2.1. Study design and participants

We conducted a prospective cohort study. This involved two groups, each consisting of 77 infants aged 4–8 weeks (the age when future interventions to treat infant u6m SAM will be anticipated to begin; Mwangome, Fegan, Fulford, Prentice, & Berkley, 2012). One group comprised infants with SAM as defined by current WHO guidelines: weight‐for‐length Z‐score (WLZ) <−3 and/or bilateral nutritional oedema (WHO, 2013); the other comprised age‐ and sex‐matched infants who were not severely malnourished (non‐SAM) defined as WLZ ≥−2 to <2 and mid‐upper‐arm circumference (MUAC) ≥125 mm. Exclusions were infants from twin/multiple pregnancies and those with obvious congenital anomalies that could affect feeding (e.g., cleft lip or palate).

Ethical approval was obtained from the Institutional Review Board of International Centre for Diarrhoeal Disease Research, Bangladesh (icddr, b; PR14112). Written informed consent was obtained from the infants' parents or legal guardians. The study is registered with the ISRCTN trial registry: http://www.isrctn.com/ISRCTN12494235.

2.2. Variables

The primary outcome was the proportion of infants who died or who had SAM (defined as per WHO criteria as WLZ <−3 and/or MUAC <115 mm and/or oedema) at age 6 months (180 completed days). Key secondary outcomes were changes in and absolute values of weight‐for‐length z‐score (WLZ), weight‐for‐age z‐score (WAZ), and length‐for‐age z‐score (LAZ).

Maternal mental health status was assessed using the WHO Self Reporting Questionnaire 20 (SRQ 20; WHO, 1994). Mothers who had a high total score (≥13) or who answered “Yes” to question no. 17 (“Has the thought of ending your life been on your mind?”) were referred to the outpatient psychiatry department at the nearby Sher‐e‐Bangla Medical College Hospital, Barisal, for appropriate management.

At cohort end line (age ≥ 6 months), infants' vital status, anthropometry, and dietary history were repeated.

2.3. Data collection procedures

2.4. Sample size estimation and statistical analysis

Sample size for the prospective cohort study was estimated for comparing outcomes in the exposed group (SAM) and unexposed group (non‐SAM) assuming that 25% of the participants in the SAM group would have SAM at 6 months of age and 6.3% of the participants in the non‐SAM group would have SAM at the same time point (the prevalence of SAM in infants u6m in Bangladesh during designing the survey; NIPORT, Mitra, & ICF, 2013). With 5% level of significance and 80% power, 77 infants were required per group, assuming approximately 25% loss to follow‐up at the end of 6 months. The following formula was used for sample size estimation: [(pA1−pA/κ+pB1−pB]z1−α/2+z1−β/pA−pB2 [available at http://powerandsamplesize.com/Calculators/Compare‐2‐Proportions/2‐Sample‐Equality].

Anthropometric z‐scores (standard deviation scores) were calculated from raw age, sex, length, and weight data using WHO, 2006 growth standards and WHO Anthro software (WHO, 2006). LAZs and WLZs were calculated for infants who were ≥45 cm long (WHO 2005). WAZs were calculated for all infants. Those with extreme values were excluded from analysis following standard WHO “cleaning rules,” WAZ <−6.0 or WAZ >+5.0, LAZ <−6.0 or LAZ >+6.0, and WLZ <−5.0 or WLZ >+5.0 SD (Crowe, Seal, Grijalva‐Eternod, & Kerac, 2014).

Potential explanatory variables were grouped under household characteristics, maternal characteristics, and infant characteristics. Differences between groups were tested using chi‐squared tests for proportions, Student's t test for normally distributed continuous variables, and Mann–Whitney U test for skewed continuous variables. Multivariable logistic regression was performed for adjustments of different variables and potential risk factors for SAM. Multiple linear regression analysis was also used to adjust the potentially confounding variables when different variables were compared for test of significance. The potential confounders were considered to be the age of the infants at the time of enrolment, monthly income, and maternal education. All the tests were considered as significant at P < 0.05. Statistical analysis was done with the Statistical Package for the Social Sciences (IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp.).

3.1. Sociodemographic characteristics

Infants in the SAM group were over a week younger at enrolment than those in the non‐SAM group (5.1 ± 1.2 weeks vs. 6.5 ± 1.2 weeks, P = 0.001). Median household monthly income was significantly lower (P = 0.007) in SAM group, and mothers were less educated (P = 0.005) also. Fewer households in the SAM group had electricity (P = 0.013). Other reported sociodemographic characteristics were comparable between groups (Table 1).