Angiogenic Balance in Pregnancy and Subsequent Breast Cancer Risk and Survival: A Population Study

A full-term pregnancy at a young age reduces breast cancer risk later in life, and full-term pregnancies after the first confer additional protection (1, 2). However, interrupted pregnancies in the first trimester are not associated with reduced risk (3), and pregnancies with preterm delivery yield less protection than full-term pregnancies (4). There is evidence to suggest that preeclampsia, a condition defined by development of hypertension and proteinuria after 20 weeks of pregnancy, may also be associated with reduced maternal breast cancer risk later in life (5-7).

For preeclampsia to develop, an imbalance between proangiogenic and antiangiogenic factors in maternal circulation seems to be crucially important (8, 9). Thus, placental factors with antiangiogenic properties [soluble fms-like tyrosine kinase-1 (sFlt-1) and soluble endoglin (s-endoglin)] may suppress proangiogenic factors [vascular endothelial growth factor and placental growth factor (PlGF)] and lead to endothelial dysfunction and preeclampsia (9). Therefore, it has been suggested that factors that suppress angiogenesis (sFlt-1 and s-endoglin), possibly causing preeclampsia in pregnancy, may be inversely associated with subsequent risk (10, 11) and progression (12) of breast cancer later in life. On the other hand, PlGF stimulates vascular growth, possibly promoting tumor invasiveness, and PlGF may therefore be positively associated with breast cancer risk and tumor progression (10, 12). These hypotheses cannot be easily tested because PlGF, sFlt-1, and s-endoglin may not be reliably detectable in the nonpregnant state, and therefore, blood has to be collected from a large number of pregnant women, who need to be followed up for cancer occurrence after the pregnancy. Because the incidence of breast cancer is relatively low during the fertile period, even a large population of pregnant women will require long-term follow-up for sufficiently many cases to develop.

We conducted a case-control study nested within a cohort of >26,000 women who were pregnant in the early 1990s. Breast cancer cases that developed during 10 years of follow-up after pregnancy were compared with age-matched controls. Levels of PlGF, sFlt-1, and s-endoglin in blood samples collected at three time points during pregnancy were related to later risk of breast cancer. Among breast cancer cases, we also assessed whether these factors could be associated with case fatality.

Between June 1992 and May 1994, 35,940 women in Norway participated in a prospective study of Toxoplasma gondii infection in pregnancy. All pregnant women who attended antenatal care in 11 of 19 counties were included in the study (13). After the original study, we attempted to contact all participating women by mail to obtain their consent to use the data and blood samples for additional scientific purposes, and 29,072 women consented. Of these, 26,744 women had available blood specimens and were eligible for this study.

We obtained information on pregnancy outcomes from the Medical Birth Registry of Norway to which information on all births in the country has been reported since 1967. Information on each pregnancy is obtained by compulsory notification on standardized forms by delivering midwives, including information on maternal and fetal health. Mean age at childbearing was 29 y (SD, 5.3 y), 41% were nulliparous, 36% were primiparous, and 23% of the women had two or more previous births.

Using the unique identity number of every Norwegian citizen, the 26,744 women with available blood specimens were linked to the Norwegian Cancer Registry to ascertain incident cases of breast cancer that occurred during ∼10 y of follow-up after the index pregnancy. A total of 145 women had developed invasive breast cancer after pregnancy. A random sample of 400 women who did not develop breast cancer during follow-up was used as controls.

The study was approved by the Norwegian Directorate for Health, the Norwegian Data Inspectorate, and the Regional Committee for Medical Research Ethics.

The first serum sample was collected around week 10 of gestation (median, 10; range, 4-18 wk), the second sample was collected around week 23 (median, 23; range, 19-34 wk), and the third sample was collected around week 38 (median, 38; range, 35-40 wk).

Among a total of 545 women (145 breast cancer cases and 400 controls), 525 women had provided blood from the first sampling period. Among these women, sufficient serum was available to analyze sFlt-1 in 508 samples, PlGF in 488, and s-endoglin in 447 samples. From the second sampling period, sufficient serum was available for sFlt-1 measurements in 435 samples, PlGF was measured in 432, and s-endoglin in 382 samples. For the third sampling, the corresponding numbers were 340 for sFlt-1, 339 for PlGF, and 300 for s-endoglin.

The sera were treated according to uniform standards and stored in one place at −20°C until analyzed. Concentrations of PlGF, sFlt-1, and s-endoglin were determined by ELISA according to the manufacturer's instructions (R&D Systems). These assays have been validated in studies of pregnant women, and minimal detectable levels for the assays of PlGF, sFlt-1, and s-endoglin were 7, 5, and 7 pg/mL, respectively. Samples with concentrations below these limits were assigned the minimal detection value. The technicians doing the tests were blinded to the clinical outcome. Within the batches, blinded quality control samples were inserted and showed that the coefficients of variation were within the limits recommended by the manufacturer. Thus, the coefficients of variation were 6% to 7.5% for s-endoglin, 5% to 8.6% for sFlt-1, and 7% to 10.5% for PlGF.

Using the three samples of maternal serum collected during pregnancy, we estimated geometric mean concentrations of PlGF, sFlt-1, and s-endoglin among breast cancer cases and controls. For each sampling period (median weeks 10, 23, and 35), we assessed the association of each angiogenic factor with the risk of breast cancer, using odds ratios (OR), by dividing PlGF, sFlt-1, and s-endoglin into quartile categories, according to the distribution among control women.

We also assessed individual longitudinal changes in PlGF, sFlt-1, and s-endoglin from one sampling period to the next. There was a general increase for each factor over time, and because the increase was normally distributed, we studied breast cancer risk (expressed as ORs) per SD increase in each factor. We used multiple logistic regression analysis to adjust for potential confounding by maternal age and parity, plurality (twins), and length of pregnancy at serum collection.

Among women who developed breast cancer during follow-up, we also assessed whether the three angiogenic factors, as measured in pregnancy, were associated with case fatality of breast cancer. In this analysis, we estimated the risk of dying (hazard ratios) from breast cancer associated with each angiogenic factor to indicate whether angiogenic profile (dominantly proangiogenic or antiangiogenic) was associated with breast cancer survival.

Stata for Windows (version 10.0; Stata Corp.) was used for the statistical analyses.

Compared with controls, mean serum concentrations of PlGF, sFlt-1, and s-endoglin in blood samples collected in the first, second, and third sampling period did not differ from women who were subsequently diagnosed with breast cancer (Table 1). Concentrations of each angiogenic factor increased with increasing length of pregnancy, but the increase was similar among cases and controls.

For each sampling period, we divided PlGF, sFlt-1, and s-endoglin into quartile categories based on concentrations among controls and estimated ORs for breast cancer using the lowest quartile as the reference. For the first and second sampling period, there was no clear association with later breast cancer risk for any of the measured angiogenic factors after adjustment for age, parity, twin pregnancies, and length of pregnancy at blood sampling (Tables 2 and 3). Similarly, there was no evidence for any association related to the third sampling period (data not tabulated).

We also studied whether longitudinal increase in each angiogenic factor from one sampling period to the next was associated with later risk of breast cancer (Table 4). We estimated increase from the first to the second, from the first to the third, and from the second to the third period and assessed ORs for later breast cancer per SD increase in each factor. For PlGF and sFlt, the results were consistent in showing that increase in these factors throughout pregnancy was not associated with subsequent breast cancer risk. For s-endoglin, however, the results were less consistent. Whereas the increase in s-endoglin from the first to the second sampling period suggested a weak positive association [OR per SD increase, 1.3; 95% confidence interval (95% CI), 0.9-1.8], the increase in s-endoglin from the second to the third sampling period was inversely associated with risk (OR per SD increase, 0.6; 95% CI, 0.4-0.9).

Among 145 breast cancer cases in this study, 22 died from the disease during follow-up. We assessed whether angiogenic factors measured in blood samples collected from pregnancy several years before diagnosis were associated with survival but found no indication for any association (data not tabulated). Thus, the hazard ratio of death per SD increase in PlGF from the first to the second sampling period was 1.1 (95% CI, 0.6-1.9), and for the corresponding increase from the first to the third sampling period, the hazard ratio was 1.1 (95% CI, 0.6-1.8). For sFlt-1, the hazard ratio per SD increase from the first to second sampling period was 0.7 (95% CI, 0.4-1.4), and the corresponding increase from the first to the third period was associated with a hazard ratio of 1.4 (95% CI, 0.9-2.2). For s-endoglin, the hazard ratio per SD increase from the first to the second sampling was 0.5 (95% CI, 0.3-1.0), and correspondingly, the hazard ratio associated with one SD increase from the first to the third sampling period was 1.3 (95% CI, 0.7-2.3).

In this nested case-control study within a large cohort of pregnant women, there was no clear evidence that angiogenic factors measured in the maternal circulation during pregnancy are associated with later risk or case fatality of breast cancer.

Previously, no prospective study has assessed the hypothesis that an imbalance in angiogenic factors could be associated with breast cancer risk later in life. The hypothesis was derived from studies that have shown reduced breast cancer risk in women with a history of preeclampsia (5-7). Because preeclampsia is associated with a relative dominance of antiangiogenic over proangiogenic factors in pregnancy (8, 9, 14), Aagaard-Tillery and coworkers (10) suggested that this imbalance could be associated with reduced cancer risk later in life. The hypothesis is attractive in suggesting that angiogenic imbalance provides a novel biological understanding for the reduced breast cancer risk observed in women with a history of preeclampsia (5-7). However, it has also been suggested that markers of angiogenic balance may be associated with the prognosis of cancer and that PlGF may promote tumor angiogenesis (12). Thus, expression of PlGF has been shown in breast cancer tissue, and it has been suggested that PlGF may be positively associated with breast cancer recurrence, metastasis, and case fatality (12). On the other hand, the relative dominance of sFlt-1 and s-endoglin relative to PlGF observed in preeclampsia may be associated with lower breast cancer risk and, among breast cancer cases, relatively smaller tumor size, lower prevalence of metastatic disease, and, possibly, reduced case fatality (11).

This nested case-control study was derived from breast cancer follow-up of >26,000 pregnant women and with repeated collection of blood roughly corresponding to the three trimesters of pregnancy. This allowed us to assess whether the typical longitudinal increase in circulating angiogenic factors (14) was associated with breast cancer risk later in life. All women who developed breast cancer during ∼10 years of follow-up were included, and controls were a randomly selected age-matched group of women within the cohort who did not develop breast cancer during follow-up.

Maternal serum was stored at −20°C for ∼10 years before analyses. The use of blinded quality control samples from the manufacturer was reassuring in showing that the coefficients of variation of each of the angiogenic factors were within the recommended limits. Both storage time and measurement error could have affected serum concentrations of PlGF, sFlt-1, and s-endoglin, but for both sources of error, it does not seem likely that women who later developed breast cancer and control women have systematically differed. However, random differences in measurement error between the groups could have biased the results toward the null.

Effects related to changes in angiogenic factors could only be studied among women with serum collected from at least two sampling periods, and it is possible that differences in collection patterns could have biased the results. However, the pattern of serum collection between breast cancer cases and controls did not differ, and therefore, this is an unlikely source of bias.

We considered potentially confounding factors and adjusted for differences in maternal age and parity between cases and controls because these factors are related to breast cancer risk. In the analyses of single measurements of PlGF, sFlt-1, and s-endoglin within each sampling period, we also adjusted for differences in pregnancy week at blood sampling. In the analyses of increase in PlGF, sFlt-1, and s-endoglin in blood collected in the first, second, and third sampling period, adjustment was made for differences in time interval between blood collections. However, the results remained essentially unchanged after these adjustments.

Follow-up of breast cancer cases was too short to allow an appropriate analysis of survival. However, among the 22 women who died from breast cancer, there was no clear evidence that angiogenic factors measured in pregnancy up to several years before the diagnosis of breast cancer are associated with case fatality.

Nonetheless, there were tendencies for s-endoglin to show associations with both risk and case fatality, but the results were not consistent from one sampling period to the next, and increase in s-endoglin between sampling periods also showed inconsistent patterns, suggesting that the results for s-endoglin should be interpreted with caution.

Few studies have been in the position to assess pregnancy factors in maternal serum as possible predictors for later breast cancer risk (15-17). Within the context of an antenatal screening program in Denmark, Melby et al. (15) conducted a nested case-control study of maternal α-fetoprotein levels in pregnancy and later risk of breast cancer. The results suggested that relatively high levels of α-fetoprotein, indicating possible inhibition of estrogen activity, were associated with reduced breast cancer risk later in life (15). In a population study of pregnant women in Sweden, however, the inverse association of α-fetoprotein with later risk of breast cancer could not be confirmed (16). In the Swedish study, concentrations of human chorionic gonadotropin were also measured, and the results suggested a possible inverse association of human chorionic gonadotropin with breast cancer risk later in life. In addition, the investigators of the Swedish data measured insulin-like growth factors (IGF) and found a positive association of IGF-I with later breast cancer risk but no association with IGF-II (17).

There was a difference in the timing of blood sampling between these populations. In the Danish population (15), blood was sampled in the second trimester, whereas blood was drawn in the first trimester in the Swedish population (16, 17). In the present study, we had blood collected from three periods that correspond fairly well to the trimesters of pregnancy. Due to the dramatic hormonal and other changes that take place during pregnancy, it is necessary to consider how the timing of blood collection may influence the results of studies that use maternal blood samples from pregnancy to assess later risk of disease.

In summary, the results of this nested case-control study, based on blood samples collected throughout pregnancy, and subsequent cancer follow-up, do not provide any clear evidence that pregnancy levels of PlGF, sFlt-1, and s-endoglin are associated with breast cancer risk later in life.

No potential conflicts of interest were disclosed.