Physical activity from menarche to first pregnancy and risk of breast cancer

Abstract

Breast tissue is particularly susceptible to exposures between menarche and first pregnancy, and a longer interval between these reproductive events is associated with elevated breast cancer risk. Physical activity during this time period may offset breast cancer risk, particularly for those at highest risk with longer menarche-to-first-pregnancy intervals. We used data from 65,576 parous women in the Nurses’ Health Study II free of cancer in 1989 (baseline) and recalled their leisure-time physical activity at ages 12–34 in 1997. Current activity was collected at baseline and over follow-up. Relative risks (RRs) were estimated using multivariable Cox proportional hazards regression models. Between 1989 and 2011, 2,069 invasive breast cancer cases were identified. Total recreational activity between menarche and first pregnancy was not significantly associated with the risk of breast cancer. However, physical activity between menarche and first pregnancy was associated with significantly lower breast cancer risk among women in the highest category of a menarche-to first-pregnancy interval (≥20 years; RR for the highest versus the lowest quartile = 0.73, 95% confidence interval = 0.55–0.97; P_trend = 0.045; P_interaction = 0.048). This was not observed in women with a shorter interval. Physical activity between menarche and first pregnancy was associated with a lower risk of breast cancer among women with at least 20 years between these reproductive events. This may provide a modifiable factor that women can intervene on to mitigate their breast cancer risk associated with a longer interval.

Physical activity is associated with a lower breast cancer risk. A 12% lower risk was reported among the most physically active women as compared with the least active women in a recent meta-analysis of 31 prospective cohort studies.¹ In this meta-analysis, the inverse association between physical activity and breast cancer risk is observed in both pre- and post-menopausal women, with relative risk (RR) = 0.77 and RR = 0.87, respectively. However, the periods of life that may be the most relevant to the protective benefits of physical activity on breast cancer risk remain uncertain. Recreational physical activity is consistently associated with reduced breast cancer risk in case-control studies in black, white² and Asian-American women³ including both pre-⁴ and post-menopausal^5,6 breast cancer. While the evidence for an association of average lifetime activity is clearest, greater early life activity during the 10 years following menarche is also strongly protective. Six prospective studies assessed the influence of timing of physical activity on breast cancer risk,^7–12 with five reporting no association for adolescent and young adult activity.^8–12 However, in the Nurses’ Health Study II (NHS-II), we previously observed an inverse relationship between physical activity from ages 18 and 22 and premenopausal breast cancer risk.⁷

What’s new?

Physical activity is associated with lower breast cancer risk, but the periods of life offering most protective benefits remain unclear. With a longer interval between menarche and first pregnancy known to be associated with elevated breast cancer risk, here the authors set to examine the influence of physical activity between the two reproductive events. Physical activity was associated with a lower risk of breast cancer among women with at least 20 years between menarche and first pregnancy. Physical activity may thus offer women an option to mitigate breast cancer risk associated with a longer interval between menarche and first pregnancy.

Both human and animal data have demonstrated that from menarche to first pregnancy represents a window of time when breast tissue is relatively undifferentiated and potentially more susceptible to carcinogenic stimuli.^13–19 Pregnancy-induced cellular differentiation and other biological changes confer a long-term reduction in breast cancer risk.^20–22 Earlier age at menarche, later age at first pregnancy, and thus a longer interval between these two reproductive events are established breast cancer risk factors.^14,16,23 Consumption of alcohol, a breast carcinogen,²⁴ between menarche and first pregnancy has been associated with greater risks of both invasive breast cancer and proliferative benign breast disease, a well-confirmed risk marker for breast cancer.¹⁵ The influence of physical activity during this particularly vulnerable window on subsequent breast cancer risk has not been explicitly evaluated.

We, therefore, examined the association between physical activity from menarche to first pregnancy and risk of breast cancer, by length of interval between these reproductive events. We hypothesized that women at particularly high breast cancer risk due to longer intervals may experience a greater benefit from physical activity, compared with their shorter interval counterparts.

Methods

Study population

The NHS-II prospective cohort was established in 1989 when 116,430 female registered nurses aged 24–44 years completed a mailed questionnaire about their medical history, reproductive history and lifestyle. Follow-up questionnaires mailed biennially updated information on these factors and captured incident medical events. The cumulative follow-up through 2009 exceeded 90% of the total possible person-years.⁷ Participants implied consent with the return of questionnaires. This study was approved by the Human Subjects Committees at the Harvard T.H. Chan School of Public Health and Brigham and Women’s Hospital.

Physical activity assessment

Physical activity was assessed in 1989, 1991, 1997 and subsequently every 4 years. Participants indicated the average time per week they spent in the following activities over the past year: walking or hiking outdoors, jogging, running, bicycling, racquet sports, swimming laps and calisthenics or aerobics. In 1997, they were also asked to recall their usual time per week spent in moderate or vigorous recreational activities and walking to and from school or work during each of the following five age periods: 12–13, 14–17, 18–22, 23–29 and 30–34 years.

Leisure-time physical activity measures have been demonstrated to have reasonable reliability and validity in a subgroup of NHS-II participants. Activity during the first three age periods that were recalled 3 years apart by 160 participants were reproducible (correlation r = 0.76 for vigorous, r = 0.70 for vigorous plus moderate and r = 0.64 for total activity).²⁵ In a validation study among 147 NHS-II participants, the measure of physical activity in the previous year was comparable with 7-day activity recalls (r = 0.79) and four 7-day activity diaries (r = 0.62).²⁶

The intensity of physical activity was analyzed as metabolic equivalent (MET) hours per week (hrs/wk) of all activities combined (referred to as total activity). According to the Compendium of Physical Activity by Ainsworth et al.,²⁷ we assigned a MET score to each activity. Intensity was then categorized based on national guidelines for physical activity.^28,29 Vigorous activity (at least 7.0 METs) included jogging, running, bicycling, racquet sports and lap-swimming. Calisthenics or aerobics and other aerobic activity were categorized as moderate (4.0–6.0 METs). Walking was assigned a MET score based on self-reported pace and categorized separately. Consistent with prior analyses of the NHS-II data,^7,30 we assigned MET scores of 7.0, 4.5 and 3.0 to vigorous, moderate and walking activities, respectively, for each of the earlier age periods.

The MET-hrs/wk for each activity type (vigorous, moderate and walking) were computed by multiplying duration (hours per week) by its MET score. Total activity was estimated by summing all activities. We used linear interpolation to estimate annual adult activity between the last life period report for ages 30–34 and the past year assessment, under the assumption that activity changes at a stable rate.³⁰ Average physical activity between menarche and first pregnancy (of at least 6 months gestational age) was estimated by multiplying physical activity (MET-hrs/week) in each age period before the first pregnancy by the length of the corresponding period, summing the contributions from each age period and dividing by the interval length, in years.

Similarly, we also estimated average physical activity from first pregnancy to menopause (or current age for premenopausal women) as a time-varying covariate to control for potential confounding by physical activity after first pregnancy.

Incident breast cancer

Incident cancer cases were ascertained on biennial follow-up questionnaires and medical record review or by a search of the National Death Index. A review of pathology reports, which were obtained for 92% of the self-reported diagnoses, confirmed 99% of self-reported breast cancers.³¹ Carcinoma in situ cases were excluded. Estrogen receptor (ER) and progesterone receptor (PR) status were abstracted from pathology reports.

Reproductive and other covariates

The 1989 baseline questionnaire captured several potential covariates, including age at menarche, pregnancy history, height, body weight, childhood body shape (somatotype), history of benign breast disease, family history of breast cancer in mother and/or sister, oral contraceptive use, breastfeeding duration, alcohol consumption, menopausal status and post-menopausal hormone use. Alcohol consumption in the past year was updated in 1991 and subsequently every 4 years. Family history of breast cancer was updated in 1997 and subsequently every 4 years. All other time-varying covariates were updated biennially.

Statistical analysis

Among 116,430 participants in the NHS-II cohort, 82,038 participated in recalling their early–life physical activity in 1997. Women participating in the 1997 recall were similar to those not participating, with the exception of nonresponders having slightly greater weight gain since age 18 (18.3 vs. 16.5 kg) and being more likely to not consume alcohol (39.3 vs. 37.3% nondrinkers). Of these, we excluded participants (n = 16,462) with a history of cancer at baseline (except non-melanoma skin cancer), no information on reproductive characteristics or remained nulliparous through 2009. Thus, 65,576 participants were included in the following analyses.

We categorized physical activity into quartiles based on the distribution among all eligible participants. The interval between menarche to first pregnancy was calculated in years and categorized as <15, 15–19 and ≥20 years. Person-years were calculated from the return date of the 1989 questionnaire until the date of breast cancer diagnosis, death, dropout, other self-reported cancer (except nonmelanoma skin cancer) or June 1, 2011, whichever occurred first.

We used Cox proportional hazards models to estimate RRs and 95% confidence intervals (CIs) for the main effects of physical activity and the menarche/first birth interval separately and for their interaction. The models were stratified by age (continuous) and questionnaire year (continuous) and were adjusted for menopausal status (premenopausal, post-menopausal or unknown), average body size between ages 5 and 10 (somatotype pictogram 1, 1.5–2, 2.5–3, 3.5–4.5, ≥5 or unknown), family history of breast cancer in mother or sister(s) (yes or no), history of benign breast biopsy, alcohol consumption (0, 0.1–4.9, 5.0–14.9, ≥15 g/day or unknown), postmenopausal hormone use (never, ever or unknown), total duration of breastfeeding (0, 0.1–11, ≥12 months or unknown), oral contraceptive use (never; past, <4; past, ≥4; current, <4; current, ≥4 years or unknown) height (continuous) and birth index (quintiles). Birth index was calculated by summing total years from each birth to current age (or age at menopause for postmenopausal women) over all births. Current body mass index (BMI kg/m²; quintiles) and weight change since age 18 (kg; quintiles) were added to the multivariable models separately, as they are considered potential intermediates of the relationship between physical activity before first pregnancy and the development of breast cancer. Tests for trend across exposure categories were conducted by using the median value for each category as a continuous variable in the multivariable model.

The interaction between the menarche-first-birth interval length and physical activity was assessed by including a cross-product term in the multivariable-adjusted model. We also estimated the joint relationship of the two exposures, with the presumably highest risk participants with ≥20 years of interval length and the lowest physical activity quartile serving as the reference group.

In a secondary analysis, we categorized participants’ change in physical activity patterns from before their first pregnancy to after, to further evaluate whether associations between physical activity and breast cancer risk could be specifically attributed to the prepregnancy time period. For this analysis, total activity for each period was divided into tertiles, with women in the first tertile defined as “low” and in the third as “high”. Finally, we conducted a sensitivity analysis restricting to incident breast cancer cases diagnosed after 1997 to address the potential for recall bias when including cancer cases prior to 1997.

Results

Among 65,576 parous women, the mean age at menarche was 12.4 years and the mean age at first pregnancy was 26.7 years. Table 1 shows age-adjusted characteristics of participants by categories of interval length (years) between menarche and first pregnancy. A longer interval (≥20 vs. <15 years) was associated with older age at first pregnancy (34.0 vs. 23.0 years), lower birth index (3.4 vs. 21.2), greater current alcohol consumption (≥15 g/d: 5.2 vs. 3.1%), less weight change since age 18 (14.2 vs. 18.0 kg) and greater current physical activity (26.4 vs. 22.7 MET-hrs/wk).

Table 1.

Characteristics¹ of women (n = 65,576) aged 25–44 years in 1989, according to length of time interval between menarche and first full-term pregnancy^†

	Interval between menarche and first pregnancy
	<15 years	15–19 years	≥20 years
No. of participants	31,029	22,712	11,835
Age, mean (SD) (years)	35.2 (4.8)	33.8 (4.6)	33.7 (4.4)
Age at menarche, mean (SD) (years)	12.7 (1.4)	12.2 (1.3)	12.0 (1.4)
Age at first full-term pregnancy,2,3 mean (SD) (years)	23.0 (2.6)	27.9 (1.9)	34.0 (3.1)
Interval between menarche and first full-term pregnancy,2,3 mean (SD) (years)	11.2 (2.5)	16.7 (1.4)	23.0 (3.0)
Birth index (SD)	21.2 (14.5)	11.1 (9.8)	3.4 (5.1)
Total duration of breastfeeding (%)
None	25.2	17.4	15.8
0.1–11 months	34.8	37.5	46.1
≥12 months	40.0	45.1	38.1
Average body size between ages 5 and 10 years,4 mean (SD)	2.5 (1.2)	2.6 (1.2)	2.7 (1.2)
History of benign breast disease (%)	29.1	27.5	28.5
Mother or sister with breast cancer (%)	5.6	6.1	6.2
Height, mean (SD) (in)	64.9 (2.6)	64.9 (2.6)	64.9 (2.6)
Current alcohol intake (%)
Never	42.0	38.2	31.6
>0–<5 g/day	41.9	42.9	43.8
5–<15 g/day	12.9	15.4	19.4
≥15 g/day	3.1	3.4	5.2
Duration and recency of oral contraceptive use (%)
Never	13.9	14.1	16.9
Past <4 years	47.5	39.6	34.4
Past ≥4 years	28.8	34.4	32.0
Current <4 years	2.8	3.6	5.3
Current ≥4 years	7.0	8.3	11.4
Premenopausal (%)	97.2	98.6	99.2
Current body mass index, mean (SD) (kg/m2)	23.9 (4.7)	23.9 (4.7)	23.6 (4.6)
Weight change since age 18, mean (SD) (kg)	18.0 (22.9)	16.6 (22.1)	14.2 (21.8)
Current physical activity, mean (SD) (MET-hrs/wk)	22.7 (32.9)	22.2 (32.7)	26.4 (38.4)
Average physical activity before first pregnancy, mean (SD) (MET-hrs/wk)	46.2 (30.5)	43.1 (27.9)	43.9 (27.4)
Average physical activity after first pregnancy, mean (SD) (MET-hrs/wk)	36.0 (26.3)	35.6 (26.4)	38.2 (28.8)

¹All variables except for age are age-standardized.

²Full-term pregnancy was defined as a pregnancy lasting at least 6 months.

³Age at first pregnancy was assessed in 1989 or during the follow-up through 2009.

⁴Childhood body size was estimated by averaging each participant’s somatograms (1–9 scale) at ages 5 and 10.

Abbreviation: SD: standard deviation.

Between 1989 and 2011, we observed 2,069 incident invasive breast cancer cases. We did not observe an association between physical activity between menarche and first pregnancy and breast cancer risk (for ≥58 vs. <24 MET-hrs/wk, RR = 0.91, 95% CI = 0.79–1.05; P_trend = 0.24; Table 2). Further controlling for age at menarche, current BMI or weight change since age 18 did not significantly change the risk estimates (data not shown).

Table 2.

Physical activity between menarche and first pregnancy and relative risk (RR) of invasive breast cancer among parous women in the Nurses’ Health Study II

Physical activity between menarche and first pregnancy, MET-hrs/wk
	Cases	Person- years	RR (95% CI)1	RR (95% CI)2
<24	590	340,430	1.00 (reference)	1.00 (reference)
24 to <38	474	304,308	0.93 (0.82, 1.04)	0.92 (0.81, 1.04)
38 to <58	501	321,677	0.94 (0.83, 1.06)	0.93 (0.82, 1.06)
≥58	504	336,097	0.92 (0.81, 1.03)	0.91 (0.79, 1.05)
			Ptrend = 0.20	Ptrend = 0.24

¹The analysis was stratified by age and questionnaire year and was adjusted for average childhood body size, alcohol intake, history of benign breast biopsy, family history of breast cancer, height, total duration of breastfeeding, duration and recency of oral contraceptive use, menopausal status, postmenopausal hormone use and birth index.

²The analysis was further adjusted for physical activity after first pregnancy.

Abbreviation: CI: confidence interval.

Breast cancer risk was associated in a dose-dependent manner with the length of time between menarche and first pregnancy (RR = 1.05 per 1-year increment, 95% CI = 1.03–1.06; P_trend < 0.0001; Table 3). Women with the longest interval (≥20 years) had a 60% greater cancer risk versus women with the shortest (<15 years; RR = 1.60, 95% CI = 1.38–1.86).

Table 3.

Length of interval between menarche and first pregnancy and relative risk (RR) of invasive breast cancer among parous women in the Nurses’ Health Study II

Interval between menarche and first pregnancy	Cases	Person-years	RR (95% CI)1
<15 years	889	641,398	1.00 (reference)
15–19 years	772	456,860	1.39 (1.25, 1.54)
≥20 years	408	204,256	1.60 (1.38, 1.86)
Per 1-year increment	2,069	1,302,513	1.05 (1.03, 1.06)
			Ptrend < 0.0001

¹The analyses were stratified by age and questionnaire year and were adjusted for average childhood body size, alcohol intake, history of benign breast biopsy, family history of breast cancer, height, total duration of breastfeeding, duration and recency of oral contraceptive use, menopausal status, postmenopausal hormone use and birth index.

Abbreviation: CI: confidence interval.

The interaction between physical activity between menarche and first pregnancy and the length of the interval between menarche and first pregnancy is detailed in Table 4. No association between physical activity and breast cancer risk was observed among women with <15 years or 15–19 years between menarche and first pregnancy (P_trend = 0.88 and 0.29, respectively); however, a significant inverse association was observed among women with intervals of ≥20 years (P_trend = 0.045). This interaction was statistically significant (P_interaction = 0.048). In the joint relationship model (Fig. 1), compared with women with an ≥20 years between menarche and first pregnancy and the lowest physical activity during this period, there was a 48% lower breast cancer risk for women with the shortest interval length and highest activity level (RR = 0.52, 95% CI = 0.41–0.66).

Table 4.

Physical activity between menarche and first pregnancy and relative risk (RR) of invasive breast cancer according to the length of interval from menarche to first pregnancy among parous women in the Nurses’ Health Study II

Menarche to first pregnancy interval	Physical activity, MET-h/wk	Cases	Person-years	RR (95% CI)1
<15 years	<24	239	168,187	1.00 (reference)
	24 to <38	193	142,287	0.97 (0.80, 1.18)
	38 to <58	217	153,540	1.05 (0.86, 1.28)
	≥58	240	177,384	1.01 (0.81, 1.25)
				Ptrend = 0.88
15–19 years	<24	229	122,058	1.00 (reference)
	24 to <38	180	109,198	0.87 (0.71, 1.07)
	38 to <58	177	113,451	0.81 (0.65, 1.01)
	≥58	186	112,151	0.87 (0.68, 1.10)
				Ptrend = 0.29
≥20 years	<24	122	50,185	1.00 (reference)
	24 to <38	101	52,823	0.82 (0.63, 1.06)
	38 to <58	107	54,686	0.82 (0.63, 1.06)
	≥58	78	46,561	0.73 (0.55, 0.97)
				Ptrend = 0.045
				Pinteraction = 0.048

¹The analysis were stratified by age and questionnaire year and were controlled for average childhood body size, alcohol intake, history of benign breast biopsy, family history of breast cancer, height, total duration of breastfeeding, duration and recency of oral contraceptive use, menopausal status, postmenopausal hormone use, birth index and physical activity after first pregnancy.

Abbreviations: RR: relative risk; CI: confidence interval.

Figure 1.

The joint association between the length of interval (years) between menarche to first pregnancy with physical activity during the same time period with invasive breast cancer risk. The reference group is women with an interval between menarche and first birth ≥20 years and in the lowest quartile of physical activity (<24 MET-hrs/wk) during this time period. The multivariable Cox proposal hazards model was stratified by age and questionnaire year and were adjusted for average childhood body size, alcohol intake, history of benign breast biopsy, family history of breast cancer, height, total duration of breastfeeding, duration and recency of oral contraceptive use, menopausal status, postmenopausal hormone use, birth index and physical activity after first pregnancy.

We did not observe a significant association with high levels of both activity before and after first pregnancy (high/high) compared with low levels of both activity before and after first pregnancy (low/low; RR = 0.91, 95% CI = 0.79–1.04). However, a significantly lower risk was observed for the high/high activity pattern in women with an interval of at least 20 years between menarche and first pregnancy (RR = 0.70, 95% CI = 0.49–0.98; Supporting Information Table 1). Results were similar when we repeated analyses excluding 347 cancer cases occurring prior to 1997.

Discussion

In this large prospective cohort of parous women, we observed a significant interaction between leisure-time physical activity between menarche and first pregnancy and the length of this reproductive interval with breast cancer risk. Physical activity was significantly associated with a 27% lower breast cancer risk for high risk women with the longest time window between menarche and first pregnancy (≥20 years), comparing women in the top versus bottom quartiles of activity. This inverse association was observed particularly among women maintaining a high physical activity level before and after first pregnancy. The activity level (58 MET-hrs/wk) achieved by women in the top quartile was equivalent to 71 min/day of vigorous activity or 110 min/day of moderate activity, which is twice the World Cancer Research Fund/American Institute of Cancer Research recommendation of 30 min/day of vigorous or 60 min/day of moderate activity.³²

Prior studies assessing physical activity at specific age periods did not distinguish whether the activity occurred before or after first pregnancy.^{7–10,12,33–36} Given the particular susceptibility of undifferentiated nulliparous breast tissue to exposures and the association between the length of time between menarche and first birth with cancer risk, our analysis focused on physical activity between menarche and first pregnancy and its interaction with the length of this time period. In an earlier prospective analysis of the NHS-II data, Maruti et al.³⁰ reported that physical activity during ages 12–22 had a stronger association with risk of premenopausal breast cancer than activity in later life. A recent analysis of the same cohort with a longer accrual of follow-up showed a suggestive inverse association of physical activity between ages 14 and 22 years with risk of younger (less than 46.9 years) premenopausal breast cancer, but not with older pre-menopausal or postmenopausal breast cancer.⁷ In contrast to these two prior analyses including both nulliparous and parous participants, our analysis was restricted to parous participants, which may in part explain our different findings. Prior studies have observed a stronger relationship between lifetime physical activity and breast cancer risk for nulliparous compared to parous women,³⁷ although the opposite has also been seen.^12,30

If breast tissue has vulnerable windows where it is particularly susceptible to the effects of carcinogens, it seems plausible that women with prolonged time frames would be at a higher risk of breast cancer. Furthermore, if lifestyle factors such as physical activity during this vulnerable window are able to offset risk, it makes sense that the largest benefit would be seen among those with the longest windows. We previously demonstrated among parous NHS-II participants that the adverse relationship between alcohol consumption between menarche and first pregnancy and breast cancer risk varied by the duration of the interval between menarche and first pregnancy.¹⁵ Among women with a longer interval (≥10 years from age 15 to age at first pregnancy), each 10 g/day increase in alcohol intake was associated with a 21% greater breast cancer risk, independent of alcohol intake after first pregnancy, while no association was observed among women with a shorter interval. The current analysis provides additional evidence for the contribution of exposures between menarche and first pregnancy to breast cancer development. We found that the relationship between physical activity during this critical window and breast cancer risk differed by the length of the window, such that an inverse association between physical activity and breast cancer was limited to women with a longer interval (≥20 years). Therefore, lifestyle prior to first pregnancy potentially impacts breast cancer risk later in life, supporting the hypothesis of a critical window for breast tissue exposures with subsequent cancer risk.

Limitations of our analysis include that early-life physical activity was recalled in 1997 at a mean age of 42 years,²⁵ which is likely be reported with some error, given that the recall is many years later. This measurement error would likely underestimate our estimates between physical activity and breast cancer; thus, the relationships may be stronger than our analyses indicate. Leisure-time physical activity measures have been demonstrated to have moderate reliability and validity in a subgroup of NHS-II participants, although adult recall of adolescent activity has not been specifically validated.³⁸ Second, recall bias is another possible limitation given 347 of the breast cancer cases had already been diagnosed by 1997 when the early life physical activity questionnaire was administered; however, a sensitivity analysis restricting to cases occurring after 1997 gave similar findings, suggesting that recall bias is minimal. Third, given the observational design of our study, confounding by unmeasured or poorly measured covariates is possible; however, we accounted for numerous established breast cancer risk factors, many of which are validated, minimizing this concern. Fourth, statistical power may be limited by the number of events in our cohort, limiting our confidence in point estimates for some of the analyses.

Conclusion

This study supports the hypothesis that exposures during critical early life time periods elevate breast cancer risk and suggests that physical activity between menarche and first pregnancy may offset such risk among women with a longer menarche-to-first-pregnancy interval. This vulnerable window of breast development has lengthened in recent decades, with a modest decline of a few months in the age at menarche^39,40 and a simultaneous increase of 4.6 years in the average age at first birth.⁴¹ Although intervening to reduce the length of this interval may not be a feasible public health breast cancer prevention strategy, increasing physical activity in young women may benefit this group of women at higher risk.