Risk factors for moderate acute malnutrition among children with acute diarrhoea in India and Tanzania: a secondary analysis of data from a randomized trial

This secondary analysis was conducted using data from the Zinc Therapeutic Dose Trial (ZTDT) (NCT03078842). ZTDT was an individually randomised, parallel-group, double-blind, controlled trial of three doses of supplemental zinc (5, 10 and 20 mg) conducted among children aged 6–59 months at sites in Delhi, India and Dar es Salaam, Tanzania. Details of the methodology of the trial, and the primary results that showed non-inferiority of lower dose zinc on diarrhoea duration and stool output, have been published elsewhere [12, 13]. Briefly, this trial was conducted from January 2017 to April 2019 and included 4,500 children with diarrhoea lasting < 72 h. After enrolment, children were followed for 60 days with initial close follow up for 15 days to ensure compliance to therapeutic zinc. Subsequently, illnesses were assessed at days 30, 45, and 60. Participant anthropometry was assessed at enrolment (baseline) and 60 days after enrolment.

Participants were recruited from outpatient health facilities in Sangam Vihar, a resettlement colony on the outskirts of South Delhi and Harsh Vihar, a semi-urban locality in North East District, Delhi, India. In Dar es Salaam, Tanzania, recruitment sites were outpatient health facilities in a densely populated peri-urban area (Temeke Municipal Hospital, Mbagala Round Table, and Mbagala Rangi Tatu).

Body weight, length/height, and mid-upper arm circumference (MUAC) were measured during screening and 60 days later during a scheduled clinic visit with study staff. Body weight was measured using an electronic baby scale (ADE, ADE GmbH & Co, Hamburg, Germany) or weighing machine (SECA, SECA GmbH & Co, Hamburg, Germany) with reading increment of ± 10 g. If the child was dehydrated, the weight was obtained after rehydration. Length/height was measured by a stadiometer with reading increment of ± 0.1 cm. MUAC was measured in participants’ left upper arm, midway between the olecranon and acromion processes using standard WHO/UNICEF measuring tapes with a reading increment of ± 0.1 cm. Daily calibrations of the anthropometry equipment were conducted to ensure that the equipment produced accurate measurements. Weight measurement tools were calibrated with known weights of 3, 5, 10, and 15 kg. Stadiometers were calibrated using measuring rods of known length measuring 50 cm and 75 cm. Readings of all anthropometric measurements were made in triplicate and the mean of the three readings was used.

The primary outcome measure, MAM, was defined as weight-for-height Z score ≤ -2 and > -3 or mid-upper arm circumference < 12.5 and ≥ 11.5 cm according to WHO standards [7]. We performed separate analyses for the presence of MAM at baseline (prevalent MAM) and the development of MAM during follow-up (incident MAM). Assessment of the continuous variable weight-for-length Z scores (WHZ) at baseline and during follow-up was also assessed.

We included variables collected in the parent trial to conduct an exploratory analysis of factors theoretically associated with the development of MAM, both prevalent and incident. In addition, we excluded prevalent MAM and SAM at baseline from analyses of risk factors for incident MAM. Improved water was defined as water from sources like piped water that, by nature of their construction or through active intervention, are protected from outside contamination. Improved sanitation facilities were defined as those that hygienically separate human waste from human contact and include flush or pour-flush [14]. Dysentery was defined as acute diarrhoea with blood or mucous in stool. Household wealth index was defined as a measure of a household’s standard of living. A country-specific household wealth index was constructed by means of a principal components analysis of these factors: household ownership, household assets, drinking water source, and sanitation.

Descriptive statistics were used to summarize baseline characteristics of caregivers and children who participated in the trial. Frequencies were presented for categorical variables and mean ± standard deviation (SD) for continuous variables. Univariable log-binomial regression analyses were used to estimate risk ratios (RRs) and 95% confidence intervals (CIs) for predictors of the outcome of interest to produce relative risks. Variables significant at the P < 0.20 were retained in a multivariable log-binomial regression analysis model to calculate adjusted RRs. P values < 0.05 were considered statistically significant. Analyses were performed using SAS software Version 9.3 (SAS Institute, Cary, NC, USA).

This study’s results should be interpreted in the context of several limitations. Since this was not a planned secondary analysis, the original trial was not powered a priori to primarily detect changes in anthropometry among enrolled children. Additionally, there are other potentially unmeasured confounders that were not accounted for in this study. Specifically, we did not measure dietary intake or the use of ready to use supplemental feeds, which may have prevented the development in MAM in some children. However, ready to use supplemental feeds are infrequently administered to children without malnutrition or with MAM in low- and middle-income countries including India and Tanzania [7, 15]. Furthermore, local ecology or geographical environment, residential culture, maternal knowledge of diarrhoea treatment, and government policies combating malnutrition and diarrhoea which are factors that may have a significant impact on development of MAM were not collected in our study and thus were not factored into our regression model. Finally, the incidence of MAM after enrolment (2.5%) was very rare, making our analyses of incident MAM likely underpowered.