c), lipid profile, and full blood count were taken after 12 h of fasting. Salivary cortisol was assessed about mid-morning. A subset of children (n=60) underwent an oral glucose tolerance test with 1·75 mg per kg bodyweight of Polycal glucose powder (Nutricia, Dublin, Ireland).19 Data collectors were not blinded to the case or control status of the children because of the logistics of the study. The HIV status of children and mothers was established from health passports; if status was not known or if the test was done before the age of 18 months, HIV testing was offered by a trained counsellor. Puberty was recorded as a binary variable, as reported by the participant or guardian (onset of menarche in girls, voice change in boys). Socioeconomic status was derived from asset scores calculated with questions taken from the Malawi Demographic Health Survey.20 Sample size was predetermined by the cohort size and survival. Controls were more difficult to recruit than were cases because they had no previous personal connection with the study team and were restricted to the number of eligible children in the family and community. However, with 320 cases, 217 sibling controls, and 184 community controls, we calculated that the study had a post-hoc power of at least 90% to detect a Z score difference of 0·5 between the cases and controls based on reference data for growth and lung function outcomes at the 5% level of statistical significance. Given that Z scores of 0·5 are considered clinically significant in lung function testing and that the FuSAM study showed a difference in weight-for-age Z score (WAZ) of 0·55 and height-for-age Z score (HAZ) of 1·13 between cases and sibling controls,11 we deemed our sample size to be satisfactory. Post-hoc calculations suggested that the sample size was adequate for all outcomes except physical activity (steps per day), which was underpowered. We did statistical analyses with Stata version 12.1. To calculate the WAZ, HAZ, and body-mass index (BMI)-for-age Z score (BAZ), we used WHO AnthroPlus, which includes new WHO reference curves for children aged 5–19 years.21 We analysed BIA data with BIA vector analysis, which graphically illustrates the association between resistance and reactance measures (phase angle) to give an indication of variability in hydration versus lean mass.22 We converted spirometry outcomes to Z scores with the Global Lung Function Initiative (GLI) African-American reference values, which adjust for height, age, sex, and ethnicity.18 In the main analysis, we compared each type of control with case children by use of simple and multivariable linear regression analysis. We included age, sex, HIV status, and socioeconomic status as potential confounders in all multivariable regressions. We also included puberty as a potential confounder for body composition; puberty and sitting height were additional potential confounders for lung function outcomes. We used ordered logistic regression to analyse differences between cases and controls for time completed in the exercise test and school grade achieved, which we used as a measure of educational attainment since children in Malawi do not move to the next grade until they pass, irrespective of their age. Because anthropometry was collected for some sibling controls during the 1-year follow-up study,11 we used repeated measures mixed regression models to analyse longitudinal growth between cases and sibling controls between 1 and 7 years post-discharge. In all analyses, we deemed a p value of less than 0·05 as showing a statistically significant difference between groups. The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.