In this paper we have described the development of an FFQ and a food composition database for the Black Zimbabwean population. We have also provided a copy of the FFQ and food composition table for the benefit of other researchers in the field.
Our sample included a wide cross-section of Zimbabwean society – men, women, rich and poor, urban and rural. We did this to get a list of nutrient rich foods that were eaten by a wide cross-section of people in the community that would help in categorizing people by nutrient intake. There was a wide spread of intakes of porridge and white bread and are examples of discriminating foods (Table ).
We tried to phrase the questions in the FFQ so that the respondent could picture the food we were asking about in his or her mind. For example the portion size of meat in Zimbabwe is considerably smaller than that in western countries. Matemba (fish) is eaten in very small quantities almost like a condiment. We took these factors into account when designing the questions and analyzing nutrient content. The questions were short, used average portion sizes, and categories of intake so that the interviewer simply checked a category to estimate frequency of intake, and did not have to estimate portion size. As a result its administration (by a trained interviewer) took on average 10 minutes. The instrument has face validity. For example, sadza is the staple food in Zimbabwe and almost all the participants reported that they ate it at least 5 times a week. We are conducting a formal validation of this instrument by comparing nutrient and food intakes assessed from the FFQ to multiple 24-hour recalls.
By using the USDA food composition database as our major source of information we ensured that nutrient estimates were available for most foods, and the assays to estimate nutrients were current [3
]. By using the Zimbabwean food composition database as a guide we were able to select the food item from the USDA food composition database that was closest in nutrient content to the corresponding Zimbabwean food. The choice was probably reasonable as the nutrient estimates from the two methods were very similar for most nutrients. The discrepancies, particularly for vitamin A and folate, were probably because certain foods were not analyzed for these nutrients and hence there were missing values in the Zimbabwean table, or old methods of nutrient analysis were used. The methods to estimate vitamin A and folate from foods have changed in the USDA nutrient database [3
], but those for macronutrients have not.
Instead of the backward selection procedure to develop a food list, we collected a 24-hour dietary data, supplemented with input from local nutritionists. Such an approach has also been used for the development of FFQs [2
]. Moreover the variation of food intake in low income countries is low [2
] so it is unlikely that we missed out any important food in the FFQ. The participants in this study were persons over 35 years of age. The FFQ could likely be used among younger adults as well because it probably captures most of the foods eaten by this group as well but this caveat should be borne in mind when extending its use. Another limitation was that we did not have nutrient contents of certain foods, for example, caterpillar. We overcame this by using the closest category (sushi caterpillar) that was in the ESHA food composition table to that food. These assumptions would inevitably be the source of some errors in nutrient estimation.
The discrepancies in nutrient estimates obtained from the same FFQ data but different food composition tables underscores the need to have standard methods, not only to develop the FFQ, but also food composition tables. While there is much literature on methods for FFQ development there is much less information on food composition database development. We have described a standardized process of FFQ and food composition database development for a Black Zimbabwean population.