Item selection

We searched MEDLINE  with the terms “ quality of life” or “ health related quality of life” or “health status” in combination with “ c(o)eliac disease” or c(o)eliac disease” from 1990- 31 decembre 2004. A list of items was formulated based on restrictions of HRQOL in celiac disease described in the literature including the non-validated questionnaires on HRQOL in CD such as the Burden of Illness Scale5, Self ministered Inventory for Celiacs 6, questions on CD-specific social restrictions used in a Canadian CD-survey7 and other studies exploring CD-specific problems of well- being8,9,10. Additionally, ten members of the DZG and four gastroenterologists completed an open ended questionnaire in which they listed items that have a significant impact on HRQOL in CD in their opinion. Furthermore some items based on the Inflammatory Bowel Disease Questionnaire IBDQ11, which assess emotional and social well-being of patients with inflammatory bowel diseases, were included in the list.

Questionnaire construction and pretesting

Appropriate questions for each item were constructed. Response options were specified using a seven-point Likert scale (ranging from 1 = the worst to 7 = the best function). Previous work suggested that seven point scales combine excellent responsiveness with patient understanding. Ten members of the DZG, selected by sex (50% female) and age classes (every decade between 18 and 80 years represented) completed the first preliminary version of the CDQ and commented on each question. The four experts and one psychologist, herself affected by CD, checked the preliminary version for readability and ease of completion. Furthermore the two translators of the German version of the CDQ into the English language checked for readability and ease of completion. Changes in the wording based on patients’ and experts’ comments were carried out and a second preliminary version of the CDQ was formulated.

Item reduction
A principal component analysis with a minimal Eigenvalue > 1 of the second preliminary version of the CDQ of the first 100 respondents of the survey was performed to select items and to guide placement of items into different factors (subscales). A scree plot was used to help decide on possible numbers of factors (subscales). For the number of factors suggested by the scree plot a varimax rotation was performed. The varimax rotation facilitates separation of factors with an optimal balance of variance explained by each factor. A factor should explain at least 3% of the total variance and have at least 3 items. The items were related to a factor by the following criteria: Factor loading coefficient ? 0.4, communality ? 0.16 and difference factor loading to other factors > 0.1212.