Breast cancer risk by age at birth, time since birth and time intervals between births: exploring interaction effects

Abstract

In a Norwegian, prospective study we investigated breast cancer risk in relation to age at, and time since, childbirth, and whether the timing of births modified the risk pattern after delivery. A total of 23,890 women of parity 5 or less were diagnosed with breast cancer during follow-up of 1.7 million women at ages 20-74 years. Results, based on Poisson regression analyses of person-years at risk, showed long-term protective effects of the first, as well as subsequent, pregnancies and that these were preceded by a short-term increase in risk. The magnitude and timing of this adverse effect differed somewhat by birth order, maternal age at delivery and birth spacing. No transient increase in risk was seen shortly after a first birth below age 25 years, but an early first birth did not prevent a transient increase in risk after subsequent births. In general, the magnitude of the adverse effect was strongest after pregnancies at age 30 years or older. A wide birth interval was also related to a more pronounced adverse effect. Increasing maternal age at the first and second childbirth was associated with an increase in risk in the long run, whereas no such long-term effect was seen with age at higher order births.

Figure 1

Predicted incidence rate ratio of breast cancer for women of parity i (P_i, i=1–5, with a common effect of parity 4–5) according to time since ith birth, relative to nulliparous women. Results are adjusted for age, birth-cohort, age at ith, that is, most recent, and all previous births (assuming a common effect of age at ith birth among women of parity ⩾i), and with interaction between parity and time since birth. The predicted risk level corresponds to reference categories of ages at births, that is, the youngest categories of age at Nth birth (N=1–5) shown in Table 1.

Figure 2

Predicted incidence rate ratio of breast cancer for uniparous women (P₁) according to time since first birth in subgroups of age at first birth, relative to nulliparous women, calculated on the basis of a model (A) without and (B) with, interaction between age at and time since the first birth. Results are adjusted for age and birth-cohort.

Figure 3

Predicted incidence rate ratio of breast cancer for biparous women (P₂) according to time since second birth in (A) subgroups of age at the second birth (interaction model), and for each category of age at second birth, that is, (B) <25, (C) 25–29, and (D) ⩾30 years, within subgroups defined by time interval between first and second birth, relative to nulliparous women. Results are adjusted for age and birth-cohort, and in (A) also for age at the first birth.

Figure 4

Predicted incidence rate ratio of breast cancer for triparous women (P₃) according to time since third birth in (A) subgroups of age at the third birth (interaction model), and for each category of age at third birth, that is, (B) <30, (C) 30–34, and (D) ⩾35 years, within subgroups defined by time interval between second and third birth, relative to nulliparous women. Results are adjusted for age, birth-cohort, age at first birth and in (A) also for age at the second birth.

Figure 5

Predicted incidence rate of breast cancer for women of parity 0–3 (P₀, P₁, P₂ and P₃, respectively) by attained age, with additional contribution from each subsequent child according to time since the most recent birth, in specific combinations of ages at first, second and third birth (a₁, a₂, and a₃), respectively. (A)–(C) represent combinations with an early first birth, (D), (E) a first birth at an intermediate age, and (F) a late first birth. Results are adjusted for age and birth-cohort.