The cases are also prevalent rather
than incident, with the inherent consequence of some recalled items
being influenced by knowledge of the diagnosis. In general we did
not conduct analyses of variables likely to be particularly affected by
this (eg current mental state), but our findings with respect to recent
behaviours require cautious interpretations. The other usual challenges
to validity are less likely to be troublesome here: Table 1
showswell the lack of selection bias within the cohort; and whilewe
do not have precise information on some standard risk factors for
breast cancer, we have quite a deal that is relevant, and confounding
is not usually a major issue with this disease.