Placental Weight and Breast Cancer Survival in Young Women

Hormonal factors in general and reproductive history in particular are crucial risk factors for developing breast cancer (1). Pregnancies result in a long-term reduced risk of breast cancer. However, this risk reduction is preceded by a pregnancy-related increased risk, which seem to be more pronounced in primiparous than multiparous women (2). The possible role of reproductive factors on breast cancer prognosis is less studied.

Emerging epidemiologic evidence indicate that reproductive history has an effect on survival in premenopausal breast cancer (3). Among women with premenopausal breast cancer, >40% die from their disease within 15 years from diagnosis (4). The prognosis is poorer among women diagnosed within 10 years after last delivery (5), and even poorer among women with pregnancy-associated breast cancer, defined as breast cancer diagnosed during pregnancy or within 2 years after delivery (5-12). The underlying mechanisms for an adverse effect of recent childbirths on survival remain unclear, but hormonal influences may play a role.

The placenta is the major source of several pregnancy hormones (13), and the size and weight of the placenta could serve as surrogate markers of exposure to pregnancy hormones (14, 15). In two studies, placental weight has been linked to an increased risk of maternal breast cancer (16, 17).

We hypothesized that increased placental weight in the most recent pregnancy before diagnosis would be associated with reduced survival in breast cancer, and that this effect would be most pronounced in pregnancy-associated breast cancer. In addition, given that the short-term risk of breast cancer after childbirth is especially increased among primiparous women, we further hypothesized that the effect of placental weight on breast cancer survival would be more pronounced in primiparous than multiparous women. In consequence, we investigated whether any association between placental weight and breast cancer survival was modified by parity, or by time between the last pregnancy and diagnosis, with special emphasis on pregnancy-associated breast cancer. To study these hypotheses, we used a cohort of young, parous women with breast cancer.

Data from medical records and from four population-based Swedish registers were used for the study. All Swedish registers and medical records can be linked through an individually unique National Registration Number, assigned to all Swedish residents. The Swedish Cancer Register, the Swedish Medical Birth Register, and the Swedish Cause of Death Register are all held by the National Board of Health and Welfare.

National cancer registration in Sweden started in 1958 and reporting of all incident cancers is mandated by law for both clinicians and pathologists. Virtually all (99%) of the cancer cases are morphologically verified, and comparisons with data reported on death certificates show that >99% of all breast cancer cases are registered (18).

The Medical Birth Register is based on data from standardized medical records used in all Swedish maternal care and delivery units because the beginning of 1973, when registration started. It includes data on >99% of all births in Sweden (19). Individual variables have diverse completeness, as routines to collect information on some variables have shifted over the years. For example, sex, length, and weight of child have a near 100% completeness, whereas information on maternal smoking or placental weight is less complete.

Causes of death have been recorded in Sweden since the mid-18th century. The registration includes dates of all deaths in Sweden as the Causes of Death Register is linked to the Register of the Total Population. However, in some cases, the cause of death cannot be established due to missing death certificates. The proportion of missing data have in recent years been <1% but was larger in earlier years (20). Furthermore, a number of cases have been wrongly diagnosed, or the clinician has been uncertain of the cause of death. In a validation of breast cancer recorded as underlying cause of death in the Swedish Causes of Death Register, <2% of deaths were judged to be in disagreement with the conclusion of an end point committee that reviewed each case (21).

The Swedish Migration Register, held by Statistics Sweden, is a part of the Swedish Register of the Total Population, and contains dates of immigration and emigration for all Swedish residents.

The study base included 1,874 Swedish women previously free of breast cancer with a recorded childbirth after January 1st, 1973, and a subsequent diagnosis of primary breast cancer before December 31st, 1991. The cohort has been described in detail elsewhere (22). All women were followed up with regard to death or permanent emigration between date of diagnosis and December 31st, 2006. One case was excluded from analysis because the date of death could not be ascertained. Thus, data from 1873 women were available for analyses.

Dates of breast cancer diagnoses were collected from the Swedish Cancer Register. Dates of immigration and emigration during follow-up were obtained from the Swedish Migration Register, whereas the Swedish Causes of Death Register provided information on dates and causes of death.

For all cases, information concerning the first birth after January 1st, 1973, the preceding pregnancy, and reproductive history before that, had previously been collected from antenatal care and birth records. These data included maternal height, prepregnancy weight and smoking (both recorded at the woman's first visit to prenatal care), pregnancy weight gain, time and number of previous pregnancies and abortions, gestational length, weight, length and sex of the child, and placental weight. The dates and number of subsequent childbirths were retrieved from the Swedish Medical Birth Register. Data from this register also included age and citizenship of the mother; parity; gestational length; and offspring weight, length, and gender. For childbirths after January 1st, 1982, data were also available on maternal height, smoking, weight at delivery, and pregnancy weight gain. Between January 1st, 1982 and December 31st, 1989, the Medical Birth Register included information on placental weight.

After combining the data from medical records and the birth register, information about placental weight from the last pregnancy before diagnosis was available for 1,057 women. The proportion of missing data were smaller for potential confounders (Table 1).

All analyzed data were delivered deidentified from the registers. Both the study from which the cohort originated and the present study were approved by the Research Ethics Committee at Karolinska Institutet, Stockholm, Sweden.

The effect of placental weight and potential confounders on mortality was assessed by Cox survival models. Results are given as hazard ratios with 95% confidence intervals (95% CI). Placental weight was analyzed both as a linear continuous variable and as a categorized variable. Dividing placental weight into quartiles (<510, 510-590, 590-675, and ≥675 grams) was considered suboptimal, as placental weights often had been recorded to the nearest 50 grams. In an alternative approach, categories identical with those in our previously published study on placental weight and risk of breast cancer (16) was used (<500, 500-599, 600-699, and ≥700 grams). Age at diagnosis, height, parity, age at first birth, prepregnancy weight, smoking, and sex and weight of last child before diagnosis were all considered as potential confounders. The adjusted model further included gestational length, as placental weight corrected for gestational length was believed to better represent the exposure to sex hormones during pregnancy. Nonlinear effects among potential confounders were investigated by adding squared terms to the model. The squared terms for age at diagnosis and weight of last child had stronger association to outcome than linear terms, and were added to the model. The proportional hazard assumption was assessed by plots of Schoenfeld residuals against ranks of follow-up time, and by adding time-dependent variables to the model (23). Age at diagnosis and parity both had time-dependent effect on mortality and were added to the model as functions of log(time).

Effect modifications by parity and by time between most recent birth and diagnosis were evaluated by modeling interactions between placental weight and parity (primiparity versus multiparity) and between placental weight and interval of time between most recent birth and diagnosis, respectively.

In the categorization of time between most recent birth and diagnosis, special emphasis was made on cancers arising in close proximity to pregnancy. Pregnancy-associated breast cancer has several definitions in the literature, mainly depending on the scope of the study. In this context, the period of interest was during pregnancy or within 2 y thereafter, as this is the period, which has most frequently been associated with reduced breast cancer survival (5-12). Further categories were made as 2 to 5, 5 to 10, and ≥10 y.

Among 824 women who died during follow-up, 90.4% had breast cancer registered as cause of death. Of the remaining 79 women, 23 had no cause of death listed, whereas 56 were registered with a cause of death other than breast cancer. Because undetermined or uncertain causes of death were so frequent among those not classified as breast cancer deaths, the proportion of misclassification was assumed to be substantial in this group. To avoid this potential misclassification error, all analyses were carried out with all-cause mortality as the outcome of interest.

All statistical analyses were done using SAS statistical software.

Baseline characteristics of the 1,873 cases are listed in Table 1. Significant associations with placental weight were observed for parity, prepregnancy weight, gestational length, and birth weight. Information on placental weight in the most recent pregnancy was more frequently missing in multiparous compared with primiparous women, but otherwise, there were no large differences in the distribution of baseline characteristics between women with or without data on placental weight.

Data on placental weight was available for 1,057 women. During 14,772 person-years of follow-up (mean, 14.0 years; range, 14 days-33 years), 456 (43%) of these women died. Overall, each increase of 100 grams in placental weight resulted in a nonsignificant crude hazard ratio of 1.03 (95% CI, 0.96-1.11; Table 2). After adjustment for age at diagnosis, parity, age at first birth, prepregnancy weight, gestational length, and birth weight, the hazard ratio increased to 1.09 (95% CI, 0.99-1.19). Further adjustments did not alter the results (data available at request). In the adjusted analysis, women with placental weight of ≥700 grams were at a statistically significant 50% higher risk of death than women with placental weight of <500 grams, whereas women with placental weight of 500 to 599 and 600 to 699 grams had smaller, nonsignificant increases in risk.

Among primiparous women, each increase in placental weight by 100 grams resulted in a statistically significant increased hazard ratio of death of 1.18 (95% CI, 1.03-1.36; Table 3). After adjustment, the hazard ratio increased to 1.26 (95% CI, 1.09-1.47). Among primiparous women, having had a placental weight of ≥700 grams resulted in a more than doubled risk of death after diagnosis of breast cancer, compared with a placental weight of <500 grams. Among multiparous women, placental weight was not significantly associated with mortality. The interaction between placental weight and primiparity was statistically significant (P = 0.009 when placental weight was used as a continuous variable).

A further analysis investigated whether the effect of placental weight on breast cancer mortality was modified by time since most recent childbirth (Table 4). Cases with pregnancy-associated breast cancer had a statistically significant increased risk of death with increasing placental weight (adjusted hazard ratio, 1.30 for each extra 100 grams; 95% CI, 1.06-1.59; P = 0.01). When comparing categories of placental weight with <500 grams as the reference, the hazard ratio increased stepwise to 4.09 (95% CI, 1.63-10.23) for placental weight ≥700 grams. Among cases with >2 years between most recent birth and diagnosis, no significant associations between placental weight and mortality were detected. Still, the interaction between placental weight and time since most recent childbirth did not reach statistical significance (P = 0.23 based on placental weight as a continuous variable). When time since most recent childbirth was dichotomized (during pregnancy or ≤2 years compared with >2 years from last childbirth), P for interaction equaled 0.06.

We found a statistically significant association between placental weight in the most recent pregnancy and risk of death after a diagnosis of breast cancer among primiparous women.

Regardless of parity, we found a statistically significant increased risk of death with increasing placental weight among women diagnosed with breast cancer during pregnancy or within 2 years thereafter. For women diagnosed later, we could not detect any significant effect, although the lowest hazard ratios predominantly were found among women with the smallest placentas.

Placental weight could only be ascertained for 56% of the women in our initial cohort. There were two major reasons for the missing data; information was lacking because the placenta was never weighed, or we failed to retrieve information on placental weight from most recent birth (we had data on placental weight from medical records regarding the first birth between 1973 and 1991 and from the Birth Register regarding all births between 1982 and 1989). The latter explanation also accounts for the fact that multiparous women were more likely than primiparous women to have missing information on placental weight in the most recent pregnancy. Because parity was not associated with risk of death, and the patients with missing data on placental weight did not differ from the rest of the group in any other baseline characteristic, we believe the missing data to be nondifferential. Furthermore, because all data had been recorded before diagnosis, reporting bias seems unlikely.

A limitation of the present study was the lack of information on tumor characteristics, which could have provided valuable insights to possible underlying mechanisms. Previous studies have shown that tumors arising shortly after last pregnancy are more likely to be node-positive (7, 9, 11) or nonlocalized (24), to have a high histologic grade (7, 9, 24), to be p53 positive (7), to have a high S-phase fraction (7), and to have high mitotic count (7), whereas results on receptor status are conflicting. However, proximity to childbirth seems to be a predictor of poor breast cancer survival even after tumor characteristics have been taken into account (7, 9, 11, 25).

Results from earlier studies on reproductive factors and breast cancer survival are to some extent conflicting. In some studies, low age at first childbirth has been associated with poorer prognosis (26, 27), whereas other investigators have found no such association (12, 25). Likewise, some (10, 27), but not all (5, 12, 25, 26) studies have found evidence of a worse prognosis with increasing parity. Most widely studied is the prognostic influence of time since last childbirth. Several studies have found a reduced survival among women who are diagnosed within 2 years from last childbirth (6-12), or during pregnancy (5, 6, 8), and to some extent, as long as 5 years (27), 8 years (6), or 10 years (5) after last childbirth, compared with those with less recent births.

Given previously reported findings of a poor breast cancer survival in the first years after childbirth, our data support the theory that cancers arising during or shortly after pregnancies are influenced by pregnancy hormones. The association between placental weight and survival in our data were also stronger among primiparous women. As the effect of pregnancies on breast cancer risk is stronger for the first pregnancy than for any subsequent (2), this effect modification further supports the idea of a hormonal influence on breast cancer survival. However, from the presented data, we cannot tell whether the more pronounced association between placental weight and breast cancer mortality in primiparous women is due to a general effect of the first child-birth or is due to that women with only one child-birth are more vulnerable to hormone or other exposures during pregnancy.

In our data, women with pregnancy-associated breast cancer were also more likely to be primiparous (35% compared with 30% in the rest of the cohort; P = 0.20). The small numbers did not allow us to further investigate risk of death by year of follow-up among primiparous women. Hence, we cannot determine whether the effect modifications by parity and by proximity to childbirth were independent.

Placental weight was used as an indirect marker of pregnancy hormone exposure. Both exposure to endogenous and exogenous estrogens and insulin-like growth factors have been implicated in breast carcinogenesis (1, 28). Placental weight has been associated with estriol (14, 15, 29, 30), and progesterone (15, 29) levels in late pregnancy, sex hormone–binding globulin (15), human placental lactogen (30, 31), human chorionic gonadotropin (30), insulin (32), insulin-like growth factor 1 (32, 33), and with less certainty also with estradiol levels (15, 29). We can, however, only speculate which hormone or hormones that may be important for the association between placental weight and survival in breast cancer.

Although our results may not have direct clinical implications, they indirectly support the notion of an important role of hormonal factors in the biological behavior of breast cancer.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank the Unit of Population and Welfare at Statistics Sweden and the Epidemiological Centre at the National Board of Health and Welfare for sharing data from the nationwide registers.