SEARCH

SEARCH BY CITATION

Keywords:

  • breast cancer;
  • prognosis;
  • reproductive factors;
  • premenopausal;
  • postmenopausal

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Statistical Analyses
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Reproductive factors that have a well-documented effect on breast cancer risk may also influence the prognosis of the disease, but previous studies on breast cancer survival have yielded conflicting results. We combined information from two population-based registries and obtained information on 16,970 parous women with invasive breast cancer. Cox regression analysis was used to assess breast cancer survival in relation to age at diagnosis, age at first birth, time since last birth and parity. We stratified the analyses by age at diagnosis (<50 and ≥50 years) as an approximation for menopausal age. In women diagnosed before 50 years of age, breast cancer survival was reduced with younger age at diagnosis (p for trend <0.001), whereas in women diagnosed at 50 years or later, survival was reduced with older age at diagnosis (p for trend 0.011). For breast cancer diagnosed before 50 years, survival was poorer in women with four or more births compared to women with one or two births (hazard ratio 1.3, 95% confidence interval 1.1–1.6). A short time since last birth was associated with reduced survival (p for trend 0.05), but adjustment for stage and grade attenuated the association. Among women diagnosed at 50 years or later, we found no association with survival for any of the reproductive factors. In summary, reproductive factors were associated with survival from breast cancer diagnosed before but not after age 50 years. Young women had a particularly poor prognosis throughout the study period.

Breast cancer is the most common malignancy and the leading cause of cancer-related death in women worldwide.1 Late age at first birth and low parity are well-established risk factors for breast cancer, and the increasing incidence in western countries may partly be explained by changes in reproductive patterns.2, 3 Pregnancy is known to have a dual effect on breast cancer risk, with an increased risk for about 5–10 years after a pregnancy, followed by a lifelong protective effect.4, 5 The occurrence of breast cancer shortly after a pregnancy is rare, but many women of today postpone their pregnancies until their thirties, and with the steep age related increase in risk, the number of cases that will appear shortly after a pregnancy is likely to increase.6

The influence on breast cancer risk by pregnancy related factors is thought to be mediated by hormonal mechanisms.5 It seems plausible that these factors could also influence the prognosis of the disease through hormonally induced growth stimulation or by altering the susceptibility for different subtypes of breast cancer with inherently different prognoses.7, 8 However, the influence of age at first birth, parity and time since last birth on breast cancer survival is not established in the same way as their influence on breast cancer risk.

Several studies that investigated the association of time since last birth with the prognosis of subsequently occurring breast cancer reported a poorer prognosis related to a short time interval from last birth until diagnosis.6, 9–13 There is also some epidemiological evidence that breast cancer patients with high parity may have a relatively poor prognosis,14–17 but others have failed to confirm that finding.18 With regard to age at first birth, results have conflicted; some have reported no association,17 and others have either reported better10, 15, 18 or poorer19 survival with increasing age at first pregnancy. It is possible that some of the discrepancies in findings between studies may be attributed to differences in patients' age (i.e., premenopausal or postmenopausal breast cancer).

In this population-based study from Norway, we combined information from the national Medical Birth Registry (MBR) and the Cancer Registry to assess the relation of age at first birth, time since last birth and parity with breast cancer survival among 16,970 parous women with invasive breast cancer.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Statistical Analyses
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The study population consists of women who have given their first birth since 1967 and been registered in the MBR of Norway (www.mbr.no) and subsequently been diagnosed with breast cancer, as recorded in the Cancer Registry of Norway. Women who had given birth before 1967 (this information is recorded in the registry) were not included in this study due to the lack of information on dates for their previous births. Thus, among 1,026,587 women with a first birth after 1967, 16,970 were diagnosed with invasive breast cancer during the interval from their first birth until December 31, 2008. Among these cases, 9,160 women were diagnosed before 50 years of age, and 7,810 women were diagnosed at 50 years or later, which reflects that the source population is still relatively young.

The MBR comprises compulsory reported information by midwives and doctors including maternal age, marital status, date of delivery, pregnancy outcome and complications as well as characteristics of the newborn. Using the unique 11 digit identification number of Norwegian citizens, individual information from the MBR was linked to information about cancer incidence provided by the Cancer Registry (www.krefregisteret.no). The Cancer Registry of Norway has registered information on incident cancer cases since 1953, and the reporting of cancer is mandatory by law. The information includes date at diagnosis, clinical stage at diagnosis and histological grade. The registry information is considered to be practically complete.20

Breast cancer is registered according to the International Classification of Diseases (ICD-7 code 170 and ICD-10 code C50). Information on cause-specific death among breast cancer cases was provided by the Norwegian Cause of Death Registry (www.ssb.no/dodsarsak/) [ICD-6/7 (1951–1968) code 170, ICD-8/9 (1969–1995) code 174 and ICD-10 (1996–2008) code C5021] and information on date of death or emigration was provided by the National Registry.

The study was approved by the Regional Committee for Medical Research Ethics that regulates both legal and ethical aspects of medical research in Norway.

Statistical Analyses

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Statistical Analyses
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Each of the 16,970 women with breast cancer contributed person-time from the date of diagnosis until the date of death from breast cancer, death from any other cause, emigration or the end of mortality follow-up (December 31, 2009), whichever occurred first.

We analyzed breast cancer survival separately for women diagnosed before or after 50 years of age, where 50 years was regarded as a crude approximation for menopausal age.22

Cox regression analysis was used to estimate hazard ratios (HRs) with 95% confidence interval (CIs) of death from breast cancer associated with the following study factors: calendar period of diagnosis (≤1990, 1991–1999 and ≥2000); age at diagnosis before 50 (<35, 35–39, 40–44 and 45–49 years) and at 50 years or later (50–54, 55–59, 60–64 and ≥65 years); time interval from last birth until diagnosis among women younger than 50 years at diagnosis (<2, 2–5, 6–9,10–19 and ≥20 years since last birth) and 50 years or later at diagnosis (<20, 20–29 and ≥30 years since last birth); age at first birth (<20, 20–29 and ≥30 years); and the number of pregnancies before diagnosis (1, 2, 3 and ≥4).

Age at diagnosis and time since last birth are correlated factors. In an effort to disentangle the association of age at diagnosis from that of time since last birth and to facilitate comparison with previous studies,9–11 we estimated breast cancer survival among women diagnosed before 50 years in finer strata of age (<40 and ≥40 years). In the primary analyses, we adjusted for age and calendar period at diagnosis (5-year categories), and in additional analyses, we also mutually adjusted for the other reproductive factors available (parity, time since last birth and age at first birth) and marital status at the time of the last pregnancy (cohabiting/married or single mother), as these factors were considered to be potential confounders.

In the analysis of each study factor, we used the most common category as the reference. The HR for death from breast cancer per unit increase in each study factor was also calculated with corresponding 95% CIs. p-Values for trend were calculated for the original continuous variables.

To investigate if established prognostic factors for breast cancer survival23 would influence the associations under study, we used the available clinical information in the Cancer Registry and conducted separate analyses with additional adjustment for the extent of disease (Stage I, II, III, IV and unknown) and histological grade in four categories of differentiation (well, moderate, poor and undifferentiated) at diagnosis. Histological grade was extracted from the morphology codes in the Cancer Registry, which has changed during the study period, and was available for 15% of cases diagnosed before 1993 (histological code; MoTNaC, four-digit system where the fourth digit specifies the grade of the tumor) and in 85% of cases diagnosed after 1993 [WHO ICD—oncology (ICD-O-2/3) a six digit morphology system where the sixth digit specifies the grade].

Thus, information on grade was available for 6,020 (65.7%) women diagnosed before 50 years of age, and for 6,687 (85.6%) women diagnosed at 50 years or later. To evaluate whether differences in grading information influenced our results, we stratified the analyses according to calendar period of diagnosis (before and after 1993).

Finally, we conducted additional analyses with shorter follow-up (3 and 5 years) and analyses conditioned on surviving 3 or 5 years after diagnosis to assess a possible time-dependent variation in the estimated associations.

Proportional hazards assumptions were evaluated by log minus log plots and Schoenfeld residuals and showed no violation of the assumptions. All statistical tests were two sided and all analyses were performed using STATA for Windows (Version 11© StataCorp LP, Texas, United States, 1985–2007).

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Statistical Analyses
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Table 1 shows baseline characteristics of the breast cancer patients of the study. Among women diagnosed before the age of 50 years, mean age at diagnosis was 42.7 years (SD 5.3 years), and among women diagnosed at 50 years or later, mean age at diagnosis was 56.3 years (SD 4.8 years). Mean follow-up from date of diagnosis was 9.35 years (SD 7.0 years) in the younger and 5.75 years (SD 3.8 years) in the older age group. There was a total of 3,738 deaths during follow up, and 3,161 (84%) of the deaths were due to breast cancer. Tables 2 and 3 show the associations (HRs) of each study factor with the risk of dying from breast cancer in women before and after the age of 50 years, respectively. In general, there were no substantial changes after adjusting for other reproductive factors.

Table 1. Baseline characteristics of the 16,970 women with breast cancer in the study
inline image
Table 2. HR of breast cancer death by time period at diagnosis, age at diagnosis and reproductive factors among 9,170 women less than 50 years of age at diagnosis
inline image
Table 3. HR of breast cancer death by time period at diagnosis, age at diagnosis and reproductive factors among 7,810 women aged 50 years or more at diagnosis
inline image

Women who were diagnosed relatively early during the observation period had a substantially poorer survival than women who were diagnosed during the last 2 decades, regardless of age at diagnosis (p for trend <0.001).

Table 2 shows that among women diagnosed before 50 years of age, there was a reduced survival with younger age at diagnosis (p for trend <0.001).

Thus, women diagnosed before the age of 35 years were twice as likely to die from breast cancer compared to women who were diagnosed between 45 and 49 years of age (HR 1.96, 95% CI 1.61–2.39).

A relatively short time interval since last birth also appeared to be associated with poorer survival (p for trend <0.001) in the age and period adjusted analysis, but further adjustment for age at first birth, parity and marital status attenuated the association. In a separate analysis restricted to women with information on stage and histological grade, the association of time since last birth was completely attenuated after adjustment for stage and grade. Further stratification on age at diagnosis (<40 and ≥40 years) did not change this result.

A late age at first birth (≥30 years of age) was associated with poorer survival (HR 1.21, 95% CI 1.06–1.38); however, when stage and histological grade were included in the analysis, the association was attenuated.

The association of parity with breast cancer survival showed that among the 514 women with four or more births, survival was poorer (HR 1.32, 95% CI 1.10–1.59) than among women with one or two births. This finding was not substantially altered after adjustment for potentially confounding factors, including stage and histological grade at diagnosis.

In corresponding analyses for women diagnosed at 50 years of age or later (Table 3), survival was reduced with increasing age at diagnosis (p for trend 0.011), but there were no clear associations of any of the reproductive factors with survival from breast cancer.

Stratified analyses by diagnostic period (before or after 1993) provided similar results in both strata as the combined analysis of both age groups (data not shown).

In general, there was little evidence for time dependency for the associations of reproductive factors with breast cancer survival among women younger than 50 years, although survival related to a young age at diagnosis was somewhat poorer during the first 5 years of follow up (Table 4, analyses stratified on 3 years of follow-up not shown). Also, the poorer survival associated with high parity was only apparent during the first 5 years of follow-up and not for the subsequent period. For women 50 years or older, no corresponding time dependency was detected (results not shown).

Table 4. HR of breast cancer death by age at diagnosis and reproductive factors among Norwegian women aged less than 50 years at diagnosis, according to different follow-up periods: follow up the first 5 years and follow-up from 5 years and onward
inline image

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Statistical Analyses
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

In this large registry-based study of parous women with breast cancer, a young age at diagnosis was strongly associated with poorer survival. Among women diagnosed before the age of 50 years, survival was also poorer for women with high parity. Among women who were 50 years or older at diagnosis, we found no clear association between any of the reproductive factors and breast cancer survival.

The poorer prognosis related to an early age at diagnosis has also been documented by others.24–26 However, the biological processes underlying this finding are not well understood and probably include a combination of inheritance,27 gene expression of the tumors in young women28 and the hormonal milieu for breast tumors developing in young compared to older women.

The time interval from last birth until diagnosis has been related to the prognosis of breast cancer in other large registry-based studies.6, 9, 10 In one of these, it was possible to adjust for clinical characteristics,9 whereas in others, such adjustment was not made.6, 10 Women diagnosed with breast cancer close to a pregnancy tend to be younger and have a more advanced stage of disease,29 a higher histological grade30 and a higher proportion of estrogen receptor (ER) negative tumors.11, 31 Those findings are consistent with our observation that age in itself, and tumor grade and stage, appear to explain much of the variation in prognosis related to time since last birth.

Whether the addition of more detailed tumor characteristics or clinical information would have influenced the estimated associations further is not clear and warrants further research. In the meantime, a low age at diagnosis combined with a relatively short time interval since last birth should not be neglected as potentially important for the prognosis of breast cancer patients.

The results of this study suggest that a higher age at first birth may be associated with reduced breast cancer survival among younger breast cancer patients; however, this finding was attenuated when the analyses included stage and grade at diagnosis. In other studies, age at first birth has not been consistently associated with the prognosis of breast cancer. Some investigators have reported no association,6, 17 whereas others have found poorer survival associated with an early age at first birth,10, 18 at least among younger women. In a similar but smaller study, we previously reported that late age at first birth may be associated with a poorer prognosis among women with postmenopausal breast cancer.19 However, the postmenopausal women in that study were on average much older at diagnosis, very few had received modern treatment and the results may not be fully comparable.

It appears that a young age at first birth is primarily associated with a reduced risk of low grade, ER positive breast cancer,30, 32 which in general is considered to have a better survival than high grade, ER negative tumors.33 Thus, women with an early first birth who develop breast cancer may be more likely to have ER negative, high-grade tumors, and this could explain the poorer prognosis associated with an early first birth reported in some studies. On the other hand, a late age at first pregnancy has been associated with higher risk of more aggressive ER-positive breast tumors in some studies.34, 35 Therefore, both early and late age at first pregnancy may be associated with poorer survival, but the underlying biological mechanisms may differ.

We found a relatively poorer prognosis of breast cancer in women with relatively high parity (four or more births vs. one or two births). Similar findings have been reported by some,6, 14–17 but not by others.18 In our study, adjustment for time since last birth and available clinical factors (stage and grade) did not substantially influence the association. In relation to breast cancer risk, high parity has a clear protective effect,2 but the protection may be limited to ER- positive tumors.32 Indeed, it has been suggested that high parity may increase the risk for triple negative breast cancer,36 which is a subtype known to have a poor prognosis,7, 37 and that subtype is also relatively more common in younger women.38, 39 It has also been suggested that high parity may have a protective effect on small and low-grade tumors,30, 34 with the possible consequence that women with high parity will tend to have a relatively advanced and aggressive disease. We found that the poorer prognosis of women with high parity was only apparent during the first 5 years of follow-up. This finding supports that high parity could be associated with prognostic factors (high grade, ER negativity, lymph node positivity and tumor size) that primarily have a detrimental effect on breast cancer survival the first 2–5 years after diagnosis.33

The strengths of this study include the large underlying population base of all parous women in Norway with a first birth since 1967, the high number of women with invasive breast cancer and the reliable and virtually complete mortality follow-up.

However, the source population is still young and we cannot exclude the possibility that the associations of reproductive factors with survival from postmenopausal breast cancer may change with increasing follow-up. Thus, our results concerning the role of reproductive factors on breast cancer survival in postmenopausal women should be interpreted with caution and not be generalized to older postmenopausal women.

A weakness of our study is the lack of information on socioeconomic status (SES), which may have confounded the results, as SES is associated with both reproductive behavior,40 breast cancer risk41, 42 and with survival from breast cancer.43, 44 Some of the survival difference might be explained by more advanced disease at diagnosis in women with low SES.43, 44 In our study, the adjustment for tumor stage could, at least partly, take this potential confounding into account.

Differences in breast cancer survival by SES level may also arise from suboptimal treatment.43 However, cancer treatment in Norway is publically financed and most prognostic discrepancies related to socioeconomic position may be leveled out.

We had relatively crude categorical information on tumor characteristics, and no information on tumor size, nodal status or protein expression of the tumors. Therefore, we cannot exclude the possibility that including more detailed clinical information could have influenced the results.

In this study, we found that a young age at diagnosis was associated with reduced breast cancer survival, and among women diagnosed before the age of 50 years, survival was reduced in women with high parity. We found that time since last birth may be less important for the prognosis than reported by others, but our findings suggest that among premenopausal breast cancer patients, age at diagnosis, the interval since last birth and parity should be considered in the clinical prognostication.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Statistical Analyses
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

M.D.K.A is a research fellow financed by the Norwegian University of Science and Technology (NTNU). The authors thank the Cancer Registry of Norway and MBR of Norway for providing the data.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Statistical Analyses
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References