Prospective Study on Physical Activity and Risk of In Situ Breast Cancer

Physical inactivity is one of the few established risk factors for breast cancer amenable to intervention. Approximately 73 studies have been conducted worldwide on breast cancer risk and physical activity, including about 33 cohort studies (1, 2). On average, a 25% reduction in breast cancer risk has been observed among physically active women in comparison with the least active women (3). An expert panel convened by the World Cancer Research Fund and the American Institute for Cancer Research classified the evidence for the inverse association of breast cancer with physical activity in postmenopausal women as probable (4). In premenopausal women the evidence was more limited, with a suggestion of a somewhat weaker association (4).

In situ breast cancer is considered as risk factor or precursor of subsequent invasive breast cancer (5). Thus, an association with physical activity and in situ breast cancer risk would suggest that this factor acts at a relatively early stage of the carcinogenic process. To date, only 3 cohort studies (6–8) and 4 case–control studies (9–12) have assessed the association of physical activity with risk of in situ breast cancer. Generally, these studies provided little evidence of an inverse association (6, 8–11). The studies showed wide variations in design, age range of participants, and the assessment of physical activity. Besides, only one of the cohort studies also investigated other types of physical activity in addition to recreational (6). In addition, the number of cases of in situ breast cancer varied from 450 (8) to 593 (7) for the cohort studies, and from 227 (11) to 1,689 (9) for the case–control studies, limiting the statistical power for some studies to detect associations of modest magnitude.

Given this limited evidence and the relevance of this topic for further understanding biologic mechanisms of carcinogenesis and breast cancer prevention, we examined the association between various types (occupational, household, recreational, and total) of physical activity and in situ breast cancer risk in the large European Prospective Investigation into Cancer and Nutrition (EPIC), in which a total of 1,059 in situ breast cancers were diagnosed during a median follow-up interval of 11.7 years.

EPIC is an ongoing multicenter prospective cohort established to investigate the associations between dietary, lifestyle, genetic, and environmental factors and risk of specific cancers and other chronic diseases. The design and baseline data collection methods have been previously described (13). Approval for this study was obtained from the Ethical Review Boards of the International Agency for Research on Cancer and from local institutions in the participating countries. There were 366,521 women, mostly of ages 35 to 70 years, enrolled between 1992 and 2000 in 23 subcohorts in 10 European countries (Denmark, France, Germany, Greece, Italy, Norway, Spain, Sweden, The Netherlands, and the United Kingdom). These participants were recruited from the general population from defined areas in each country with some exceptions: women who were members of a health insurance scheme for state school employees in France; women attending breast cancer screening in Utrecht, The Netherlands, and Florence, Italy; blood donors in some Italian and Spanish centers and a high number of vegans and vegetarians in the Oxford “Health conscious” cohort. Women provided written informed consent at the time they completed the standardized baseline questionnaires on diet, reproductive characteristics, sociodemographic variables, lifestyle, and medical history and were invited to a center to provide a blood sample.

The present study is based on data from 345,158 women after excluding a priori women with prevalent cancer at any site apart from nonmelanoma skin cancers at baseline examination (n = 19,853) or were missing a diagnosis or censoring date (n = 2,892). Women were excluded if it was unclear whether their breast cancer diagnosis was a primary incidence case (n = 4) and did not contribute to the underlying time at risk variable (n = 3). Three EPIC study centers, the full Norwegian subcohort (n = 35,890), and the Umeå center (Sweden, n = 12,513) were excluded because they had nonstandardized physical activity questionnaire data. Furthermore, we excluded women from all centers with missing nondietary questionnaire data (n = 526) or missing physical activity questionnaire data (n = 12,395). Thus, the final study sample consisted of 283,827 women from 9 countries.

Physical activity data were obtained either by face-to-face interviews (Germany, Greece, and Spain) or were self-administered (Denmark, France, Italy, The Netherlands, and the United Kingdom) using a standardized questionnaire in all centers included in this analysis. A description of the ascertainment of physical activity and its validity and reproducibility has been described in detail elsewhere (14–16). Data on current occupational activity included employment status and the level of physical activity done at work (nonworker, sedentary, standing, manual, heavy manual, and unknown). Housewives were categorized as nonworkers. In the Danish centers, the questionnaire focused on type of work activity done within the last year, and participants who did not answer this question were categorized as nonworkers.

For nonoccupational physical activities, the number of hours per week during the past year were collected in all centers and comprised household activities, including housework, home repair (do-it-yourself activities), gardening, and stair climbing, and recreational activities including walking, cycling, and sports combined as done in winter and summer separately. Because the intensity of recreational and household activities was not directly recorded, a metabolic equivalent (MET) value was assigned to each reported activity according to the Compendium of Physical Activities (17). A MET is defined as the ratio of work metabolic rate to a standard metabolic rate of 1.0 kcal (4.184 kJ)/kg/hour. One MET is considered a resting metabolic rate obtained during quiet sitting. The MET values assigned to nonoccupational data were: 3.0 for walking, 6.0 for cycling, 4.0 for gardening, 6.0 for sports, 4.5 for home repair (do-it-yourself work), 3.0 for housework, and 8.0 for stair climbing. These mean MET values were obtained by estimating the average of all comparable activities in the Compendium. The mean number of hours per week of summer and winter household and recreational activities were estimated and then multiplied by the appropriate MET values to obtain MET-hours/week of activity.

To obtain an overall estimate of household activity, housework (including child and older adult care), home repair, gardening, and stair climbing were combined. Walking (including walking to work, shopping, and leisure time), cycling (including cycling to work, shopping, and leisure time), and sports activities were combined to derive overall recreational physical activity. Household and recreational activities in MET-hours/week were combined and cohort participants classified according to EPIC-wide quartiles of nonoccupational physical activity.

In addition, 2 indices of total physical activity, combining occupational and nonoccupational activities, were considered in the analyses. Both indices have been developed previously and categorize individuals into inactive, moderately inactive, moderately active, and active groups. The “Cambridge index” combines occupational activity with hours spent doing cycling and sports (16, 18). The second index, the “total physical activity index”, cross-classifies combined household and recreational activity combined with occupational activity (15, 19).

Incident breast cancer cases were identified through a linkage with population-based cancer registries in most countries (Denmark, Italy, Spain, Sweden, The Netherlands, and the United Kingdom). For these centers, follow-up was completed as follows: December 2004 (Asturias), December 2006 (Florence, Varese, Ragusa, Naples, Granada, and San Sebastian), December 2007 (Murcia, Navarra, Oxford, Bilthoven, Aarhus, and Copenhagen), January 2008 (Utrecht), June 2008 (Cambridge), and December 2008 (Turin and Malmo). Active follow-up of study participants and next-of-kin, as well as of social security records and cancer and pathology registries was used in France, Germany, and Greece. For the centers using active follow-up, end of follow-up was the last known contact, date of diagnosis, or date of death, whichever occurred first. Mortality data were coded according to the 10th revision of the International Statistical Classification of Diseases, Injuries, and Causes of Death (ICD-10), and cancer incidence data were coded according to the International Classification of Diseases for Oncology (ICD-O-2). A total of 9,947 incident breast cancer cases had been reported, out of which 1,059 were in situ breast tumors. A total of 866 cases were defined as ductal and 78 cases as lobular carcinomoa in situ.

Cox proportional hazards models estimating HR and 95% confidence intervals (CI) were used to evaluate the associations between physical activity and the risk of in situ breast cancer. Age was defined as the primary-dependent time variable. Women were considered at risk from the time at recruitment until breast cancer diagnosis or censoring (age at death, loss to follow-up, end of follow-up, or diagnosis of other cancer entities, including invasive breast cancer). All models were stratified by age in 1-year categories and by study center, to control for potential variability between centers and to be less sensitive to potential violations of the proportional hazards assumption. The proportional hazards assumptions were verified by including interaction terms between exposure and time/age and comparing the interaction model with the model without the interaction terms by means of a likelihood ratio test. The assumptions were met, apart from a slight violation in the lowest category of occupational physical activity.

Continuous variables of physical activity (e.g., MET-hours/week of activity) were categorized into quartiles using cutoff points based on the overall cohort. Trend tests for originally continuous physical activity variables were based on mean values calculated within the quartiles. Trend tests for categorical variables were estimated on scores (1–4) applied to the categories of the physical activity variables and entered as a continuous term in the regression models. To evaluate HRs for overall significance, P values for Wald χ² were computed. We considered several types of models, crude models and adjusted models. The maximally adjusted multivariable model was stratified by age and center and adjusted for the following potential confounders as assessed at baseline: body mass index (BMI, <25, ≥25–<30, ≥30 kg/m²), age at menarche (≤11, 12–14, >14 years, missing), age at first birth (<20, 20–30, >30 years, nulliparous, missing), breastfeeding (yes, no, missing), oral contraceptive use (ever, never, missing), menopausal status (premenopausal, perimenopausal, postmenopausal), age at menopause (<43, 43–46, 47–49, 50–51, 52–53, ≥54, missing), number of full-term pregnancies (0, 1, 2, 3, 4, missing), use of hormone replacement therapy (ever, never, missing), smoking status (never, former, current, unknown), alcohol consumption as grams of ethanol per day (<1.5, 1.5–9, 10–19, 20–30, >30 g/d, missing) over the 12 months before recruitment, and education (none/primary school, technical/professional school, secondary school, university, not specified). Missing values (generally <2%) were accounted for by creating an extra category in each covariable. Where applicable, models were mutually adjusted for other types of physical activity (i.e., occupational, household, or recreational). All potential confounders were retained in the multivariable models because the exclusion of single or multiple factors did not result in more precise estimates for the associations with physical activity; thus, there was no advantage in using more parsimonious models (20).

To explore possible confounding, interaction, or an intermediate role of BMI, we fitted the multivariable models with and without adjustment for BMI as well as stratified by BMI subgroups (<25, ≥25–<30, ≥30 kg/m²). As information on menopausal status was only available for the time of recruitment but not for the time of diagnosis, age at diagnosis was used as proxy for the menopausal status at time of breast cancer diagnosis. Thus, we investigated the possibility of effect modification by stratifying on age at diagnosis (≤50, >50 years of age). For the subgroup analysis of women diagnosed before or equal to age of 50 years, we applied right censoring at age of 50 years for all women. Left censoring for all women was conducted for the subgroup analysis for breast cancer cases diagnosed at older ages.

In general, women who visit screening programs are at a higher probability of being diagnosed with in situ breast cancer and may also be more active. In EPIC, we did not have information on whether cancers were screening detected or not. As proxy-analyses, we investigated those centers that recruited through breast cancer screening programs (centers Florence and Utrecht, 129 cases) separately from the other centers.

All analyses were conducted using SAS Statistical Software, version 9, and all statistical tests were 2-sided.

The average duration of follow-up of the 283,827 women in the analytic cohort was 11.0 (median 11.7) years with a total of 3,117,431 person-years (Table 1). The mean age at recruitment into this cohort was 51.5 years (median: 51.9). A total of 1,059 in situ breast cancer cases were diagnosed. Of these cases, 138 were diagnosed before or at 50 years of age and 921 were older than 50 years at diagnosis.

For the total physical activity index, more physically active women had more children, were more likely to have breastfed, were less likely to use oral contraceptives and hormone replacement therapy, and had a lower alcohol intake in comparison with inactive women (Table 2). Smoking habits, age at menarche, age at menopause, and age at first birth were comparable. Furthermore, active women had a lower education level and a higher BMI than less active women. This result for the combined total physical activity index is driven primarily by household physical activity (data not shown). For recreational physical activity and the Cambridge index, which does not include household activity, active women had a lower BMI than less active women (data not shown).

In crude and adjusted multivariable models, risk of in situ breast cancer showed no association with any of the physical activity subtypes (occupational, household, or recreational physical activity; Table 3), nor was there any association with combined recreational and household activity or total physical activity (both indices). For the total physical activity index, active compared with inactive study participants had a multivariable adjusted HR of 1.07 (95% CI, 0.81–1.40). Further analyses by age at diagnosis showed similar results. For cancers diagnosed before or at the age of 50 years, the HR for the highest activity group was estimated to be 1.32 (95% CI, 0.69–2.52) for total physical activity. For cancers diagnosed later, the corresponding HR was 1.03 (95% CI, 0.77–1.39). For combined recreational and household physical activity as well as for the separate activity components, any physical activity above the reference level resulted in crude and adjusted risk estimates less than 1 for cancers with early diagnosis and more than 1 for breast cancer diagnosed at age 50 years or older. In general, all estimates were nonsignificant apart from the multivariable adjusted estimate for moderately inactive recreational physical activity (second quartile, HR = 0.58; 95% CI, 0.35–0.96) for breast cancers that were diagnosed before or at age of 50 years and moderate physical inactivity in the household (HR = 1.21; 95% CI, 1.02–1.45) for breast cancer diagnosed at later age. When restricting the group of women, ages less than 50 years, at diagnosis to those who were premenopausal at baseline recruitment similar results were obtained (data not shown). In general, analyses for cancers with early diagnosis had limited power. When considering only women that were postmenopausal at recruitment (including 515 cases) no associations between physical activities and in situ breast cancer risk were observed (data not shown).

Extensive sensitivity analyses, for example, by restricting the case definition to cases diagnosed at least 2 years after baseline (n = 912) and separate analyses for those centers that recruited through breast cancer screening programs from the other centers, were conducted (results not shown). Overall, no further information was gained from any of these analyses with some being restricted by power. We reported only models that were adjusted for BMI because models with and without adjustment for BMI yielded almost identical results. The same held true for height and weight adjustment instead of BMI adjustment (data not shown). When examining effect modifications by predefined categories of BMI, no apparent heterogeneity in the null association of physical activity with in situ breast cancer risk across categories was observed (Table 4). Most of the estimates for the obese participants, the smallest group (n = 92), were nonsignificantly more than 1, however, with no indications for trends.

The results of this large European prospective cohort study, including 1,059 incident in situ breast cancer cases, do not support the hypothesis of an association of in situ breast cancer risk with any subtype of physical activity, nor with total physical activity. Furthermore, no evidence for effect modification by age at diagnosis or BMI was found.

To our knowledge, this is the largest prospective study on risk of in situ breast cancer based on a comprehensive assessment of physical activity. Another strength of this study is the availability of data on a wide range of other risk factors for breast cancer. A major limitation of the study is the self-reported physical activity assessment 1 year before study entry that may not represent long-term activity. Also, while all types of activity were assessed in this study at the time of recruitment, there was no information on the duration and frequency of occupational activity. This precluded estimating a quantitative sum of all types of activity for total activity in MET-hours/week. However, several previous analyses of physical activity in EPIC have shown inverse associations with several cancers known to be associated with physical activity, (19, 21–23), including our own recent investigation of invasive breast cancer (23), supporting the validity of the exposure assessment. Another limitation is the lack of information on cancer detection through mammography screening for most centers.

Studies that have assessed the association of physical activity and risk of in situ breast cancer represent a wide range of methodologic approaches. They differed in their design, sample size, age range of participants, and in the assessment of physical activity, which may explain some of the heterogeneity in their results. The number of cases of in situ breast cancer varied from 227 (11) to 1,689 (9), raising the issue of sufficient power to detect associations in some of the studies.

Our results are consistent with 2 (6, 8) of 3 cohort studies (6–8) that have examined the association of physical activity and risk of in situ breast cancer. One study was based on 450 in situ cases (8) and assessed recreational physical activity only, defined as walking and exercise. For another cohort, results have first been published for 1,176 in situ cases of the breast based on a simplistic single-variable–based exposure assessment of the frequency of any vigorous physical activity (24). Subsequently, more detailed data on nonrecreational physical activity only were reported for a subcohort with 570 cases (6). Only in the first analysis, a weak association with nonsignificant risk estimates and trends was reported. In contrast to our results, 1 cohort study (7) conducted in the United States based on 593 cases found a significant reduction in risk of in situ breast cancer in the highest quintile of lifetime sports activity (HR = 0.69; 95% CI, 0.48–0.98) and with strenuous sports in the past 3 years (HR = 0.57; 95% CI, 0.33–0.99), but not with moderate sports activity (either lifetime or past 3 years). Other types of physical activity were not assessed in this study.

Also in agreement with our results, 3 (9–11) of 4 case–control studies (9–12) did not find an association between physical activity with risk of in situ breast cancer. Only in 1 case–control study of 567 mostly premenopausal women, an inverse association of recreational physical activity with in situ breast cancer was reported (OR 0.65; 95% CI, 0.48–0.90), but there was no evidence of a trend with increasing level of exercise (12). The other case–control studies [1 with premenopausal (11), 1 with mostly postmenopausal women (9), and the other with exclusively postmenopausal women (10)] found no association of lifetime recreational or occupational physical activity with in situ breast cancer (9, 11) or of leisure time or total physical activity with in situ breast cancer (10), respectively.

Some additional studies reported results for invasive and in situ cancers combined and for invasive cancers only. They reported similar (inverse) associations with both types of analyses (25–27). These results are not necessarily inconsistent with our result of no association between in situ breast cancers and physical activity. Given the small percentage of in situ carcinoma in most of the studies, their impact on overall results is expected to be small. Thus, an association found for invasive and in situ cancers combined does not necessarily indicate an association also for in situ breast cancers alone.

We found no indications for effect modifications by BMI. Only 2 previous studies investigated this issue (7, 8) and briefly reported results without presenting data. In agreement with our results, both of them stated that no effect modifications by BMI were seen. We did not find differential associations between physical activity and in situ breast cancer risk by age at diagnosis (as surrogate for menopausal status at diagnosis). Most of the previous cohort and case–control studies have only included postmenopausal (6, 8, 10) or predominantly postmenopausal populations (9). Only 1 study was conducted in a population (7) of both pre- and postmenopausal women, but data are not reported according to menopausal status. In general, our study and previous studies do not support differential effects by menopausal status.

If these null results for physical activity and in situ breast cancer risk did not occur by chance, they may raise several suggestions with regards to further understanding breast cancer etiology. Considering the contrast between our results for carcinoma in situ and the recognized inverse association for invasive breast cancer, one possible explanation could be that in situ and invasive breast cancers are in fact separate diseases with different etiologies. Carcinomas in situ of the breast are a heterogeneous group of tumors with and without invasive characteristics (28). Information on direct comparisons of risk factors for in situ and invasive breast cancer is sparse, but remarkable similarities were seen for most risk factors (5). Physical activity has not been among the considered risk factors in any of the studies. Interestingly, the only factor to show a clear difference in its association with in situ versus invasive breast cancer risk was also energy-balance–related factor BMI, with a trend for increasing risk with increasing BMI for invasive but not for in situ cancer (5, 29). Another hypothetical explanation for our null finding is that any beneficial effect of physical activity may operate at a relatively late stage in the development of the disease. The factors that lead to in situ tumors to progress into invasive breast cancer remains unclear and poorly defined (28, 30). Furthermore, the exact biologic mechanisms for the association of physical activity with breast cancer are in general still speculative. Several currently hypothesized models focus on the promotion and progression of initiated cells, such as the ones that implicate sex hormones, insulin resistance, adipokines, and chronic inflammation, as possible mediators of physical inactivity (1, 3, 31, 32). Specifically for steroid hormonal pathways, epidemiologic studies have found direct associations between postmenopausal endogenous levels of steroid hormones and breast cancer (33). Postmenopausal women who are physically active in comparison with inactive women may have lower concentrations of serum estrone, estradiol, and androgens (34, 35) and higher sex hormone–binding globulin (36). Furthermore, substantial epidemiologic, clinical, and experimental evidence has clearly established a late-stage growth promoting effect of estrogens (37), which would be in line with our findings.

In conclusion, this large prospective study conducted in a heterogeneous population of European women provides no evidence for an association between physical activity and risk of in situ breast cancer. Given the recognized inverse association of physical activity and invasive breast cancer, these null results for in situ breast cancer, probably, suggest that, unlike other risk factors, any beneficial effects of physical activity may operate later in the carcinogenic process.

No potential conflicts of interest were disclosed.

Conception and design: K. Steindorf, R. Ritte, A. Tjonneland, K. Overvad, J. Chang-Claude, R. Tumino, C. Navarro, E. Ardanaz, H.B. Bueno-de-Mesquita, K.-T. Khaw, T.J. Key, R. Kaaks

Development of methodology: K. Steindorf, R. Ritte, R. Tumino, H.B. Bueno-de-Mesquita

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Ritte, A. Tjonneland, N.F. Johnsen, K. Overvad, F. Clavel-Chapelon, A. Lukanova, J. Chang-Claude, H. Boeing, A. Trichopoulou, D. Trichopoulos, G. Masala, V. Krogh, A. Mattiello, R. Tumino, S. Polidoro, J.R. Quirós, M.-J. Sánchez, C. Navarro, E. Ardanaz, P. Amiano, H.B. Bueno-de-Mesquita, E. Monninkhof, K.-T. Khaw, N. Wareham, T.J. Key, R.C. Travis, R. Kaaks

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Steindorf, R. Ritte, C. Navarro, E. Ardanaz, F.J.B. van Duijnhoven, A.M. May, V. Fedirko, I. Romieu, P.A. Wark, R. Kaaks

Writing, review, and/or revision of the manuscript: K. Steindorf, R. Ritte, A. Tjonneland, N.F. Johnsen, K. Overvad, J.N. Østergaard, F. Clavel-Chapelon, A. Fournier, L. Dossus, A. Lukanova, J. Chang-Claude, H. Boeing, A. Wientzek, A. Trichopoulou, T. Karapetyan, D. Trichopoulos, G. Masala, V. Krogh, A. Mattiello, S. Polidoro, J.R. Quirós, N. Travier, M.-J. Sánchez, C. Navarro, E. Ardanaz, P. Amiano, H.B. Bueno-de-Mesquita, F.J.B. van Duijnhoven, E. Monninkhof, A.M. May, K.-T. Khaw, N. Wareham, T.J. Key, R.C. Travis, K.B. Borch, V. Fedirko, S. Rinaldi, I. Romieu, P.A. Wark, T. Norat, E. Riboli, R. Kaaks

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Ritte, K. Overvad, H. Boeing, A. Trichopoulou, R. Tumino, J.R. Quirós, E. Ardanaz, H.B. Bueno-de-Mesquita, K.-T. Khaw, R. Kaaks

Study supervision: K. Steindorf, R. Ritte, R. Tumino, M.-J. Sánchez, H.B. Bueno-de-Mesquita, T.J. Key

The authors thank Jutta Schmitt and Jutta Kneisel for their assistance during data collection, and all the EPIC cohort participants for their contributions to data collection at baseline recruitment and during follow-up.

The coordination of EPIC is financially supported by the European Commission (DG-SANCO) and the International Agency for Research on Cancer. The national cohorts are supported by Danish Cancer Society (Denmark); Ligue contre le Cancer, Mutuelle Générale de l'Education Nationale, Institut National de la Santé et de la Recherche Médicale (France); Deutsche Krebshilfe, Deutsches Krebsforschungszentrum and Federal Ministry of Education and Research (Germany); the Hellenic Health Foundation, and the Stavros Niarchos Foundation (Greece); Italian Association for Research on Cancer (AIRC) and National Research Council (Italy); Dutch Ministry of Public Health, Welfare and Sports (VWS), Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF), Statistics Netherlands (The Netherlands); ERC-2009-AdG 232997 and Nordforsk, Nordic Center of Excellence program on Food, Nutrition and Health. (Norway); Health Research Fund (FIS), Regional Governments of Andalucía, Asturias, Basque Country, Murcia (no. 6236) and Navarra, ISCIII RTICC “Red Temática de Investigación Cooperativa en Cáncer” (R06/0020; Spain); Swedish Cancer Society, Swedish Scientific Council and Regional Government of Skåne and Västerbotten (Sweden); and Cancer Research UK, Medical Research Council (United Kingdom).

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.