Interpretation of results is difficult when there is much missing data, and it is best to take precautions to minimise the problems. For example, questions about sexual interest, ability or activity may cause problems. In some clinical trials these questions might be regarded as of little relevance, and it may be reasonable to anticipate similar low levels of sexual problems in both treatment arms. This raises the question of whether it is advisable to exclude these potentially embarrassing items so as to avoid compromising patient compliance over questionnaire completion.
Fayers et al. (1997b) reported that in a wide range of UK MRC cancer trials, approximately 19% of patients left blank a question about "(have you been bothered by) decreased sexual interest". Females were twice as likely as males to leave this item blank. Possible interpretations are that some patients found the question embarrassing, or that they did not have an active sexual interest before their illness and therefore regarded the question as not applicable. Perhaps wording should be adapted so that lack of sexual activity before the illness (not applicable) can be distinguished from other reasons for missing data.
Was this article helpful?