Dietary assessment - Food frequency questionnaire
Food frequency questionnaires (FFQ) are designed to assess habitual diet by asking about the frequency with which food items or specific food groups are consumed over a reference period (e.g. 6 months or a year). FFQs may be based on an extensive list of food items or a relatively short list of specific foods. The foods listed should be a) major sources of a group of nutrients of particular interest or b) foods which contribute to the variability in intake between individuals in the population, and c) commonly consumed in the study population. The length of the list of foods can range from about 20 to 200 items. Questionnaires can be self-administered using paper or wev-based formats, or interviewer administered, either face-to-face or telephone interview. The frequency of food consumption is assessed by a multiple response grid in which respondents are asked to estimate how often a particular food or beverage is consumed. Categories ranging from ‘never’ or ‘less than once a month’ to ‘6+ per day’ are used and participants have to choose one of these options.
FFQs were originally designed to provide descriptive qualitative information about food-consumption patterns. These then developed to provide nutrient information by ascribing an ‘average’ portion size. Some FFQs, known as semi quantitative, include portion size estimates where respondents are asked to indicate the frequency of consumption of specific quantities of foods (e.g. ½ a cup, ¾ cup etc) or to assess their usual portion size based on a specified measure.
Some FFQs may include supplementary questions on cooking methods and specific types of fat, bread, milk and additions to foods e.g. salt. Brand name information may also be requested e.g. the variety of margarine used, though this can require considerable effort to code and analyse. Other FFQs may have an open-ended section where respondents may record consumption of other foods not included on the food list. This ensures that the total diet of the individual is captured and may also help to identify those whose diet is very unusual, for whom the FFQ may not be appropriate.
FFQs have been used widely in epidemiological studies investigating links between diet and disease. For this purpose it is more important to rank the intake of individuals relative to others in the population (e.g. high, medium, or low intake) or as quantiles (e.g. fifths of the distribution of intake) than to determine the absolute intake. Two well-known FFQs are the Harvard or Willett questionnaire (Willett et al, 1985) and the Block questionnaire (Block et al, 1986). These are both American FFQs and have been used in many studies; the Willett questionnaire or an adaptation of it has been particularly widely used (e.g. Stein, 1992; Tucker et al, 1999; Lamb et al, 2007; Barclay et al, 2008; Linos et al, 2008). Participants can be ranked into broad categories of intakes of specific foods and food components or of nutrient intake which needs to be calculated using in-house programmes based on appropriate food composition tables. In large epidemiological studies, data on diet from FFQs are compared with specific disease outcomes like cancer diagnosis, or risk factors for disease such as total or HDL-cholesterol levels. FFQs may also be used to identify food patterns associated with inadequate intakes of specific nutrients using statistical methods such as principal components analysis (e.g. Flood et al, 2008; Nettleton et al, 2008; Northstone et al, 2008; Cutler et al, 2009), cluster analysis (e.g. Wirfält et al, 2009) or reduced rank regression (Vujkovic et al, 2009). As responses to FFQs are standardised, data can be entered and analysed in a comparatively short period of time, often in an automated process, allowing dietary data on a large number of people to be collected relatively inexpensively. There is also less need for nutritional expertise in data entry, though for incorporation of additional foods and interpretation of results nutritional researchers are strongly recommended. The usefulness of a particular FFQ is limited if it is poorly designed and does not contain a comprehensive food list representative of the eating habits of the entire population under study. Modifications to an FFQ require considerable effort as the data entry and in-house nutrient analysis programmes must be modified.
Since FFQs are often designed to assess the ranking of intakes within a population, they cannot be relied on to produce reliable estimates of absolute intake. Over-estimation is common, particularly for foods eaten less often or for foods perceived as 'healthy' such as fruit and vegetables. There is some evidence that over-estimation increases with the length of the food list. Measurement error with FFQs is likely to be greater than that with other methods such as diet diaries, and it has been suggested that there may have been false negative results in studies of diet-disease associations (Day et al, 2001; Lietz et al, 2002; Subar, 2003; Bingham et al, 2003; Cade, 2004).