Practical considerations

Guidance on the development, validation and utilisation of food-frequency questionnaires has been compiled by Cade et al (2002).

Adaptation of an existing questionnaire
The development of a new FFQ is costly in terms of time and resources and therefore the use of an existing questionnaire might be appealing. However, several points should be considered before selecting an existing questionnaire for use in a study. These include:

Do the original objectives of the questionnaire meet the requirements of the new study?
Who were the original target population and how does this compare with the new target population?
Has the questionnaire been validated/calibrated and if so, in which population group and how?
Is the expertise to modify the analysis package available?

Food lists
Developing or adapting the food list from an existing FFQ is crucial to achieving a dietary assessment of acceptable accuracy. A comprehensive food list will enable a more detailed examination of nutrient intakes and diet-disease relationships although the huge variability in peoples’ diets means that the food list cannot be finite and some grouping of foods is necessary. It is important to consider carefully the level of grouping of foods. The aggregation of some foods with dissimilar eating patterns (e.g. tomatoes and tomato juice) can make it cognitively difficult for respondents to report frequency of consumption and may lead to an underestimation of intake. Also, for some food items, asking questions about single items can help respondents differentiate between similar foods with different nutritional profiles (e.g. full fat milk, semi-skimmed milk etc).
There may be situations in which the purpose of the FFQ is very specific and a shortened food list is preferable such as the assessment of intakes of calcium and other nutrients potentially involved in bone health (Taylor et al, 1998), but this limits the use of the questionnaire and the data derived to that use only.

Portion sizes
The potential contribution of questions on portion sizes is debatable for several reasons:

The ability of the population to accurately assess portion sizes is often limited
The variability in portion sizes in the population
The lack of availability of current average portion size data. While some data is available, this becomes outdated very quickly.

There is disagreement among researchers about the need to ask about portion size on FFQs (Willett, 1998) and it has been suggested that portion size questions do not necessarily improve the performance of the method (Molag et al, 2007). The source of variation in the diet may be greater in terms of differences in frequency of intake compared to differences in portion size between individuals. For some foods such as vegetables, it is difficult to estimate how much is eaten, particularly when they are a part of mixed dishes. In addition, the use of a medium reference portion for some foods is vague and respondents may have difficulty in judging their portion size in relation to the reference portion (Teufel, 1997).

Work with focus groups has suggested that providing objective portion sizes to respondents (e.g. ½ a cup) may be of limited use to individuals who rely on visual cues (Vuckovic et al, 2000). Some foods are particularly difficult to quantify, like chips (fries) or pasta. In the UK, unlike the USA, measures such as cups and tablespoons are not commonly used in cooking and will be less familiar to respondents. Providing instructions on the questionnaire about how to consider the food, such as a main versus a starter may help reduce difficulties (Vuckovic et al, 2000). Photographs showing examples of portions, like a slice or a tablepsoon, may be useful. Portion sizes listed may not correspond well with portions actually consumed and there is a debate about the use of sex-specific standard portions (Marks et al, 2006; Molag et al, 2007). In general, the collection of portion size data with FFQs is controversial (Coates et al, 1997). Portion sizes are also likely to change with age, and hence this needs to be considered if children or a wide age range are being studied. Cade et al (2002) found that the validity of FFQs was improved when subjects were able to describe their own portion size compared with no portion size specified (use of average portion weights to compute intakes), or portion size specified on the questionnaire.