Beyond the twinning effect: invited commentary on "Levels of maternal serum alpha-fetoprotein (AFP) in pregnant women and subsequent breast cancer risk".

In 1987, Jacobson et al. (1) suggested that mothers of twins might be protected against subsequent breast cancer. This prediction was based on laboratory studies of the effects of alpha-fetoprotein (AFP) on the growth of human breast cancer implanted in animals, combined with the well-known approximate doubling of the levels of maternal serum AFP during twin versus singleton pregnancies. This theory was put to the test quite promptly in the setting of the Cancer and Steroid Hormone Study. The results provided support for the hypothesis that a twin pregnancy is associated with a reduced risk of breast cancer (1); nonetheless, a specific role for AFP could not be inferred with confidence, because twinning is only a surrogate for AFP and twin pregnancies differ from singleton pregnancies in a number of ways other than in AFP levels. In the correspondence (2, 3) that followed publication of the Jacobson et al. (1) paper in 1989, we argued that speculation about alternative explanations for the finding was no substitute for studies in which actual levels of AFP during pregnancy are compared for cases and controls. The paper by Richardson et al. (4) in this issue of the Journal seems to be the first to address directly a possible protective effect for AFP; it is a welcome addition to the epidemiologic literature on pregnancy and breast cancer. The results of the Richardson et al. study (4) support the notion that AFP may underlie the inverse association between twin pregnancies and breast cancer. However, as do many pioneering studies, this study raises a number of new questions. The remainder of this commentary concentrates on two, namely, 1) whether AFP is the causally relevant variable and 2) what we ought to make of the finding regarding an interaction between AFP and age at first full-term pregnancy. We can be fairly confident that the pattern of results