2.1. Data Sources and Compilation Like most countries, food supply derived from the food balance sheets is a reliable and perhaps a very good option available to follow and analyze the trends of dietary changes at the national level. In this study, due to the lack of a long-term national dietary intake dataset, food availability data in Bangladesh were obtained from the FAO’s food balance sheets documented in the Food and Agriculture Organization Corporate Statistical Database (FAOSTAT) from 1961, the baseline year, to 2013, the latest available year. The food balance sheet data for Bangladesh was downloaded as comma separated values (.csv) files from the FAOSTAT database. The food balance sheet is compiled yearly and is an international resource of a country’s food availability during a certain period. Food availability is derived from production, supply, usages and wastages of food and provides information on the apparent consumption instead of actual consumption in the diet. Assessing the pattern and trend in the availability of food components is a useful tool for the assessment of changes in diet and can be applied to define dietary patterns. Food availability in a country is calculated based on: the amount of food exported, used for animals and in agriculture, or wasted, and is deducted from the amount of food produced and imported; the annual amount of available energy from food items is then divided by the total population of a country in the same period. Per capita, food availability data are represented as both kcal/day/person and kg/day/person in the FAOSTAT database. For the food intake, we obtained the dataset from FAOSTAT of the fully adjusted food supply with kg per capita per year values for each food item. We then divided the values in kg per capita per year by 365.25 to obtain values in kg per capita per day and multiplied this amount by 1000 to obtain g per capita per day for each food item.
2.2. Operational Definition Since the data came from national food balance sheets rather than from a nationwide dietary survey, these intake data refer to “average food and nutrients available for consumption.” This does not indicate the food actually consumed, rather it indicates national availability for food (g/day/person). Hence, in the remainder of this article “apparent food consumption or intake” should be read as “food available for consumption” or “national availability of food”.
2.3. Food Groups and Recommended Intake For this study, dynamics and trends in apparent food intake or food availability were evaluated for eleven food items including cereals, starchy roots, pulses, fish, eggs, meat, vegetables, fruits, milk, vegetable oils, and sugar. Moreover, the food items that were grouped into these eleven food groups. In the meat group, the availability of meat and offal for consumption were added and combined in one group. We have not calculated the trends for animal fat intake because of its extremely low level of availability; from 1961 to 2013, animal fat intake was less than 1.0 g/day/person except in the year 1978 (about 1.01 g/day/person). Moreover, the nuts availability trend was also not analyzed for its extremely low availability (less than 3.0 g/day/person) in the diet. The recommended food intakes in the diet for the Bangladeshi population were derived from the desirable dietary pattern (DDP) for the Bangladeshi population.
2.4. Trends Analysis To analyze the temporal trends and to identify significant changes in trends, we used Windows-based statistical software, the Joinpoint Regression Program (version 4.6.0.0, National Institute of Health, Bethesda, MD, United States), for performing the joinpoint regression by using joinpoint models. This software tests whether an apparent change in a trend is statistically significant (p <0.05). With the joinpoint regression analysis, it is possible to identify years when a significant change in the linear slope of the trend is detected over the study period. The best fitting points, called “joinpoints,” are chosen when the rate changes significantly. The analysis starts with the minimum number of joinpoints and tests whether one or more joinpoints (in this study up to 5) are significant and must be added to the model. To describe linear trends by period, the estimated annual percent change (APC) is then computed for each of those trends. Moreover, we calculated the average annual percent change (AAPC) as a summary measure of the trends over the period for each of the food items. The average annual percent change calculated as a geometric weighted average of APCs of various segments was used to quantify the trends of food intake changes in the diet of the Bangladeshi population over the entire available period of the FAOSTAT data from 1961 to 2013.