2.1. Study Design, Participants, and Setting This cross-sectional study was nested in the larger project of 15-year follow-up of MINIMat trial (Maternal and Infant Nutrition Interventions in Matlab, ClinicalTrials.gov identifier ISRCTN16581394). MINIMat was a factorial randomized trial with 6 arms that primarily examined the effects of food and micronutrient supplementation for pregnant women on hemoglobin level at 30 weeks’ gestation, birth weight, and infant mortality. The procedural details and results have been reported elsewhere. A total of 4436 pregnant women were recruited between November 2001 and October 2003. This resulted in 3267 singleton live births with valid birth anthropometrics, constituting the MINIMat cohort that has been repeatedly followed up. The 15-year follow-up was carried out from September 2017 to June 2019. A total of 2465 (75.5% of the eligible) adolescents completed the household survey. Over a period of 15 years some loss to follow-up was inevitable. Those with higher-educated mothers and those belonging to wealthier households were less likely to take part in subsequent follow-ups of the trial [30]. These differences, however, were small and unlikely to distort the findings. Results section presents participant flow into current study in detail. Matlab is a rural sub-district, located about 55 km to the southeast of the capital city of Dhaka. The International Center for Diarrheal Disease Research, Bangladesh (icddr,b), formerly the South-East Asia Treaty Organization (SEATO) Cholera Research Laboratory, has been operating a Health and Demographic Surveillance System (HDSS) in Matlab since 1966. Matlab has a low-lying, deltaic topography crisscrossed by the river Gumti and its branches. Rice farming is the main occupation of people, except for a few villages that depend on fishing. A three-month long agricultural lean period usually occurs between September and November.
2.2. Data Collection Trained interviewers with at least 12 years of formal schooling identified eligible adolescents using unique MINIMat identification numbers. Mother/guardian-adolescent dyads were interviewed at their residences with a pre-tested, structured questionnaire containing pre-coded questions. Insights from previous follow-ups and qualitative study undertaken at the initial phase of the 15-year follow-up guided questionnaire preparation and context-specific adaptation of standard instruments. The interviewers received comprehensive training on questionnaire survey techniques and data collection using tablet computer devices. Supervisors conducted random field visits to check data collection procedure in real-time. A web portal also allowed off-site monitoring of the progress in data collection and quality of collected data.
2.3. Assessment of Dietary Diversity Dietary diversity was assessed at individual level through a 24 h recall of consumed foods, using locally adapted version of a standard instrument endorsed by the Food and Agriculture Organization of the United Nations (FAO). This 10-food-group instrument has been applied widely and validated for adolescents as well. It offers an improvement over 2011 FAO guidelines by prescribing a cut-off to differentiate adequate and inadequate DD. In our adapted version, we separated “Fish” from “Meat, poultry, and fish” to avoid masking of the plausibly lower consumption of meat by an anticipated higher consumption of fish. Whereas families in Matlab mostly buy meat from the market, fish remains more accessible due to its abundant natural availability in Matlab. Having fish in the same food group with meat and poultry may, therefore, obscure affordability issues and potential association with socioeconomic status. Of note, separate grouping of fish and other type of meats is also suggested in the 2011 FAO guideline. We also collapsed “Pulses” with “Nuts and seeds” into the food group called “Legumes, nuts, and seeds” to better capture existing culinary norms of the setting. This grouping is also supported by the 2011 FAO guideline.
2.4.1. Socioeconomic Status (SES) SES was derived from household asset score; a continuous, numerical variable computed by principal component analysis of data on ownership of a range of durables (e.g., mobile phone, radio, television, refrigerator, bicycle, etcetera); access to electricity and sanitary latrine; and nature of fuel used. Asset scores were converted to tertiles (lower, intermediate, and upper) representing a relative measure of household SES: the poorest, the middle-status, and the richest, respectively.
2.4.2. Household Food Security We employed the Household Food Insecurity Access Scale (HFIAS) to distinguish food insecure households from food secure ones. This is a 9-item, experience-based scale with a 1-month recall period and has been validated for use in LMICs. In accordance with the guideline , food secure households were those that either had not endured any food insecurity experiences or rarely worried about running out of food. The rest were categorized as food secure.