ASSESSMENT OF POSTPARTUM TIME-DEPENDENT DISEASE RISK IN CASE-CONTROL STUDIES - AN APPLICATION FOR EXAMINING AGE-SPECIFIC EFFECT ESTIMATES

Among subjects who have experienced a biological event, such as menarc he, menopause or a delivery, one cannot distinguish the effects of tim e since the event from age at the event due to the linear dependency a mong these time variables and age at study ('current age'). This is a well-known problem that also exists in the determination of the short- and long-term influence of childbirth on subsequent disease risk, sin ce one must take into account in the analysis both current age and age at delivery. We describe an approach to assess in case-control studie s the effect of a full-term pregnancy on time-dependent disease risk b y including nulliparous women in the analysis and considering current age as a modifier of the effect of age at delivery. One then uses curr ent age-specific odds ratio estimates that compare uniparous to nullip arous women to examine whether the relative rate of disease varies ove r time after a delivery. Analytic options include stratified analysis and modelling with interaction terms for unconditional or conditional logistic regression analysis. As an example, we have applied this anal ysis to a large case-control study that utilized record linkage betwee n the Cancer Registry and the Fertility Registry of Sweden and that do cumented a transient increase in breast cancer risk after a childbirth , followed by a long-term reduction in this risk.