Objective: To determine whether changes in the side effects frequency information in a patient information leaflet affects the interpretation of the frequency, the recall of this information and the number of side effects reported.
Design: Randomized controlled trial.
Setting: A field experiment was carried out in which three versions of a package insert concerning diclofenac were randomized and tested in 33, 30 and 34 patients respectively. Version 1 contained the original side effects information. In version 2 the frequency descriptors were manipulated (lowered, e.g. 'sometimes' was replaced by 'seldom'). This was because a preliminary investigation had shown that patients make higher assessments of side effect frequencies than the leaflet writers want to convey. In version 3 an introduction was added with general information on side effects.
Results: The frequency assessments differed enormously between subjects. Especially the lower frequencies were virtually impossible to convey by means of words. The lower-ranking frequency descriptors did not only result in lower frequency assessments, but also led to lower recall score for side effects mentioned in the text and to a smaller number of side effects actually reported by patients. The only effect of the introduction was to lower the frequency assessments for the side effects following the introductory paragraph.
Conclusions: Side effects with high assessed frequencies were read with more attention. The advantages and disadvantages of numerical frequency statements need to be discussed.