During the past fifteen years,two new perspectives have been introduced in the study of reproductive history in relation to breast cancer risk. Lambe and colleagues 1 have provided evidence that a livebirth has dual effects on maternal breast cancer risk—a transient increase followed by a long-term more marked decrease. This observation provided an explanation for what has been thought of as a paradox—that, among young women, parity increases rather than decreases breast cancer risk. 2 In addition, several authors have pointed out that an induced abortion may increase the risk of breast cancer but others have suggested this apparent increase simply reflects a lack of long-term protection that would have been con veyed by a full-term pregnancy. 3,4 Neither of these p e rs p e c t i ves has received biological support from an independent point of view. Because premature deliveries are not accompanied by a late gestational-age hormonal surge, they are likely to modify the dual effect.We have examined in a large population-based database whether premat u r e d e l i ve ries impart less short - t e rm breast cancer ri s k increments and less long-term breast cancer risk reduction in comparison to full-term deliveries. Members of the study cohort were identified from the mothers who delivered liveborn or stillborn babies after a gestation period of at least 28 weeks in Sweden from 1973 to 1989—a total of 1 093 466 women who had a total of 1 953 975 deliveries. However 285 592 women had at least one delivery before 1973 and detailed information on such deliveries was not available.Thus,the study base consisted of 807 874 women whose deliveries occurred after 1973. We used a nested case-control design to allow more efficient analyses. Case patients were women diagnosed with breast cancer during the same period, as ascertained from the records of the National Cancer Registry. For each case patient, five control patients were randomly selected from the study base. These women were individually matched by birth year to the index case, were alive at least to the date of the diagnosis for the index case, and had not been previously diagnosed with breast cancer. Included in the analysis were 2318 breast cancer cases who had complete information on gestation length and had only singleton births. Similarly, 10 199 control women age-matched to the case were available for a n a l y s i s. A premature delive ry is defined as having a gestation age of less than 37 weeks. Age at first birth was adjusted for in all analyses. Among all parous women,having a premature, rather than a term, delivery of their first pregnancy was associated with an overall breast cancer odds ratio of 1·17 (95% CI 0·98–1·40). Premature delivery was not associated with breast cancer risk among women younger than 40 years of age (1·03, 0·79–1·35), whereas it was associated with a significantly increased risk among those aged 40 or older (1·30, 1·02–1·65). Both results are compatible with the previously outlined expectations. Among uniparous women, a premature delivery was associated with a slightly lowered breast cancer risk within 5 years after the delivery compared with a term delivery, whereas it was associated with increased risk later on (table). A similar pattern was evident among multiparous women. The findings of our study indicate that the transient riskincreasing effect of a full-term pregnancy on breast cancer ri s k , which is mostly evident among women in their childbearing years, is less evident after a premature delivery. We have also found that a premature delivery conveys less long-term protection against breast cancer than a term delivery, possibly because it has a more modest impact on the process of terminal differentiation. 5 Our findings suggest that the effect of pregnancy outcomes,including interruption through induced abortion, on maternal breast cancer risk should be examined with variable time lags. They also allow for a coherent view of the complex relation betwe e n reproductive characteristics and breast cancer risk.