Parity and age at first childbirth in relation to the risk of different breast cancer subgroups

Abstract

The aim of the present study was to examine parity and age at first childbirth, in relation to the risk of specific breast cancer subgroups. A prospective cohort, The Malmö Diet and Cancer Study, including 17,035 women were followed with linkage to Swedish Cancer Registry until December 31, 2004. A total of 622 incident breast cancers were diagnosed during follow-up and were evaluated regarding invasiveness, tumour size, axillary lymph node status, Nottingham grade, tumour proliferation (Ki67), HER2, cyclin D1 and p27. The tumours were also examined for WHO type and hormone receptor status. Nulliparity was associated with an overall increased risk of breast cancer, although not statistically significant (the relative risk was 1.39 with a 95% confidence interval of 0.92–2.08). Nulliparity was also associated with large tumours (>20 mm) (1.89: 0.91–3.91), high Ki67 levels (1.95: 0.93–4.10), high cyclin D1 levels (2.15: 0.88–5.27), grade III (2.93: 1.29–6.64) and HER2 positive tumours (3.24: 1.02–10.25). High parity was not statistically significantly associated with any specific breast cancer subgroup. Older age at first childbirth (>30) was associated with a slightly increased risk of breast cancer (1.39: 0.94–2.07). There was a statistically significant association between late first childbirth and lobular type (2.51: 1.01–6.28), grade III tumours (2.67: 1.19–6.02), high levels of cyclin D1 (2.69: 1.18–6.12) and low levels of p27 (2.23: 1.15–4.35). We conclude that nulliparity and late first childbirth are associated with relatively more aggressive breast cancer subgroups. © 2009 UICC

High parity has generally been associated with low breast cancer risk in previous epidemiological studies.1–5 Additionally, parity seems to be an independent prognostic factor in breast cancer and both high parity and nulliparity have been associated with poor survival when compared to uniparous women.6 Early first childbirth has been associated with a low risk of breast cancer,1 whereas the prognostic value of age at first childbirth is less clear.7

When these parity parameters are investigated there are, hence, different patterns in relation to risk versus survival. These patterns might be clarified when different tumour characteristics are taken into consideration.

The Malmö Diet and Cancer Study (MDCS), a prospective population-based cohort in Malmö, Sweden, has information on parity and other reproductive factors. A total of 17,035 women participated and during follow-up 622 women were diagnosed with breast cancer. About 90% of these cases had tumour tissue samples available for tumour reclassification and additional measurements. These tumours were evaluated with regard to invasiveness, size, axillary lymph node status, Nottingham grade, tumour proliferation (Ki67), HER2, expression of an oncogene (cyclin D1), and a tumour suppressor gene (p27). Tumours were further examined for WHO histological type (ductal, lobular and tubular) and hormone receptor status (estrogen receptor [ER]α, [ER]β and progesteronereceptor [PgR]).

The aim of the present study was to examine parity and age at first childbirth in relation to the risk of developing breast tumours with different biological characteristics in terms of histopathological findings, hormone receptor status and gene expression.

Material and methods

The Malmö diet and cancer study

Between 1991 and 1996, all residents in Malmö, Sweden in certain age groups were invited to participate in a prospective study (MDCS). The female cohort consisted of 17,035 women born between 1923 and 1950. A total of 40% of women invited to the MDCS participated in the baseline examination.8

A questionnaire provided information on parity, age at first childbirth, education, occupation, marital status, age at menarche, year of menopause, exposure to oral contraceptives (ever/never), current use of hormonal replacement therapy (HRT), alcohol consumption and smoking habits.9 Menopausal status was assessed with the help of the questionnaire and medical records. A women was considered postmenopausal if: (1) she had undergone bilateral oophorectomy; or (2) if she had undergone hysterectomy but not bilateral oophorectomy, and if she was ≥55 years of age; or (3) if the above criteria were absent and she affirmed that her menstruations had ceased at least during the calendar year 2 years prior to the baseline examination; or (4) if it was unknown whether or not she had undergone a previous oophorectomy or hysterectomy and information on menstrual status was missing, and she was ≥55 years of age. There were in all 11,388 women who were postmenopausal at baseline. A woman was classified as pre/perimenopausal if she affirmed that she was still menstruating, or if her menstruations had ceased <2 years prior to baseline examinations, or if information on menstrual status was missing and the women was <55 years of age at baseline.

Information on gynaecological surgery was collected from medical records. A trained nurse at the study centre measured height and weight, and body mass index was calculated as kg/m².

The MDCS and the present analyses were approved by the Ethical Committee at Lund University (LU 51-90 and Dnr 652/2005).

Parity and age at first childbirth

Parity was assessed from the question: “How many children have you given birth to and in what years were they born?” Parity was categorised as nullipara, 1 child, 2 children and 3 or more children. Age at first childbirth was calculated from the information provided in the same question and categorised as; ≤20, >20–≤25, >25–≤30 and >30. Information on parity and age at first childbirth was missing for a small number of women, and they form separate groups in all analyses.

Follow-up

All women were followed until December 31, 2004. Tumour end-points were retrieved by record linkage with The Swedish Cancer Registry (until December 31, 2003), and due to a delay in central registration, also with linkage to its regional branch, The Southern Swedish Regional Tumour Registry for the period January 1, 2004 to December 31, 2004. Vital status was obtained from The Swedish Cause-of-Death Registry until December 31, 2004. A total of 622 incident breast cancer cases (including invasive and cancer in situ) were registered during follow-up.

Study population

Among all 17,035 women, 576 women had already been diagnosed with breast cancer at baseline and they were excluded from the analyses as prevalent breast cancer cases. Seventy-two out of the 622 incident breast cancer cases were cancer in situ (CIS) and did not provide any invasive breast cancer end-points in the analysis for all breast cancers. CIS cases did, however, contribute with person-years up until the event in all analyses. Estimated risk for CIS was investigated as a separate end-point analysis for both parity and age at first childbirth. A total of 10 women with bilateral breast cancer were not included as end-points due to difficulties in determining the relevant side to be used in the analyses of tumour size, axillary lymph nodes and histopathology. As with CIS, bilateral cases provided person-years up until the event in all analyses. The study population hence consisted of 16,459 women.

Histopathological analyses

All tumours were reevaluated concerning invasiveness and out of 622 incident breast cancers, 72 were CIS, 10 were invasive bilateral and 39 did not have sufficient tissue for further analyses. In all, 501 tumours were invasive unilateral and had sufficient tissue for further histopathological analyses.

Biomarkers that are of clinical relevance and play an important role in cell-cycle regulation were investigated with histological analyses. One senior breast pathologist reevaluated all invasive tumours (LA) and tumour type was described according to the WHO classification. The tumours were graded according to Elston and Ellis, including tubular formation, nuclear atypia and mitotic index.10 For the construction of tissue micro array (TMA), described elsewhere,11, 12 2 cores of 0.6 mm from each tumour were taken and arranged in a recipient block. Immunohistochemical (IHC) analyses were performed using specific antibodies, described previously by Borgquist et al,11 and tumours were evaluated according to the expression of ERα, ERβ, PgR, Ki67, cyclin D1 and p27. Tumours were dichotomized with the cut-off points being; 0–10% and 11–100% positive nuclei. HER2 was analysed using IHC as previously described.11 HER2 was classified according to the Swedish clinical practice, i.e. a breast tumour is considered to be HER2 positive when it scores more than 2+ on the IHC staining.13 Tumours that were ERα, HER2 and PR negative were classified as triple-negative tumours.

All arrays were evaluated independently twice by the same person and in case of discrepancy, a third evaluation was performed by the same investigator (SiB).

Information on tumour laterality, size and lymph node metastasis were retrieved from medical records and histopathological reports by one registered nurse.

Statistical methods

Different categories of parity and age at first childbirth were compared regarding the distribution of established and potential risk factors for breast cancer. These factors were also compared between breast cancer cases (CIS, invasive with tissue and invasive with no tissue or bilateral) and noncases. Each subject was followed until the event of breast cancer, death or end of follow- up, December 31, 2004. The incidence of breast cancer was calculated per 100,000 person-years in different parity classes and in different groups of age at first childbirth. Corresponding relative risks of breast cancer risk were analysed using a Cox's proportional hazards analysis yielding relative risks with 95% confidence intervals. Women with one child were the reference group in the parity analyses and women with their first childbirth before the age of 20 were the reference group in the age at first childbirth analyses. A log-minus-log curve was plotted for overall breast cancer risk in relation to parity classes and age at first birth. Both analyses met the assumption of proportional hazards. All analyses were subsequently adjusted for potential confounders.

The confounders were chosen on the base of already established and potential risk factors for breast cancer. In the analyses, there are several end-points and adjustment for the same confounders in all analyses made it possible to obtain comparable risk estimates. The confounders were tested one by one in relation to overall breast cancer risk to see which confounder resulted in the largest change of risk estimates.

Trend over parity categories was examined from nulliparous to women with 3 or more children. When analysing trend over age at first childbirth, age groups as defined above were used. Missing category was not included in trend analyses.

To examine heterogeneity, comparing risks for different tumour subgroups in relation to specific exposure categories, for example the risk associated with lobular type tumour in women with 3 or more children, was compared to the risk of ductal type of tumour in women with 3 or more children. A case–case analysis using unconditional logistic regression analysis was applied, and p values <0.05 were considered statistically significant.

Results

Distribution of risk factors in different breast cancer cases and noncases

Women with CIS were, as compared to all other groups, younger at baseline, more educated and more often married or cohabiting. They were also more likely to have had an early menarche, to be pre/perimenopausal and to be current smokers. Women with invasive breast cancer (and available tissue) had the highest age at menarche and were more likely to have a late menopause as compared to the other breast cancer groups. Noncases had the lowest exposure to HRT (Table I). All other potential risk factors were evenly distributed among cases and noncases.

Table I. Distribution of Risk Factors in Different Breast Cancer Cases and Noncases

Factor	Nonbreast cancer cases (n = 15,837)	Cancer in situ (n = 72)	Invasive breast cancer with tissue (n = 501)	Invasive breast cancer with no tissue or bilateral cancer (n = 49)
Column percent (mean and SD in italics). Percentages do not always add up to 100% due to missing.
Age at baseline, yr (16,459)
Mean (SD)	57.3 (7.9)	55.6 (7.4)	57.3 (7.2)	57.8 (7.5)
Education (16,417)
O-level college	70	58	70	65
A-level college	7	7	6	6
University	23	35	24	29
Type of occupation (16,294)
Manual worker	38	29	33	29
Nonmanual worker	54	60	59	65
Employer–self-employed	8	7	7	4
Married/cohabiting (16,455)
No	33	25	35	43
Yes	67	75	65	57
Age at menarche (16,336)
≤12	22	29	21	16
>12 to <15	53	54	51	61
≥15	24	15	28	20
Age at first childbirth (16,459)
Nullipara	13	17	13	12
≤20	17	8	14	22
>20 to ≤25	35	36	35	33
>25 to ≤30	24	24	26	20
>30	9	11	10	8
Bilateral oophorectomy (16,459)
No	99	100	98	98
Yes	1	0	2	2
Age at menopause (16,200)
Pre/Perimenopausal	33	46	32	25
≤45	13	15	11	27
>45 to <53	38	31	39	33
≥53	14	8	16	12
Exposure to OC (16,444)
No	51	47	49	57
Yes	49	53	51	41
Current use of HRT (16,459)
No	82	72	70	71
ERT	6	6	5	2
PRT	1	1	1	2
CHRT	11	21	24	22
Height (16,434)
Mean (SD)	1.64 (6.1)	1.64 (6.2)	1.64 (5.8)	1.65 (6.1)
Body mass index (16,433)
Mean (SD)	25.3 (4.5)	25.4 (4.2)	25.7 (4.3)	25.4 (4.0)
Alcohol consumption (16,424)
Nothing last year (teetotaler)	11	18	10	14
Something last year (not last month)	12	6	11	10
Something last month	76	76	79	76
Smoking (16,453)
Never	44	33	42	45
Current	28	39	27	29
Ex	28	28	31	27
Age at diagnosis, yr
Mean (SD)	–	60.7 (8.7)	62.8 (8.3)	63.6 (14.4)

Distribution of risk factors in different parity classes

Nulliparous women were more highly educated, more often nonmanual workers and had to a higher extent not been exposed to oral contraceptives, as compared to all other parity categories (Supporting Information Table 1). Nulliparous women were also more likely not to be married or cohabiting as compared to all other groups except for women with missing information on parity (Supporting Information Table 1). Women with 3 or more children were younger at first childbirth, as compared to all other groups except for women with missing information on parity (Supporting Information Table 1). All other factors were evenly distributed between different categories for parity.

Parity in relation to risk of different breast cancer subgroups

Nulliparity was associated with a high risk of breast cancer as compared to uniparous women, but this association did not reach statistical significance (Table II). There was a statistically significant increased risk of CIS, invasive grade III and invasive HER2 positive (2+ 3+) tumours, associated with nulliparity (Tables II and III). Nulliparity was also associated with large tumours, tumours with high proliferation (Ki67) as well as with tumours with high expression of the tumour oncogene cyclin D1 and/or a low expression of tumour suppressor gene p27 (Tables II and III), but these associations did not reach statistical significance. The breast cancer risk in relation to axillary lymph node status was similar in all parity groups (Table II).

Table II. Risk of Clinico-Pathological Markers in Relation to Parity Class

Number of children	Number of cases	Incidence/100,000	Unadjusted RR	RR1
1 Adjusted for age at baseline, education, socialeconomic status, marrital status, age at menarche, age at first birth, oophorectomy, age at menopause, oral contraceptive use, hormone replacement therapy use, bmi, alcoholconsumption, smoking and height. 2 All cases with unilateral invasive breast cancer with tissue samples avaible for examintation. 3 Reference group in heterogeneity analyses.
All in study2
Nullipara	65	314	1.15 (0.84–1.57)	1.39 (0.92–2.08)
1	98	276	1.00	1.00
2	216	311	1.12 (0.89–1.43)	1.16 (0.91–1.49)
≥3	110	158	1.01 (0.77–1.33)	1.10 (0.82–1.47)
Missing	12	352	1.21 (0.66–2.21)	–
Total	501	298	p-trend = 0.70	p-trend = 0.55
CIS
Nullipara	12	60	1.46 (0.68–3.16)	3.15 (1.00–9.94)
1	14	39	1.00	1.00
2	31	45	1.13 (0.60–2.13)	1.15 (0.60–2.21)
≥3	12	31	0.77 (0.36–1.67)	0.86 (0.38–1.94)
Missing	3	88	2.37 (0.68–8.28)	–
Total	72	43	p-trend = 0.20	p-trend = 0.68
Size ≤20 mm3
Nullipara	40	103	0.99 (0.67–1.46)	1.18 (0.72–1.93)
1	70	197	1.00	1.00
2	147	212	1.07 (0.81–1.42)	1.09 (0.81–1.46)
≥3	87	221	1.12 (0.82–1.53)	1.21 (0.86–1.69)
Missing	7	205	0.97 (0.45–2.12)	–
Total	351	209	p-trend = 0.43	p-trend = 0.27
Size >20 mm
Nullipara	25	121	1.53 (0.90–2.63)	1.89 (0.91–3.91)
1	28	79	1.00	1.00
2	67	97	1.22 (0.79–1.90)	1.32 (0.83–2.09)
≥3	23	59	0.74 (0.43–1.28)	0.83 (0.46–1.48)
Missing	5	147	1.82 (0.70–4.75)	–
Total	148	88	p-trend = 0.05 (0.048)	p-trend = 0.54
Axillary lymph node neg3
Nullipara	46	222	1.14 (0.78–1.65)	1.56 (0.95–2.58)
1	70	197	1.00	1.00
2	145	209	1.06 (0.79–1.40)	1.11 (0.83–1.49)
≥3	71	181	0.91 (0.66–1.27)	1.02 (0.72–1.45)
Missing	9	89	1.26 (0.63–2.53)	–
Total	341	203	p-trend = 0.34	p-trend = 0.92
Axillary lymph node pos
Nullipara	19	92	1.26 (0.70–2.27)	1.30 (0.63–2.70)
1	26	73	1.00	1.00
2	67	97	1.32 (0.84–2.07)	1.31 (0.82–2.09)
≥3	36	92	1.25 (0.75–2.07)	1.30 (0.76–2.22)
Missing	3	89	1.18 (0.36–3.92)	–
Total	151	90	p-trend = 0.63	p-trend = 0.36
Grade I3
Nullipara	13	63	0.68 (0.36–1.28)	0.74 (0.33–1.64)
1	33	93	1.00	1.00
2	63	91	0.97 (0.64–1.48)	0.95 (0.61–1.47)
≥3	31	79	0.85 (0.52–1.38)	0.90 (0.53–1.51)
Missing	5	147	1.57 (0.61–4.04)	–
Total	145	86	p-trend = 0.75	p-trend = 0.67
Grade II
Nullipara	30	145	1.37 (0.85–2.21)	1.34 (0.73–2.43)
1	38	107	1.00	1.00
2	100	144	1.34 (0.92–1.95)	1.44 (0.98–2.12)
≥3	53	135	1.25 (0.83–1.90)	1.36 (0.87–2.12)
Missing	5	147	1.25 (0.49–3.18)	–
Total	226	135	p-trend = 0.75	p-trend = 0.21
Grade III
Nullipara	21	102	1.40 (0.78–2.48)	2.93 (1.29–6.64)
1	26	73	1.00	1.00
2	53	77	1.04 (0.65–1.67)	1.05 (0.65–1.72)
≥3	26	66	0.90 (0.52–1.55)	1.00 (0.56–1.78)
Missing	2	59	0.77 (0.18–3.25)	–
Total	128	76	p-trend = 0.20	p-trend = 0.97

Table III. Risk of Immunohistochemical Markers in Relation to Parity Class

Number of children	Number of cases	Incidence/100,000	Unadjusted RR	RR1
1 Adjusted for age at baseline, education, socialeconomic status, marrital status, age at menarche, age at first birth, oophorectomy, age at menopause, oral contraceptive use, hormone replacement therapy use, bmi, alcoholconsumption, smoking and height. 2 Reference group in heterogeneity analyses.
Ki67 low (≤10%)2
Nullipara	39	189	1.09 (0.73–1.62)	1.29 (0.77–2.17)
1	62	175	1.00	1.00
2	137	198	1.13 (0.83–1.52)	1.18 (0.87–1.62)
≥3	72	184	1.04 (0.74–1.47)	1.15 (0.80–1.65)
Missing	9	264	1.41 (0.70–2.86)	–
Total	319	190	p-trend = 0.94	p-trend = 0.48
Ki67 high (>10%)
Nullipara	24	116	1.80 (1.02–3.19)	1.95 (0.93–4.10)
1	23	65	1.00	1.00
2	53	77	1.18 (0.72–1.92)	1.21 (0.73–2.01)
≥3	25	64	0.98 (0.56–1.72)	1.02 (0.56–1.87)
Missing	1	29	0.43 (0.06–3.21)	–
Total	126	75	p-trend = 0.09	p-trend = 0.98
HER2 (0 to 1+)2
Nullipara	44	213	1.16 (0.79–1.69)	1.27 (0.79–2.05)
1	66	186	1.00	1.00
2	167	241	1.29 (0.97–1.71)	1.30 (0.97–1.75)
≥3	78	199	1.06 (0.76–1.47)	1.13 (0.80–1.60)
Missing	8	235	1.16 (0.55–2.42)	–
Total	363	216	p-trend = 0.87	p-trend = 0.55
HER2 (2+ to 3+)
Nullipara	15	73	1.98 (0.94–4.17)	3.24 (1.02 –10.25)
1	13	37	1.00	1.00
2	15	22	0.59 (0.28–1.25)	0.68 (0.32–1.48)
≥3	13	33	0.91 (0.42–1.96)	1.16 (0.51–2.67)
Missing	2	59	1.61 (0.36–7.28)	–
Total	58	35	p-trend = 0.02	p-trend = 0.74
Cyclin D1 low (≤10%)2
Nullipara	46	223	1.22 (0.83–1.78)	1.35 (0.83–2.18)
1	65	184	1.00	1.00
2	147	213	1.15 (0.86–1.55)	1.16 (0.86–1.57)
≥3	83	212	1.15 (0.83–1.59)	1.19 (0.85–1.69)
Missing	9	264	1.40 (0.70–2.83)	–
Total	350	208	p-trend = 0.87	p-trend = 0.34
Cyclin D1 high (>10%)
Nullipara	17	82	1.50 (0.78–2.86)	2.15 (0.88–5.27)
1	20	56	1.00	1.00
2	42	61	1.07 (0.63–1.81)	1.29 (0.74–2.25)
≥3	16	41	0.71 (0.37–1.38)	0.92 (0.46–1.87)
Missing	0	–	–	–
Total	95	57	p-trend = 0.06	p-trend = 0.88
P27 low (≤10%)2
Nullipara	23	112	1.32 (0.76–2.26)	2.03 (0.99–4.14)
1	30	85	1.00	1.00
2	63	91	1.07 (0.70–1.66)	1.18 (0.75–1.85)
≥3	42	107	1.26 (0.79–2.02)	1.54 (0.93–2.55)
Missing	6	176	2.07 (0.86–5.02)	–
Total	164	98	p-trend = 0.88	p-trend = 0.10
P27 high (>10%)
Nullipara	38	184	1.24 (0.82–1.89)	1.19 (0.70–2.03)
1	53	150	1.00	1.00
2	122	176	1.17 (0.85–1.62)	1.19 (0.85–1.66)
≥3	55	140	0.93 (0.64–1.36)	0.94 (0.63–1.40)
Missing	4	117	0.72 (0.26–1.98)	–
Total	272	162	p-trend = 0.36	p-trend = 0.75

Parity was not associated with any specific breast cancer subgroup defined by ERα, ERβ, PgR and histological type (Supporting Information Table 3).

Women with no information on parity were very few and missing information was not statistically significantly associated with any of the investigated tumour subgroups. Multivariate analyses including subjects with missing information on parity were considered unsuitable due to a very small number of cases in this category.

The analyses did not show any statistically significant trend over parity groups with regard to risk of any breast cancer subgroup (Tables II and III). When analysing heterogeneity between different breast cancer subgroups in relation to association with specific exposure categories, there were no statistically significant findings.

An additional analysis was made for triple-negative (ERα-/PR-/HER2-) breast tumours, but this analysis did not show any statistically significant association with parity (data not shown).

All confounders were tested one by one in the analysis for overall breast cancer risk. Age at first childbirth was the confounder that affected the adjusted risk estimates the most (data not shown). When only adjusting for socioeconomic status, the relative risks were: 1.12 (0.82–1.54) for nulliparous women, 1.12 (0.89–1.43) for women with 2 children and 1.03 (0.78–1.35) for women with 3 or more children.

Distribution of risk factors in different classes of age at first birth

Women older than 30 years of age at first childbirth were more often highly educated and had been less exposed to oral contraceptives, as compared to all other age categories (Supporting Information Table 2). Women younger than 20 years of age at first childbirth were more often current smokers as compared to all other age categories (Supporting Information Table 2). All other factors were evenly distributed between different categories for age at first childbirth.

Age at first childbirth in relation to risk of different breast cancer subgroups

Older age at first childbirth was associated with a high breast cancer risk as compared to women with their first childbirth before the age of 20, but this association did not reach statistical significance (Table IV). The risk associated with grade III tumours, was higher among women aged 20–25, and women >30, as compared to women <20 years of age at first childbirth (Table IV). Older age at first childbirth was also associated with high expression of cyclin D1, and low expression of p27 (Table V). Late first childbirth was associated with lobular breast cancer (Supporting Information Table 4).

Table IV. Risk of Clinico-Pathological Markers in Relation to Age at First Birth

Age at first childbirth	Number of cases	Incidence/100,000	Unadjusted RR	RR1
1 Adjusted for age at baseline, education, socialeconomic status, marrital status, age at menarche, parity, oophorectomy, age at menopause, oral contraceptive use, hormone replacement therapy use, bmi, alcoholconsumption, smoking and height. 2 All cases with unilateral invasive breast cancer with tissue samples avaible for examintation. 3 Reference group in heterogeneity analyses.
All in study2
≤20	69	246	1.00	1.00
>20–≤25	176	296	1.20 (0.91–1.59)	1.19 (0.89–1.58)
>25–≤30	129	314	1.28 (0.95–1.71)	1.27 (0.93–1.73)
>30	50	323	1.32 (0.92–1.90)	1.39 (0.94–2.07)
Missing	12	340	1.30 (0.71–2.42)	–
Total	436	295	p-trend = 0.10	p-trend = 0.08
CIS
≤20	6	21	1.00	1.00
>20–≤25	26	44	2.05 (0.84–4.97)	2.16 (0.87–5.32)
>25–≤30	17	41	1.94 (0.77–4.92)	2.16 (0.82–5.72)
>30	8	52	2.43 (0.84–7.00)	2.53 (0.81–7.86)
Missing	3	85	4.28 (1.07–17.13)	–
Total	60	41	p-trend = 0.16	p-trend = 0.18
Size ≤20 mm3
≤20	50	178	1.00	1.00
>20–≤25	128	215	1.21 (0.87–1.67)	1.20 (0.86–1.68)
>25–≤30	94	229	1.28 (0.91–1.81)	1.28 (0.89–1.85)
>30	32	207	1.16 (0.75–1.81)	1.26 (0.78–2.03)
Missing	7	198	1.04 (0.47–2.30)	–
Total	311	211	p-trend = 0.33	p-trend = 0.25
Size >20 mm
≤20	19	68	1.00	1.00
>20–≤25	47	79	1.17 (0.69–1.99)	1.11 (0.65–1.92)
>25–≤30	34	83	1.22 (0.70–2.14)	1.17 (0.65–2.12)
>30	18	116	1.72 (0.91–3.29)	1.68 (0.83–3.38)
Missing	5	141	2.03 (0.75–5.46)	–
Total	123	83	p-trend = 0.12	p-trend = 0.18
Axillary lymph node neg3
≤20	40	143	1.00	1.00
>20–≤25	122	205	1.44 (1.01–2.05)	1.42 (0.99–2.05)
>25–≤30	88	214	1.50 (1.03–2.18)	1.48 (0.99–2.19)
>30	36	233	1.64 (1.04–2.57)	1.70 (1.05–2.77)
Missing	9	255	1.67 (0.81–3.45)	–
Total	295	200	p-trend = 0.03	p-trend = 0.046
Axillary lymph node pos
≤20	26	93	1.00	1.00
>20–≤25	52	87	0.94 (0.59–1.51)	0.92 (0.57–1.50)
>25–≤30	37	90	0.97 (0.59–1.61)	1.00 (0.59–1.70)
>30	14	90	0.98 (0.51–1.87)	1.08 (0.54–2.19)
Missing	3	85	0.91 (0.27–3.01)	–
Total	132	89	p-trend = 0.98	p-trend = 0.74
Grade I3
≤20	19	68	1.00	1.00
>20–≤25	55	92	1.36 (0.81–2.30)	1.24 (0.73–2.11)
>25–≤30	40	97	1.44 (0.83–2.48)	1.20 (0.67–2.14)
>30	13	84	1.25 (0.62–2.52)	1.09 (0.51–2.33)
Missing	5	141	2.07 (0.77–5.57)	–
Total	132	89	p-trend = 0.41	p-trend = 0.89
Grade II
≤20	37	132	1.00	1.00
>20–≤25	67	113	0.85 (0.57–1.27)	0.84 (0.55–1.26)
>25–≤30	65	158	1.20 (0.80–1.79)	1.17 (0.76–1.81)
>30	22	142	1.08 (0.64–1.83)	1.12 (0.63–1.98)
Missing	5	141	0.98 (0.38–2.50)	–
Total	196	133	p-trend = 0.27	p-trend = 0.26
Grade III
≤20	13	46	1.00	1.00
>20–≤25	54	91	1.96 (1.07–3.59)	2.10 (1.14–3.89)
>25–≤30	24	58	1.26 (0.64–2.48)	1.54 (0.76–3.12)
>30	14	90	1.96 (0.92–4.17)	2.67 (1.19–6.02)
Missing	2	57	1.16 (0.26–5.16)	–
Total	107	73	p-trend = 0.45	p-trend = 0.11

Table V. Risk of Immunohistochemical Markers in Relation to Age at First Birth

Number of children	Number of cases	Incidence/100,000	Unadjusted RR	RR1
1 Adjusted for age at baseline, education, socialeconomic status, marrital status, age at menarche, parity, oophorectomy, age at menopause, oral contraceptive use, hormone replacement therapy use, bmi, alcoholconsumption, smoking and height. 2 Reference group in heterogeneity analyses.
Ki67 low (≤10%)2
≤20	44	157	1.00	1.00
>20–≤25	110	185	1.18 (0.83–1.67)	1.16 (0.81–1.67)
>25–≤30	85	207	1.32 (0.92–1.90)	1.30 (0.88–1.91)
>30	32	207	1.33 (0.84–2.09)	1.40 (0.86–2.30)
Missing	9	255	1.51 (0.74–3.11)	–
Total	280	190	p-trend = 0.13	p-trend = 0.13
Ki67 high (>10%)
≤20	21	75	1.00	1.00
>20–≤25	39	66	0.87 (0.51–1.49)	0.88 (0.51–1.51)
>25–≤30	29	71	0.94 (0.54–1.65)	1.02 (0.56–1.86)
>30	12	78	1.04 (0.51–2.11)	1.22 (0.57–2.64)
Missing	1	28	0.36 (0.05–2.66)	–
Total	102	69	p-trend = 0.90	p-trend = 0.57
HER2 (0 to 1+)2
≤20	55	196	1.00	1.00
>20–≤25	128	215	1.10 (0.80–1.50)	1.07 (0.78–1.48)
>25–≤30	95	231	1.18 (0.84–1.64)	1.16 (0.82–1.66)
>30	33	213	1.09 (0.71–1.68)	1.20 (0.75–1.92)
Missing	8	226	1.06 (0.50–2.23)	–
Total	319	216	p-trend = 0.48	p-trend = 0.34
HER2 (2+ to 3+)
≤20	6	21	1.00	1.00
>20–≤25	14	24	1.10 (0.42–2.87)	1.23 (0.46–3.28)
>25–≤30	15	37	1.71 (0.66–4.41)	2.03 (0.74–5.56)
>30	6	39	1.81 (0.59–5.62)	2.06 (0.60–7.12)
Missing	2	57	2.62 (0.52–13.19)	–
Total	43	29	p-trend = 0.14	p-trend = 0.11
Cyclin D1 low (≤10%)2
≤20	53	189	1.00	1.00
>20–≤25	121	203	1.08 (0.78–1.49)	1.08 (0.78–1.51)
>25–≤30	92	224	1.19 (0.85–1.66)	1.21 (0.85–1.74)
>30	29	187	0.99 (0.63–1.56)	1.10 (0.67–1.79)
Missing	9	255	1.31 (0.65–2.67)	–
Total	304	206	p-trend = 0.65	p-trend = 0.44
Cyclin D1 high (>10%)
≤20	12	43	1.00	1.00
>20–≤25	30	50	1.17 (0.60–2.29)	1.15 (0.58–2.27)
>25–≤30	20	49	1.13 (0.55–2.31)	1.14 (0.54–2.42)
>30	16	103	2.44 (1.16–5.17)	2.69 (1.18–6.12)
Missing	0	–	–	–
Total	78	53	p-trend = 0.04	p-trend = 0.04
P27 low (≤10%)2
≤20	23	82	1.00	1.00
>20–≤25	50	84	1.03 (0.63–1.68)	1.15 (0.69–1.90)
>25–≤30	43	105	1.28 (0.77–2.12)	1.61 (0.94–2.76)
>30	19	123	1.49 (0.81–2.74)	2.23 (1.15–4.35)
Missing	6	170	2.06 (0.83–5.09)	–
Total	141	96	p-trend = 0.11	p-trend = 0.01
P27 high (>10%)
≤20	42	150	1.00	1.00
>20–≤25	96	161	1.08 (0.75–1.55)	1.01 (0.70–1.47)
>25–≤30	69	168	1.12 (0.76–1.64)	1.01 (0.67–1.52)
>30	23	149	1.00 (0.60–1.66)	0.93 (0.54–1.60)
Missing	4	113	0.69 (0.25–1.92)	–
Total	234	159	p-trend = 0.82	p-trend = 0.88

Women with older age at first childbirth were significantly more likely to be lymph node negative as compared to women with a first childbirth before 20 years of age (Table IV). There were statistically significant trends related to increasing age at first childbirth with regard to the risk of tumours with negative axillary lymph nodes, over expression of cyclin D1, low p27 expression and lobular type (Table V). There was no association between age at first childbirth and specific subgroups by invasiveness, Ki67, HER2, ERα, ERβ or PgR status (Tables IV, V and Supporting Information Table 4).

In the heterogeneity test, the risk associated with a older age at first childbirth (>30) was statistically significantly higher in cyclin D1 over expressing tumours as compared to the risk of cyclin D1 negative tumours, p value 0.046 (TableV).

Triple-negative breast tumours were evaluated with regard to age at first childbirth but did not reveal any statistically significant association (data not shown).

All confounders were tested one by one in the analysis for overall breast cancer risk. HRT was the confounder that affected the adjusted risk estimates the most (data not shown). When only adjusting for socioeconomic status the adjusted relative risks associated with age at first birth were: 1.17 (0.89–1.55), 1.23 (0.91–1.65) and 1.26 (0.87–1.81) for the 3 categories, respectively.

Discussion

The aim of our study was to investigate parity and age at first childbirth in relation to different biological breast cancer characteristics and we found that nulliparous women had a statistically significant high risk of grade III and HER2 over expressing tumours and a high, although not statistically significant, risk of tumours with high Ki67, high cyclin D1 and low p27 expression. Further our study showed that women with late first childbirth had a higher risk of developing tumours with lobular type, grade III, high expression of cyclin D1, low expression of p27, as well as negative axillary lymph node status.

Methodological considerations

The participants in the MDCS were probably selected towards higher socioeconomic groups and women in MDCS had a higher incidence of breast cancer as compared to the rest of the female population in Malmö.8 Since there is no information on exposure of risk factors in women outside the cohort, observed incidence rates may not be applicable to all parity groups or age groups. However, since the aim was to compare tumour characteristics between different classes of parity and age at first childbirth and that there was a wide distribution with regard to these factors; we consider that it is possible to make internal comparisons between subjects with different levels of the studies' exposures and, hence obtain valid relative risks. Moreover, the analyses were adjusted for sociodemographic factors, i.e. factors related to participation. Hence, we believe that our estimations of relative risks were less likely to have been affected by a potential selection bias.

Information on parity and age at first childbirth was retrieved from a questionnaire provided at baseline examinations where all women were 44 years or older, thus unlikely to have additional children following baseline. A previous study has confirmed self-reported parity to be highly accurate,14 hence, we consider this information to be valid.

All tumour endpoints were retrieved by record linkage to The Swedish Cancer Registry. This is a nation-wide registry and all cancer cases in Sweden are to be reported to this registry. This registry has previously been validated in Malmö and the completeness was 99% regarding breast cancer.15

Data on potential risk factors for breast cancer were available and they were subsequently adjusted for in the multivariate analyses. One limitation of our study is that there was no information on the total number of pregnancies. However, two other studies have shown no association between spontaneous or induced abortion with breast cancer risk,16, 17 and a third study has shown no association between abortion and survival after breast cancer,18 indicating that the total number of pregnancies may be less relevant for breast cancer risk.

Tumour classification with regard to the biomarkers were analysed with TMA and this technique is well documented for tissue screening and two cores are considered to be sufficient to get a representative sample.19, 20

It is important to consider a type I error due to many analyses in our study. However, with regard to the different characteristics, there is a clear pattern as biologically correlated tumour characteristics, such as grade, HER2 and Ki67 status have similar associations with parity, indicating that the findings in our study were indeed true findings. However, due to few individuals in some analyses, the confidence intervals were wide and the statistical power was relatively low. This may have led to a type II error in some comparisons and these results should be interpreted with caution.

Previous studies

In accordance with the results of our study, nulliparity has been associated with increased risk of breast cancer in many studies.1, 5, 21 Nulliparous women were at higher risk of CIS, grade III and HER2 over expressing tumours. At least one other study has confirmed the findings related to grade III tumours1 but no other study has confirmed the relationship between nulliparity and HER2 status. The results of our study also suggest that nulliparous women are at higher risk of larger and high proliferating tumours. Daling et al. found similar results regarding nulliparous women and high levels of Ki67.22 To our knowledge, no other study has investigated parity in relation to cyclin D1, p27 or CIS.

High parity was associated with a lower risk of breast cancer in the present study and former studies show similar results.2 High parity was not related to any specific hormone receptor status as proposed previously and reviewed by Ma et al.23 and Ursin et al.24

Older age at first childbirth was related to larger tumours, grade III, lobular tumours and over expression of cyclin D1 along with low expression of p27. Previous studies confirm the association with large tumours and stage at diagnosis.1, 2, 5, 21, 25–28 To our knowledge, no study has investigated any association between age at first childbirth and tumour characteristics such as cyclin D1, p27, grade III and histological type.

It has previously been shown that women with triple-negative tumours have a poorer prognosis,29 one reason being that they do not respond to adjuvant therapy to the same extent as receptor positive tumours.30 In our study, we could not find any statistically significant relation between risk for triple-negative breast tumours and parity or age at first childbirth.

Potential explanations

The results of our study suggest that nulliparous and parous women develop specific kinds of breast cancer subgroups. Lagerlund et al.31 found that nulliparous women had a lower rate of attendance to mammography screening than women with two children. This could possibly lead to a later diagnosis, more advanced tumours (and probably a older age at diagnosis), and low breast cancer risk. However, mean age at diagnosis for nulliparous women was slightly lower as compared to women with one child, and the breast cancer risk was higher in nulliparous women, indicating that there was no major detection bias in our study.

Another possible explanation is that nulliparous women and women with a late first childbirth are more susceptible to promoting factors, leading to more aggressive tumours. Breast tissue undergoes proliferation and maturation during pregnancy32 and it has been shown that time since birth may have an impact on breast cancer prognosis.33

It is possible to hypothesise that the mechanism behind the observation in nulliparous women and women with a late age at first childbirth are similar as both these categories have breast tissue that is not fully evolved. Their breast tissue could, hence, be more susceptible to initiating and promoting factors.34

Russo et al. have evaluated different genomic changes in breast epithelium of women and found that the genome in parous women with no breast cancer is different from both nulliparous and parous women with breast cancer.35 It is possible to hypothesise that such differences will contribute to protect against specific phenotypes for breast tumours in these women.

Moreover, at least one in vivo study has shown that pregnancy induces alterations in the genes of the breast epithelium. The study shows a downregulation of growth hormone genes (e.g. IGF-1) and upregulation of growth inhibitor genes (e.g. TGF-β3).36

Conclusions

Nulliparous women have an increased risk of breast tumours with a prognostically unfavourable profile. Late first childbirth was associated with a high risk of more aggressive breast cancer subgroups.