Prospective study of physical activity in different periods of life and the risk of ovarian cancer
Version of Record online: 19 JAN 2006
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 12, pages 3153–3160, 15 June 2006
How to Cite
Weiderpass, E., Margolis, K. L., Sandin, S., Braaten, T., Kumle, M., Adami, H.-O. and Lund, E. (2006), Prospective study of physical activity in different periods of life and the risk of ovarian cancer. Int. J. Cancer, 118: 3153–3160. doi: 10.1002/ijc.21800
- Issue online: 27 MAR 2006
- Version of Record online: 19 JAN 2006
- Manuscript Accepted: 28 NOV 2005
- Manuscript Received: 22 MAR 2005
- National Cancer Institute, U.S. Grant Number: CA 52449
- Norwegian Cancer Society. Grant Number: DNK 90050
- Aakre Foundation
- Swedish Council for Planning and Co-ordination of Research
- Swedish Cancer Society, STINT (The Swedish Foundation for International Cooperation in Research and Higher Education), Organon, Pharmacia, Medical Products Agency and Schering-Plough
- physical activity;
- ovarian cancer cohort study;
Only few studies have assessed the role of physical activity in the etiology of ovarian cancer, and the results have been inconclusive. We studied associations between physical activity and risk of ovarian cancer in 96,541 women aged 30–49 at enrollment in a prospective study in Norway and Sweden. Participants reported physical activity level at ages 14, 30 and at enrollment, and participation in competitive sports. Complete follow-up through 2001/2002 was achieved by linkage to national registries. The relation between physical activity and ovarian cancer incidence was assessed using multivariate Cox proportional hazard models. During an average 11.1 years of follow-up, there were 264 ovarian cancer cases (including 81 borderline tumors) diagnosed at a mean age of 49 years. Highly physically active women at cohort enrollment had a similar risk of ovarian cancer as women reporting no activity (multivariate relative risk RR = 1.08, 95% CI 0.53–2.18). Physical activity at age 30 or at age 14 did not either afford any protection from ovarian cancer, nor did a consistently high level of activity from younger ages until enrollment. Results were similar for invasive and borderline tumors, and for different subgroups of women classified according to other known risk factors for ovarian cancer. In our study of primarily premenopausal women, physical activity at different ages did neither reduce nor increase risk of ovarian cancer. In the context of the inconsistent scientific literature, our findings probably reflect that physical activity is not causally related with ovarian cancer. © 2006 Wiley-Liss, Inc.
Ovarian cancer is a common neoplasm, ranking seventh for incidence and sixth for mortality in Western countries.1 Because ovarian cancer is often diagnosed in late stages when cure is no longer possible, it is the leading cause of mortality among gynecological malignancies.2 No screening method has yet been proven effective; therefore, identifying modifiable risk factors is one strategy to reduce morbidity from this high lethal malignancy.
Ovarian cancer is at least partially a hormone-related disease, occurring less frequently in women with longer periods of anovulation due to increasing number of pregnancies or prolonged use of hormonal contraceptives.3 Increasing circulating levels of estrogens in both pre- and postmenopausal women have been suggested to increase ovarian cancer risk, although the literature is sparse. Moderate levels of physical activity contribute to weight control, and appear to improve immune function. It also possibly decreases urinary estriol and progesterone levels, and increases catechol-o-methyltransferase (COMT) activity, which is associated with estrogen metabolism. Vigorous physical activity in premenopausal women can lead to anovulation, luteal-phase insufficienty and amenorrhoea or irregular menstrual cycles, and lower endogenous estrogen levels, urinary LH levels, and impair immune function.4, 5 The hormonal and immunological effects of physical activity on postmenopausal women are less known.
To date, 3 prospective studies have examined the association between leisure time physical activity and the risk of ovarian cancer; one cohort study showed a modest increase in ovarian cancer risk with higher levels of activity,6, 7 and 2 others reported no clear association.8, 9 Two other cohort studies using occupation to classify physical activity found conflicting results: that sedentary jobs with low levels of energy expenditure were associated with increased risk10 or was not associated with ovarian cancer.11 As summarized in Table I, the results of case-control studies have been inconclusive as well.12, 13, 14, 15, 16, 17, 18, 19
|Author, date, study and location||Study dates||Number of cases (age range)||Physical activity definition (Number of categories analysed in each study)||Relative risk (Highest versus lowest categories) (95% CI)||Adjustment for confounding||Comments|
|Pukkala et al. (1993), Finland||1958–92||51 (Ages not stated)||Current occupational title||SIR (vs. general population) for physical education teachers: 1.7 (0.8–3.2)||Age||Based on national incidence figures. Physical education teachers were supposed to be more physically active than language teachers.|
|Language teachers: 1.6 (1.1–2.1)|
|Zheng et al. (1993), China||1980–84||595 (Aged 30 or more years)||Current occupational activity (3 categories: low, moderate and high)||SIR = 102 (CI not given) for high energy expenditure (and short sitting time)||Age||Based on occupational data on census|
|Mink et al. (1996), USA||1985–92||97 (Aged 55–69 years in 1985)||Recreational activity: moderate (4 levels according to frequence per week)||1.6 (0.9–2.9)||Age, smoking, education, parity, family history, waist/hip ratio||Postmenopausal women only. Significant tests for trend|
|Vigorous (4 as earlier)||2.5 (1.0–6.3)|
|Overall index (3 levels)||2.1 (1.2–3.5)|
|Bertone et al. (2001), USA||1980–1996||377 (30–55 years in 1976)||Recreational physical activity frequency||7 hours per week compared to less than 1 hour: 0.80 (0.49–1.32)||Age, parity, age at menarche, oral contraceptive use and duration of use, menopausal status and use/duration of use of HRT, tubal ligation, smoking||No evidence of any trends|
|Vigorous vs. low intensity in 1980||1.48 (0.89–2.48)|
|Anderson et al. (2004), USA (same study as Mink et al.)||1986–2000||233 (55–69 years in 1985)||Leisure time (3 levels)||High vs. low: 1.42 (1.03–1.97)||Age, family history ovarian cancer, hysterectomy, parity, smoking, HRT|
|Vigorous (4 levels of frequence per week)||Vigorous 4 times per week. rarely/never 2.38 (1.29–4.38)|
|Hannan et al. (2004), USA||1979–1998||121 (Age not stated)||Leisure time||Quintile 5 vs. Quintile 1||Age, menopausal status, duration of oral contraceptives and HRT use, parity||No evidence of trend|
|Met H/day (5 levels)||0.71 (0.41–1.22)|
|2–14 hr/day vigorous activity vs. none||0.74 (0.42–1.30)|
|Case-control studies||Odds Ratio (95% CI)|
|Cottreau et al. (2000), USA||1994–98||767 (Ages 20–69 years)||Recreational index at different ages combined (3)||Lifetime 0.73 (0.56–0.94)||Age, parity, estrogen use, family history, BMI, education, race and tubal ligation||Population-based significant test for trend|
|Decreased risk with increasing activity at age1 14–17, 22–29 and 30–39 years|
|Tavani et al. (2001), Italy||1992–99||1031 (Ages 18–79)||Occupational (3)||Age, education, study center, BMI, reproductive factors, family history, energy intake and estrogen use||Hospital-based significant test for trend for occupational activity (but not for recreational activity)|
|15–19 years||0.89 (0.65–1.20)|
|30–39 years||0.67 (0.47–0.98)|
|50–59 years||0.76 (0.48–1.21)|
|5–19 years||1.0 (0.80–1.26)|
|30–39 years||0.86 (0.65–1.12)|
|50–59 years||0.83 (0.58–1.18)|
|Zhang et al. (2003 and 2004), China||1999–2000||254 (Aged under 75 years)||Daily sitting duration (occupational and leisure) (3)||More than 10 hours vs. less than 4 hours: 1.77 (1.0–3.1)||Age, BMI, family history, hormonal status, reproductive history, total energy intake, physical activity (MET-hour)||Controls recruited from hospital visitors, outpatients, and community|
|Total weekly activity MET-hr (3)||0.54 (0.34–0.87)|
|Strenuous sports (3)||0.33 (0.14–0.75)|
|Vigorous work hours/week (3)||0.73 (0.41–1.31)|
|Moderate activity hours/week (3)||0.42 (0.22–0.81)|
|Activity inducing sweating (times/week) (3)||0.56 (0.7–0.84)|
|Freedman et al. (2002), USA||1984–1995||39,002 (Age not stated)||Occupational physical activity (4)||High vs. sedentary 0.82 (0.75–0.91)||Age, sex, race, residence, SES status and occupation (indoors, outdoors, mixed and farmer)||Death certificate based study|
|Bertone et al. (2002), USA||1991–1994||327 (40–79 years)||Hours/week||Age, parity, tubal ligation, use ovarian stimulating hormones, pelvic exams and family history ovarian cancer||Community controls selected from Medicare and driver's license lists.|
|overall activity (4)||0.95 (0.42–2.15)|
|Age 20||0.98 (0.48–2.9)|
|Age 12||1.10 (0.65–1.88)|
|Vigorous activity (4)||0.85 (0.39–1.86)||Analysis of MET-hours/week in different ages did not reveal any patterns|
|Age 20||0.86 (0.41–1.78)|
|Age 12||0.88 (0.60–1.27)|
|Riman et al. (2004), Sweden||1993–1995||655 (50–74 years)||Leisure time physical activity||Age, parity BMI, age at menopause (categorical), duration of oral contraceptive use, ever use of HRT, age categories of leisure time physical activity|
|Last years prior to enrollment|
|Pan et al. (2005), Canada||1994–1997||442 (20–76 years)||Recreational activity||10-year age group, providence of residence, education, alchol consumption, cigarette pack-years, BMI, total calorie intake, total vegetable consumption, number of live births and menopause status||No differences between pre- and postmenopausal women|
We studied the association between physical activity and ovarian cancer incidence over an average of 11 years of follow-up in the large, population-based prospective Women's Lifestyle and Health Study. From the biological effects described earlier, we hypothesized that moderate to high levels of physical activity would protect against ovarian cancer.
Subjects and methods
The present investigation is based on data from the Women's Lifestyle and Health Study, a prospective cohort study in Norway and Sweden, described in detail previously.20, 21, 22 The study population includes 57,582 Norwegian women from 34–49 years of age at recruitment and 49,259 Swedish women from 30–49 years of age, who were monitored with regard to vital status, cancer incidence and emigration since 1991–92. In Norway, participants were recruited from the entire female population during 1990–91. The source population in Sweden consisted of all women residents in the Uppsala Health Care Region in 1990–91. Women were randomly selected from the population registers. A total of 196,000 women were invited to participate, and 106,841 (54.5%) returned the questionnaire.
From the initial cohort, we excluded 1,669 women with a history of cancer other than nonmelanoma skin cancer at enrollment. We further excluded 8 women because of lack of information on vital status, 17 women who had died or emigrated before the start of the follow-up, 476 women with bilateral oophorectomy before cohort enrollment and 8,130 women who were missing information on current physical activity at enrollment. Thus, the final analysis was based on data from 96,541 (50,644 Norwegian and 45,897 Swedish) women. The responsible data monitoring boards and ethical committees in both countries approved the study design, and all women gave an informed consent prior to participating in the study.
Exposure and covariate data in the cohort are based on self-reported information collected in a questionnaire administered at cohort enrollment. The questionnaires sent to Norwegian and Swedish women included a core set of similar or identical questions, including data on physical activity, demographic characteristics, smoking habits, reproductive history, history of cancer in the mother or sister, history of oral contraceptive use, history of menopausal hormone therapy, height and weight. Information on tubal sterilization, estimated to be less than 5% in the cohort according to National figures, was not available for the combined database.
The women rated their level of physical activity at 3 time points: age 14, age 30, and at enrollment. Norwegian women ranked activity on a scale of 1 (very low) to 10 (very high). On the Swedish questionnaire, women ranked their level of physical activity on a 5 point scale: 1 (very low) was described as mainly sitting, 3 (normal) characterized by several long walks per week and 5 (very high) as sports or jogging several times per week. We adjusted the 10-level Norwegian scale into 5 levels to be comparable to the Swedish scale, grouping together categories 1 and 2, 3 and 4, etc. The 5 categories were labeled none, low, medium, high and vigorous level of physical activity. Women were also asked whether they had participated in competitive sports, and if so, the number of years they participated.
Combining the self-reported data at age 14, age 30, and at enrollment, we further categorized individuals on changes in physical activity over time. First, we separated the levels of physical activity into women who participated in none or low physical activity (inactive), and those who participated in moderate, high or vigorous activity (active) for each time point. We then compared physical activity levels between age 14 and 30, age 14 and enrollment, and between age 30 and enrollment. From this comparison, women could be categorized as those who remained inactive, women who were active and became inactive, women who were inactive and became active and those who remained active over time.
Using the 5 point scale of physical activity in each age (14, 30 and cohort enrolment), we created variables for the change in physical activity level between each age intervals (age 14 to age 30, age 30 to age at cohort enrolment and age 14 to age at cohort enrolment). Thus, a woman could be classified as changing positively one score (+1) in physical activity if she changed from inactive at age 14 to low activity level in age 30, or changing negatively one score (−1) if her physical activity decreased from very active at age 30 to moderately active at cohort enrolment. Respectively, a change from score 1 to 3, from 2 to 4 and from 3 to 5 were all considered as a change of +2. The resulting variable ranges from −4 (decreases in physical activity levels) to +4 (increases in physical activity level), with zero, indicating no change, and was treated as continuous variables in the statistical analysis.
Women who reported a natural menopause at enrollment were considered postmenopausal during the follow-up, whereas women who reported a bilateral oophorectomy were excluded. All other women were considered premenopausal, regardless of age, hysterectomy or use of hormone therapy.
Follow-up of the cohort was achieved through linkages with existing nationwide health registries. Because each resident in Norway and Sweden is assigned a unique national identification number, one can link the data from the cohort with these registers for virtually complete follow-up with respect to death and emigration. From the total population registries, we received information on dates of death for women who died during the follow-up period, and dates of emigration for women who moved out of their respective countries. The population registries were updated through January 2003 for Norway and through June 2003 for Sweden. The national cancer registries, established in the 1950's in both countries, provided data on prevalent cancer cases at cohort enrollment and on cancers diagnosed in the cohort during follow-up. During the period studied, these registries are also estimated to be close to 100% complete.23, 24, 25 The outcome of interest for this analysis was the time needed in the development of ovarian cancer. The start of follow-up was defined as the date of return of the questionnaire during 1991–1992. Observation time was calculated from date of entry into the cohort until the occurrence of ovarian cancer, or censoring on account of other cancer incidence, emigration, death, or end of the observation period (December 31, 2001 for Norway and 31 December 2002 for Sweden), which ever occurred first. The difference in termination dates of follow-up was determined by the availability from the cancer registries of complete national cancer incidence data in Sweden and Norway in September 2004.
The relation between physical activity and ovarian cancer was evaluated by fitting Cox proportional hazard models (R software v2.1.1 under Win XP and SuSE Linux 9.2). The following covariates, considered as possible confounder of the association between physical activity and ovarian cancer, were included in the Cox regression models: BMI (0–18.5, 18.5–25, 25–30 and ≥30), age (years)*, body height (cm)*, education (High/Low for ≥12 years and <12 years in school), alcohol intake (0, 0–1 and >1 drink/day where each drink correspond to 10 g alcohol), smoking status (never, former, current), duration of smoking (5 year periods up to maximum 20 years of smoking), parity (0, 1, 2, 3 and ≥4 children), breast feeding (month)*, age at 1st child (years)*, menopausal status (pre- or post-menopausal), age at start of menstruation (years)*, oral contraceptive use (never, former, current), duration of oral contraceptives (years)* and use of hormone replacement therapy (Yes/No). All variables were measured at enrollment. Variables marked * were treated as continuous variables in the models, whereas the other variables were treated as categorical. All models were fitted allowing different baseline hazards for each of the 4 BMI categories. We calculated hazard ratios and associated two-sided 95% Wald confidence intervals. The assumption of proportional hazards was evaluated by graphical evaluation of Schonfeldt residuals. Besides the analyzes of the association between physical activity at enrollment, at the age 30 and at the age 14 for the entire data sample, we also analyzed subsets of borderline and invasive tumors separately and a model, including competitive physical activity instead of physical activity at enrollment.
To assess possible effect modification, we first estimated hazard ratios comparing active versus inactive women, and separate models by invasive cancer vs. borderline tumor, age at enrollment (less than 40, 40 or more years) attained age or age at follow-up (less than 50 and 50 or more years), smoking status (Never, former-current), adult body mass index (less than 25, or 25 and more), oral contraceptive use (never users and former-current users) and education level (low and high). We then assessed effect modification on a multiplicative scale by modeling interaction terms and calculating log likelihood ratio tests to assess whether interaction terms were significantly different than 0.
Of the initial sample of 96,541 women, a maximum of 8,545 records were excluded from the analyses because of missing values for at least one of the covariates of interest. For the model, including competitive physical activity, 4,339 more records were excluded.
In our cohort of 96,541 women, altogether 264 ovarian cancer cases (164 in Norway and 100 in Sweden) were diagnosed during an average of 11.1 years of follow-up (996,416 person-years); 183 invasive and 81 borderline ovarian tumors. The mean age at diagnosis was 49 years. Since we found no evidence of a differential effect of physical activity on borderline and invasive ovarian cancers (data not shown), we present results for all ovarian tumors combined.
The mean age at enrollment into the cohort was 40.5 years, and almost all the women (n = 93, 090 or 96%) were premenopausal at cohort enrolment. Over half of the women had at least 11 years of education. About 37% of the women were never smokers, and current smokers represented 27% of the cohort at enrollment. The mean age at first birth was 24.4 years, and 12% of the women were nulliparous. Oral contraceptives had been used at some point in time by 66% of the women, and the mean duration of use was 5.7 years. Norwegian and Swedish women were similar in most of the characteristics, except for higher prevalence of smoking and alcohol consumption among Norwegian women and more use of oral contraceptives in Swedish women. Table II shows that about 95% of women were physically active at enrollment, 80% reporting at least moderate levels of activity, with even higher proportions at younger ages. About 16% of women reported participating in competitive sports, and nearly 10% had competed for 5 or more years. Vigorous activity was more common at age 14 than later in life, and none or low levels of physical activity increased with age. About half of the women maintained at least a moderate level or physical activity from age 14 to cohort enrolment. More than 70% of women reported sustained moderate to vigorous activity from age 14 to later time points (Table II).
|Physical activity level (%)|
|At age 14||2.9||9.6||45.4||24.5||17.6|
|At age 30||2.5||11.6||55.0||22.4||8.4|
|Changes in physical activity level (%)|
|Inactive-nochange||Active toInactive||Inactive toActive||Active-nochange|
|At age 14 to age 30||4.3||9.9||8.2||77.6|
|At age 14 to enrollment||3.7||17.0||8.8||70.5|
|At age 30 to enrollment||9.2||11.4||4.9||74.4|
Demographic characteristics and ovarian cancer risk factors by physical activity level at enrollment are shown in Table III. Compared to women with none or low physical activity levels, more active women were slightly younger, more educated, had lower body mass index (BMI), were taller, were less likely to smoke, consumed less alcohol, were more likely to use oral contraceptives and breastfed longer. There were no clear trends across physical activity levels for parity, age at menarche or age at first birth.
|N (%)||Physical activity level at enrollment|
|None N = 4,730(4.9%)||Low N = 15,323(15.8%)||Moderate N = 48,694(50.2%)||High N = 20,692(21.3%)||Vigorous N = 7,569(7.8%)|
|Mean age at enrollment, years||40.9||40.8||40.4||40.5||40.1|
|Education for 11 or more years (%)||44.5||56.2||53.2||58.1||52.7|
|Mean body mass index (kg/m2)||24.8||24.0||23.3||22.5||22.2|
|Mean height (cm)||165.9||166.4||166.2||166.6||166.3|
|Current smoker (%)||37.3||29.4||27.6||24.4||23.7|
|Consume >1 alcoholic drinks per day (%)||6.6||6.8||4.8||5.4||4.8|
|Age at menarche (years)||13.1||13.2||13.1||13.1||13.2|
|Mean age at first birth1 (years)||23.8||24.7||24.4||24.0||24.0|
|Mean breast feeding1 (months)||10.6||12.0||12.2||13.0||12.5|
|Current users of oral contraceptives at enrollment (%)||6.6||6.4||8.8||7.3||8.9|
|Postmenopausal at enrollment (%)||5.2||3.8||3.3||3.6||3.4|
|Participated in competitive sports for 4 or more years||9.8||8.0||8.0||11.0||17.5|
Compared to moderately active women, women with higher levels of physical activity at enrollment had a similar risk of ovarian cancer (Table IV). Extensive adjustment for potential confounders changed these risk estimates minimally. There was no evidence of a trend over categories of increasing levels of physical activity. Contrasting the highest (vigorous) to the lowest category of physical activity (none) at enrollment did not reveal any association (multivariate RR 1.08, 95% CI 0.53–2.18). There was no evidence of association between physical activity levels at ages 14 and 30 and ovarian cancer risk (Table IV). Compared to women who never participated in competitive sports, competing for up to 4 years (RR 1.07, 95% CI 0.63–1.82) or 5 or more years (RR 0.96, 95% CI 0.60–1.54) did not affect ovarian cancer risk (data not shown in the tables).
|No. of ovarian cancer cases1(Total 264)||Age-adjusted RR (95% CI)||Multivariable2 RR (95% CI)|
|Physical activity at enrollment|
|None||15||1.09 (0.64–1.86)||0.96 (0.54–1.71)|
|Low||43||0.97 (0.69–1.37)||0.91 (0.64–1.32)|
|Moderate||137||1.0 (ref)||1.0 (ref)|
|High||49||0.84 (0.61–1.17)||0.77 (0.54–1.09)|
|Vigorous||20||0.96 (0.60–1.54)||1.03 (0.64–1.66)|
|Physical activity at age 30|
|None||5||0.90 (0.37–2.20)||0.76 (0.28–2.07)|
|Low||31||1.14 (0.77–1.69)||1.05 (0.69–1.58)|
|Moderate||137||1.0 (ref)||1.0 (ref)|
|High||66||1.17 (0.87–1.57)||1.10 (0.80–1.50)|
|Physical activity at age 14|
|None||5||0.65 (0.26–1.58)||0.57 (0.21–1.54)|
|Low||31||1.22 (0.82–1.81)||1.16 (0.76–1.75)|
|Moderate||120||1.0 (ref)||1.0 (ref)|
|High||62||0.98 (0.72–1.33)||0.92 (0.67–1.28)|
|Vigorous||43||1.00 (0.71–1.42)||0.99 (0.68–1.42)|
Change in physical activity levels from active to inactive or vice-versa between different ages (14 to 30, 14 to enrollment, 30 to enrollment) did not afford protection or increase in risk from ovarian cancer (Table V). Contrasting the highest levels of physical activity with the lowest levels, or any of the other levels of physical activity with each other both at ages 14 and 30 did not reveal any patterns. Finally, the changes in physical activity levels at any ages using a continuous scale did not reveal any significant association (data not shown).
|No. of ovarian ca cases2||Person-years||Age-adjusted RR||Multivariable3 RR (95% CI)|
|Age 14 to Age 30|
|Inactive-no change||13||45,681||1.13 (0.64–1.69)||1.01 (0.55–1.86)|
|Active to inactive||23||104,279||1.04 (0.67–1.601.82)||0.98 (0.62–1.54)|
|Inactive to active||23||87,581||1.10 (0.71–1.69)||1.08 (0.69–1.71)|
|Age 14 to enrollment|
|Inactive-no change||13||39,556||1.29 (0.74–2.27)||1.15 (0.62–2.12)|
|Active to inactive||45||179,545||1.1 (0.73–1.40)||0.95 (0.67–1.35)|
|Inactive to active||23||94,229||1.01 (0.65–1.56)||0.99 (0.63–1.56)|
|Age 30 to enrollment|
|Inactive-no change||26||98,317||1.20 (0.80–1.81)||1.08 (0.70–1.68)|
|Active to inactive||32||120,967||0.96 (0.66–1.39)||0.91 (0.61–1.35)|
|Inactive to active||10||52,339||0.78 (0.41–1.47)||0.74 (0.38–1.45)|
We found no evidence that physical activity at enrollment provided protection from ovarian cancer in subgroups defined by type of tumor (invasive or borderline), age at enrollment, age at follow-up, smoking, adult body mass index, education and use of contraceptives – and all tests for interaction were nonsignificant (data not shown). Since the vast majority of cohort members were premenopausal, our statistical power for studying the effect of physical activity on ovarian cancer risk in the post-menopause period was very limited. We performed nevertheless and interaction test for physical activity and risk of ovarian cancer according to menopausal status, and the results were nonsignificant. The analysis restricted to women who were premenopausal at cohort enrollment were almost identical similar to the overall results.
In our cohort of primarily premenopausal women, we found no evidence of an association between level of physical activity at cohort enrollment, at age 14 or at age 30 and the risk of developing ovarian cancer. Changes in physical activity levels over lifetime, and participation in competitive sports were not associated with ovarian cancer risk either. There was no evidence that physical activity would affect risk differently in borderline or invasive tumors, nor in different subgroups of women classified according to age at cohort enrollment, menopausal status, age at follow-up, smoking history, body mass index, use of oral contraceptives and education.
Most women in our study (over 70%) were at least moderately physically active though life, but there was a tendency of becoming inactive with increasing age. Compared to inactive women, those who were physically active in our study were younger and displayed characteristics of higher socio-economic status, such as being taller, slimmer, smoking and drinking less, having more education and breastfeeding for longer periods. These associations have been found in other studies and support the validity of our physical activity measure, despite its simplicity.
Our follow-up was relatively long (11 years, 996,416 person-years) and virtually complete because of the existing population-based high quality registries, including cancer registries in both countries. In our study, physical activity levels at different ages were assessed at cohort enrolment in 1991–1992 only. Thus, changes in activity level were not captured and exposure misclassification, which per se may attenuate a possible association between physical activity and ovarian cancer risk, cannot be ruled out. Our physical activity questionnaire has not been validated: in fact what we accessed was the women's perception of their own physical activity. Reassuringly, our crude measurement of physical activity has proven to be an excellent predictor of overall mortality: risk of death decreased over 5 categories of physical activity at enrollment (p for trend <0.0001) and was reduced by half in the highest compared with the lowest category,26 as one would expect. In addition, we did not find any association between physical activity and premenopausal breast cancer in our cohort.27 These reassuring results on other physical activity-related health outcomes make us believe that misclassification of the physical activity variable in our study has probably not biased our results towards the null. Information on tubal sterilization was not available in our database. On the basis of recent data among control women in nationwide Swedish studies, we estimate that less than 5% of the women in the cohort would have had tubal sterilization. Thus, we do not believe that the physical activity patters of women who undergo tubal sterilization would be so dramatically different from that of other women in the cohort women that it could bias our results in anyway. We excluded from our analysis a relatively large number of women with missing information on covariates. We have redone some of the main analyses attributing a categorical value to the missing information for key covariates, and, reassuringly, our results remained similar, indicating no association between physical activity levels and ovarian cancer risk. Our analysis in subgroups of women classified according to use of contraceptives did not review any protection of physical activity among never users of the pill, as one could expect. In fact, we found no evidence of effect of physical activity in any other subgroups (according to BMI, smoking, age at enrolment, attained age and menopausal status).
Our study contrasts with previously published studies due to the high proportion of women who were at least moderately active in all periods of life and were leaner than most of other European and North–American populations.5, 28
All previous cohort studies evaluating the effects of recreational physical activity on ovarian cancer risk were conducted in the USA,6, 7, 8, 9 one of them suggesting that physical activity increases ovarian cancer risk6, 7 and 2 suggesting no association.8, 9 The 2 cohort studies, which accessed occupational physical activity, suggested an increase in risk ovarian cancer among women with less physically demanding jobs,10 or no evidence of association.11
There are several possible explanations for the discrepancy between our results, indicating no association between physical activity and ovarian cancer risk, and the results from other cohort studies – in particularly, those with individually collected data, like ours, and reaching inconsistent results (i.e. either of increased ovarian cancer risk, no association or suggestion of decreased risk). Our study includes mostly premenopausal women at cohort enrolment (96% of all participating women) and at follow-up time when ovarian cancer was eventually diagnosed, and with physical activity levels measured retrospectively in different periods of premenopausal life. The other main cohort studies,6, 7, 8, 9 which can be compared to ours, included mostly postmenopausal women, both at time of collection of information on physical activity levels and at cancer diagnosis. In premenopausal women, physical exercise lowers endogenous estrogens levels, and strenuous exercise may delay menarche, induce irregular menstrual cycles and in extreme levels, suppress ovulation.5 In postmenopausal women, the ovarian estrogens production falls dramatically, and the principal circulating estrogens, estrone, are produced by the aromatisation of androstenedione in the adipose tissue.5 The effects of physical activity on postmenopausal women are on lowering body fat or altering body fat distribution, and therefore altering extragonadal estrogen production, and may be increasing pituitary production of gonadotropins and serum gonadotropin levels. Physically active women do have lower estrone levels,5 and also reduced progesterone and increased androgen levels.29 Thus, the effect of physical activity in the physiology of pre- and postmenopausal women differs, as potentially the impact of physical activity on ovarian cancer risk according to menopausal status.
Physical activity has been assessed in very different ways in each of the cohort studies summarized earlier, and the inconsistent findings between the studies may also be, at least in part, explained by exposure misclassification, and, possibly, selection bias. In The IOWA Women's Health Study,6, 7 physical activity was measured at cohort enrolment among postmenopausal women only. Women participating in vigorous activities twice a week or moderate activity 4 times a week were considered as highly active, regardless of the time expended in such activities. In the Breast Cancer Detection Demonstration Project – BCDDP follow-up cohort9 – the typical physical activity of a weekday and a weekend day during the past year before cohort enrolment was assessed accounting for frequency and duration of activity. In the Nurses Health Study,8 measurements of frequency, duration and intensity of recreational physical activity were repeatedly assessed over a 14-year period, and changes in physical activity levels were considered in the summing up of person-years every second year. In this study, like in ours, no association between physical activity and ovarian cancer was found either in premenopausal or in postmenopausal women, and no difference in the risk of ovarian cancer for activity taking place before menopause compared with after menopause was observed.8
Thus, the literature does not indicate that physical activity has an independent strong effect on ovarian cancer risk. Although the methods for measuring physical activity differed markedly between studies, there was no indication that more sophisticated measurement scales yielded different results. Moreover, results on other established risk factors deriving from these same studies all show expected results.
Physical activity may have a distinct impact on subtypes of ovarian cancer, as suggested by the results of the Nurses Health Study.8 However, none of the studies published so far, including ours, had enough ovarian cancer cases to allow a precise analysis by ovarian cancer subtypes. Even considering all ovarian cancer subtypes together, still, the numbers of cases in each of the cohort studies separately is relatively small, which may lead to inability to detect true associations. Hence, chance may be one possible explanation for the discrepant results between studies.
In summary, our study does not support the hypothesis that physical activity has an independent role in determining a premenopausal women's risk of developing ovarian cancer. Moreover, we did not confirm previous cohort6, 7 or case control studies,16 suggesting that physical activity may increase ovarian cancer risk. Given numerous other beneficial health effects, physical activity should continue to be promoted for people at all ages.
The authors wish to thank all of the women who contributed to the study.
- 2American Cancer Society. Cancer facts and figures, 2004.
- 3Ovarian cancer. In: AdamiHO, HunterD, TrichopolosD, eds. Textbook of Cancer Epidemiology. Oxford: Oxford University Press, 2002. 378–94., .
- 5IARC handbooks of cancer prevention. Weight control and physical activity, vol. 6. Lyon: IARC Press, 2002. 316p.
- 24Pilot study for the evaluation of completeness of reporting to the cancer registry. Incidence of cancer in Norway, 1978. Oslo, Norway: The Cancer Registry of Norway, 1981. p 11–14..