Limitations and strengths of the study
Our study is not free from limitations. One of these is information bias, which may result from e.g. self-reporting age, age at marriage, and education. BMI is a crude index which does not consider the distribution of fat and which can vary from individual to individual. This study used cross-sectional data which could not confirm the cause and effect
relationships. The results may be biased as some variables, such as smoking, physical activity and dietary habits, are not
adjusted for underweight and overweight indicators in the multinomial logistic regression [27]. Although propensity score
matching and inverse weighted estimator are two possible solutions to reduce selection bias, we did not apply them
keeping in mind that many authors frequently applied multiple logistic regressions in similar studies. Generally multiple logistic regressions are easier to conduct than propensity score matching for addressing several outcome variables. The ideal number of children could be related to the number of children actually born, because women with more children may report a greater number of children as ideal to justify their fertility [58]. Moreover, all the indicators should be critically checked for intra-urban disparity (e.g. slum versus non-slum), since the rural-urban disparity remained almost the same over a long period. Unfortunately, relevant data were not available to enable an analysis of the slum versus non-slum disparity. Perhaps the increasing slum population in urban areas is a factor in this regard. It is likely that the overall impact of interventions in urban areas is underestimated for the non-slum population and overestimated for the slum population. Although divisional disparities are found, these are not focused like the other two equity indicators. It should also be mentioned that Bangladesh is a country with huge population growth. Generally trends are valid provided the population growths are more or less constant for different sectors with appropriate sampling weights for all surveys. Briefly, the strengths of the study are related to the use of several indicators based on multiple data sets. This gives an idea of existing disparities, which should be urgently addressed.