The methods used to assess diet in large studies such as EPIC must be of good technical quality, feasible and acceptable to local populations. From substantial work carried out in the years before EPIC it was accepted that there was no single best method. Therefore, it was necessary for various methods to be tested and evaluated in pilot studies and these were carried out by collaborators between 1990 and 1992. These extensive methodological studies were designed to test the validity of diet and lifestyle questionnaires, which were then adjusted and developed in the light of the results obtained. Three dietary methods were adopted: a self-completed dietary questionnaire, an interview-based dietary questionnaire or a food frequency questionnaire (FFQ, where the participants estimate their average frequency of intake of a list of foods over the previous 12 months) combined with a seven-day record (diary). The questionnaires contained common diet and lifestyle questions across each centre (core questions) and some additional centre-specific questions, which reflect particular interests of individual scientists.

Being a multi-centre study it is essential that the dietary measurements obtained from the centre-specific questionnaires are comparable. In order to calibrate, or adjust, the questionnaire measurements from each questionnaire to a common standard a specially designed computer program (EPIC-SOFT) was developed. This program was adapted for each participating country and translated into 9 languages. The program enabled the research interviewer to pursue a standardized interview to determine precisely what the participant had consumed in the previous 24 hours (24 hour diet recall). 8-10% of the participants in each centre were interviewed in this way. Common methods were used to classify and export the data from the interviews from each centre. Since there is little experience in the use of such standardization methods a working group on dietary patterns was established with the aim of reviewing the results and exploring analytical methods.

Statistical methods The EPIC data are complex and their full use requires the adaptation and development of advanced statistical methods. Intensive work to develop appropriate statistical analysis techniques has been on-going and a statistical working group meets regularly to discuss current methodological problems. For example, statistical methods have been developed to correct for bias in estimates of dietary components obtained from the questionnaires, making the cohort-specific estimates more comparable between the study centres.