Physical activity and the risk of ovarian cancer: A case-control study in Canada

Authors

  • Sai Yi Pan,

    Corresponding author
    1. Surveillance and Risk Assessment Division, Centre for Chronic Disease Prevention and Control,Public Health Agency of Canada, Ottawa, Ontario, Canada
    • Surveillance and Risk Assessment Division, Centre for Chronic Disease Prevention and Control, Public Health Agency of Canada, 120 Colonnade Road, Locator 6702A, Ottawa, Ontario, Canada, K1A 0K9
    Search for more papers by this author
    • Fax: +613-941-2057.

  • Anne-Marie Ugnat,

    1. Surveillance and Risk Assessment Division, Centre for Chronic Disease Prevention and Control,Public Health Agency of Canada, Ottawa, Ontario, Canada
    Search for more papers by this author
  • Yang Mao,

    1. Surveillance and Risk Assessment Division, Centre for Chronic Disease Prevention and Control,Public Health Agency of Canada, Ottawa, Ontario, Canada
    Search for more papers by this author
  • The Canadian Cancer Registries Epidemiology Research Group

    Search for more papers by this author
    • The Canadian Cancer Registries Epidemiology Research Group comprises a principal investigator from each of the Provincial Cancer Registries involved in the National Enhanced Cancer Surveillance System: Bertha Paulse, Newfoundland Cancer Foundation; Ron Dewar, Nova Scotia Cancer Registry; Dagny Dryer, Prince Edward Island Cancer Registry; Nancy Kreiger, Cancer Care Ontario; Erich Kliewer, CancerCare Manitoba; Diane Robson, Saskatchewan Cancer Foundation; Shirley Fincham, Division of Epidemiology, Prevention and Screening, Alberta Cancer Board; and Nhu Le, British Columbia Cancer Agency.


Abstract

We evaluated the impact of recreational and occupational physical activity on ovarian cancer risk using data from a population-based case-control study of 442 cases with histologically confirmed incident ovarian cancer and 2,135 controls aged 20–76 years, conducted in 1994–1997 in Canada. Frequency and intensity of physical activity were collected through self-administered questionnaires. Compared to women in the lowest tertiles of moderate, vigorous and total recreational activity, those in the highest tertiles had multivariable-adjusted odds ratios (and 95% confidence intervals) of 0.67 (0.50–0.88), 0.93 (0.70–1.24) and 0.73 (0.58–0.98), respectively. There were statistically significant trends of decreasing risk with increasing levels of moderate and total recreational activity, with similar patterns for premenopausal and postmenopausal women. A significant reduction in risk associated with higher level of moderate recreational activity was observed for serous, endometrioid and other but not mucinous types of tumors. The analyses in one province with the largest number of cases and controls indicated that occupational activity was associated with reduced ovarian cancer risk by lifetime activity and by various life periods (early 20s, early 30s, early 50s and 2 years before interview). Our study suggests that occupational and regular moderate recreational physical activity reduce ovarian cancer risk. © 2005 Wiley-Liss, Inc.

Although the evidence for a protective effect of physical activity on colon and breast cancer is convincing, the evidence for an inverse association between physical activity and ovarian cancer is insufficient.1, 2 If proven to reduce ovarian cancer risk, physical activity—a modifiable lifestyle factor—could provide a method to prevent this fourth most common neoplasm in women, which has a poor prognosis.

The influence of physical activity on ovarian cancer risk was examined in 10 studies with inconsistent results.3, 4, 5, 6, 7, 8, 9, 10, 11, 12 Five studies suggested a decreased risk of ovarian cancer with high levels of physical activity,3, 4, 5, 6, 7 3 found no association,8, 9, 10 and 2 observed a positive association.11, 12 The discrepancy could be due to the fact that different studies assessed different parameters (including type, frequency, duration, intensity) of physical activity in populations with different profiles.

Using data from the Canadian National Enhanced Cancer Surveillance System (NECSS), which collected detailed information on recreational physical activity, occupational physical activity (in 1 province) and other risk factors for cancers, we further evaluated the relation of physical activity to ovarian cancer risk.

Material and methods

Study population

The NECSS was a multicomponent, collaborative project of Health Canada and the provincial cancer registries. The case-control component included individual data from 21,020 Canadians with 1 of 19 types of cancers and 5,039 population controls aged 20–76 years. The respective ethics review boards of each province reviewed and approved the study proposal. The data were collected from 1994–1997 in 8 of the 10 Canadian provinces (Alberta, British Columbia, Manitoba, Newfoundland, Nova Scotia, Ontario, Prince Edward Island and Saskatchewan). The detailed methods were described elsewhere.13 The current analysis was based on 442 incident cases of ovarian cancer and 2,135 female controls from all 8 provinces except Manitoba.

Ovarian cancer cases were recruited and selected from a review of pathology reports received by the provincial cancer registries. Those women with primary ovarian cancer (defined as C56 by the International Classification of Diseases for Oncology, 2nd ed. [ICD-O-2]) newly diagnosed from 1994–1997 in the 7 participating provinces were included in our study. All ovarian cases included in the NECSS were histologically confirmed by a surgical pathologist's report or cytology. The histologic subtypes were grouped into 4 categories based on ICD-O-2: mucinous, serous, endometrioid and others. The histologic codes were verified by each provincial cancer registry. The cancer registries identified most cases within 1–3 months of diagnosis to decrease the loss of subjects caused by severe illness and death. The registries identified 782 women with ovarian cancer. Physicians refused consent to contact 49 (6.3%) subjects, and 89 cases (11.4%) had died before they could be sent questionnaires. Questionnaires were sent to 644 cases; 20 questionnaires were returned because of an incorrect address and no updated address could be found through publicly available sources. Of the 624 eligible cases contacted, 442 completed and returned the questionnaires (70.8% of eligible cases and 56.5% of ascertained cases).

The NECSS used frequency matching to the overall case group (19 types of cancers) to select population controls with similar age and sex distribution, so that there would be at least 1 control for every case within each sex and 5-year age group for any specific cancer site within each province. The sampling strategy for control selection varied by province depending on data availability, data quality (completeness and timeliness) and the confidentiality restrictions of provincial databases. The provincial cancer registries recruited 3,578 potential female controls without cancer in the 7 provinces studied and mailed these females the same questionnaires as those sent to cases. For 286 (8.0%) of these women, the questionnaires were returned with an incorrect address and no updated one could be found. In all, 2,339 controls completed and returned the questionnaires, representing 71.0% of those contacted and 65.4% of those ascertained. However, 204 controls were excluded from our analysis because they had both ovaries removed at least 2 years before interview.

Data collection

Self-administered questionnaires, with telephone follow-up when necessary for clarification and completeness, were used to collect data. The registries used the same protocol to collect data for both cases and controls.

Each subject was assigned a reference date defined as 2 years before interview. The questionnaires were designed to obtain detailed data on risk factors for cancers. The questionnaire collected information on education, average family income over the last 5 years, marital status, ethnic group, height, weight, physical activity, alcohol consumption, diet (69-item food-frequency questionnaire) and vitamin and mineral supplements at the reference date. Questionnaires also gathered information about smoking history, menstrual and reproductive history, employment history, residential history and history of occupational exposure to some specific carcinogens. In the province of Ontario, information was also collected on hormone replacement therapy (HRT), oral contraceptive use and first-degree relatives (mother, father, sister and brother) diagnosed with cancer.

Assessment of physical activity

Respondents were asked in which seasons, how often, on average, they participated in each of the 12 most common types of leisure-time physical activity in Canada. Individual activities included walking for exercise, jogging or running, gardening or yard work, home exercise or exercise class, golf, racquet sports, bowling or curling, swimming or water exercise, skiing or skating, bicycling, social dancing and other strenuous exercise. Respondents indicated their usual frequency of participating in each of the above activities by choosing one of the following categories: never, less than once per month, 1–3 times per month, 1–2 times per week, 3–6 times per week or every day. We estimated the intensity of each activity by assigning a specific metabolic equivalent (MET) value to each reported activity. The MET values used here were abstracted from the Compendium of Physical Activities.14

We combined the intensity and frequency to form a physical activity index, i.e., the MET scores were multiplied by the midpoint of the reported frequency of the activity, then converted to frequency of activity per week and summed to create a relative unit of recreational physical activity per week.15 For example, a woman who reported participation in walking (4.0 MET) 3–6 times per week, in golf (4.5 MET) 1–2 times per week and in racquet sports (8.0 MET) 1–3 times per month would be assigned 28.46 relative units of physical activity per week ([19.5 × 4.0 + 6.5 × 4.5 + 2 × 8.0]/4.33). We categorized levels of recreational activity as moderate (MET ≥3 to ≤6), vigorous (MET >6) and total (moderate plus vigorous).16 The variable used in the analysis was the sum of each category of moderate, vigorous and total (moderate plus vigorous) physical activity.

In the province of Ontario with the largest number of cases and controls, information on occupational physical activity (including that of homemakers, students and volunteers) was also collected in Ontario for 4 life periods: early 20s, early 30s, early 50s and reference date, and recorded as sitting, light, moderate or strenuous to represent subjects' usual type of activity in their daily work, job or occupation. Questionnaires gave examples for each category of activity. We assessed the cumulative lifetime occupational activity by summing the original category scores across appropriate life periods. We then divided the cumulative lifetime activity by the number of life periods summed to create another index: averaged lifetime occupational activity.

Statistical analysis

We evaluated the risk of ovarian cancer associated with moderate and vigorous recreational physical activity in all provinces studied and for occupational physical activity in Ontario. Unconditional logistic regression was used to compute the odds ratios (ORs) and corresponding 95% confidence intervals (CIs) with the software package SAS (version 8). Variables were categorized by tertiles based on the distribution of the variables in the control population. Variables of cumulative and average lifetime occupational activity were categorized by quartiles based on the distribution in the control population in Ontario.

We assessed the potential confounding effect of a wide range of factors, including age, province of residence, educational level, family income adequacy, marital status, ethnic group, alcohol consumption, smoking, body mass index, total calorie intake, menopausal status, age at menarche, age at first birth, number of pregnancies, number of live births and years of menstruation. We retained variables in the final models that are considered biologically important, even if they did not reach statistically significant levels in the initial assessment. We adjusted the final multivariate models for 10-year age groups, province of residence, education (years, <10, 10–12, >12), alcohol consumption (servings per week, continuous), smoking pack-years (continuous), body mass index (kg/m2, <25, 25–29, ≥30), total calorie intake (kcal per week, continuous), total vegetable consumption (servings per week, quartiles), number of live births (0, 1, 2, 3, ≥4), and menopausal status (postmenopausal, premenopausal). We conducted tests for trend for all models of categorized data by treating the different categories as a single ordinal variable.

Since menopause is related to the level of sex hormones, which might be one of the possible mechanisms underlying the association between physical activity and ovarian cancer,2, 17 we further explored possible effect modification on the association between physical activity and ovarian cancer by menopausal status. Because body mass index has been suggested to be related to levels of insulin, insulin-like growth factors (IGFs) and their binding protein (IGFBP)-3, which may also be affected by physical activity,2, 17 we also assessed the effect modification by body mass index. Analyses for Ontario subjects only were also adjusted for cancer in first-degree relatives (yes, no), HRT (ever used for 6 months or more, never) and oral contraceptive use (ever used for 6 months or more, never). The analyses for occupational physical activity were also adjusted for recreational physical activity at corresponding period.

Results

Table I displays selected characteristics of cases with ovarian cancer and of controls. Cases and controls were similar in mean age, total years of education, alcohol consumption, age at menarche, years of menstruation and marital status. Compared to controls, cases were more likely to have smoked slightly longer pack-years, to be obese, to have consumed more calories and slightly less vegetables, to be postmenopausal and to have had fewer live births.

Table I. Characteristics of Ovarian Cancer Cases and Controls (n = 2,577),NECSS, Canada, 1994–1997
CharacteristicsCases (n = 442)Controls (n = 2,135)
  1. NECSS, National Enhanced Cancer Surveillance System; SD, standard deviation.

Age (years) (mean ± SD)55.1 ± 12.355.2 ± 12.5
 20–292.32.7
 30–399.38.4
 40–4921.724.2
 50–5925.821.9
 60–6927.828.0
 70–7613.114.7
Alcohol drinking (drinks/week) (mean ± SD)2.2 ± 4.32.6 ± 5.4
Smoking pack-years (mean ± SD)8.4 ± 12.57.5 ± 12.8
Total calorie intake (kcal/day) (mean ± SD)1874.6 ± 742.81823.5 ± 969.4
Total vegetable intake (servings/week) (mean ± SD)21.0 ± 13.021.7 ± 15.2
Menstruation years (mean ± SD)31.8 ± 8.031.8 ± 7.8
Age at menarche (mean ± SD)12.9 ± 1.612.9 ± 1.6
Premenopausal women (%)29.938.3
Postmenopausal women (%)70.161.7
Body mass index (kg/m2) (mean ± SD)26.3 ± 6.625.2 ± 5.2
 <25.049.057.8
 25.0–29.930.128.7
 ≥30.020.913.5
Total education (years) (mean ± SD)12.3 ± 3.312.2 ± 3.6
 <1016.318.7
 10–1240.639.4
 >1243.141.9
No. of live births (%)
 024.115.1
 112.39.9
 228.229.0
 320.919.3
 ≥414.526.7
Marital status (%)
 Married67.267.3
 Common-law2.73.2
 Divorced/separated11.39.3
 Widowed10.013.9
 Single8.65.9
 Other0.20.3
Never smoked (%)46.850.8
Ex-smoker (%)33.529.0
Current smoker (%)19.720.1
Low family income (%)14.916.7
Lower-middle family income (%)17.416.6
Upper-middle family income (%)22.223.5
High family income (%)19.914.8
Family income not reported (%)25.628.4

Table II presents the odds ratios for ovarian cancer associated with moderate, vigorous and total recreational physical activity, overall and stratified by menopausal status. Compared to women in the lowest tertiles of moderate, vigorous and total activity, women in the highest tertiles had multivariable-adjusted ORs (and 95% CIs) of 0.67 (0.50–0.88), 0.93 (0.70–1.24) and 0.73 (0.58–0.98), respectively. There were statistically significant trends of decreasing ovarian cancer risk with increasing levels of moderate and total recreational activity (p for trend: 0.006 and 0.037). However, we did not find a statistically significant decrease in ovarian cancer risk for vigorous activity. The risk reduction associated with physical activity was similar among pre- and postmenopausal women.

Table II. Odds Ratios for Ovarian Cancer Associated with Moderate and Vigorous Recreational Physical Activity, Overall and by Menopausal Status, NECSS Study, Canada, 1994–1997
Recreational physical activity(MET units1/week)All women (n = 2,577)Premenopausal (n = 949)Postmenopausal (n = 1,628)
Cases (n)Controls (n)Age- and province- adjustedMultivariable-adjusted2Cases (n)Controls (n)Age- and province- adjustedMultivariable-adjusted3Cases (n)Controls (n)Age- and province- adjustedMultivariable-adjusted3
OR95% CIOR95% CIOR95% CIOR95% CIOR95% CIOR95% CI
  • NECSS, National Enhanced Cancer Surveillance System; OR, odds ratio; CI, confidence interval.

  • 1

    See Material and Methods for calculation of MET units/week.

  • 2

    OR adjusted for 10-year age group, province of residence, education, alcohol consumption, cigarette pack-years, body mass index, total calorie intake, total vegetable consumption, number of live births and menopause status.

  • 3

    OR adjusted for all above variables except menopause status. –

  • 4

    Reference category.

Moderate (tertiles)
 0–7.4941687121.00 1.00 502811.00 1.00 1184311.00 1.00 
 7.50–24.161607120.880.69–1.150.890.69–1.17452950.810.52–1.300.800.50–1.291154170.890.66–1.230.900.66–1.24
 ≥24.171147110.680.51–0.890.670.50–0.88372410.620.37–1.030.580.35–0.99774700.680.47–0.950.670.46–0.94
 p for trend  0.0090.006  0.0810.042  0.0290.021
Vigorous
 041317401.00 1.00 351971.00 1.00 965431.00 1.00 
 >0–4.151776981.160.89–1.511.140.87–1.50462950.880.54–1.450.850.51–1.421314031.170.85–1.621.170.84–1.63
 ≥4.161346970.920.69–1.220.930.70–1.24513250.810.49–1.330.810.48–1.35813720.880.62–1.250.890.62–1.28
 p for trend  0.510.56  0.410.44  0.480.54
Total (tertiles)
 0–11.6141717111.00 1.00 462701.00 1.00 1254411.00 1.00 
 11.62–34.631427120.870.67–1.120.900.69–1.16442750.850.53–1.370.870.53–1.41984370.840.61–1.160.880.63–1.21
 ≥34.641297120.750.58–0.970.730.58–0.98422720.720.38–1.100.690.36–1.05874400.730.53–1.000.700.52–0.99
 p for trend  0.0320.037  0.1540.075  0.0570.046

Because some studies suggest that risk factors of ovarian cancer may differ by tumor type, we examined whether the activity-related ovarian cancer risks varied by histologic types of ovarian tumors (Table III). For moderate physical activity, increasing activity level was associated with a statistically significant trend of decreasing risk of serous, endometrioid and other types of ovarian tumors but not of mucinous tumor. For vigorous activity, no association was observed for serous and endometrioid tumors; however, increased risks of mucinous and other types of tumors were found in some categories of vigorous activity with no consistent trend.

Table III. Odds Ratios for Ovarian Cancer Associated with Moderate and Vigorous Recreational Physical Activity, by Histology Type, NECSS Study, Canada, 1994–1997
Recreational physical activity(MET units1/week)MucinousSerousEndometrioidOthers
Controls(n)Cases(n)Province- andage-adjustedMultivariable-adjusted2Cases (n)Province- andage-adjustedMultivariable-adjusted2Cases (n)Province- andage-adjustedMultivariable-adjusted2Cases (n)Province- andage-adjustedMultivariable-adjusted2
OR95% CIOR95% CIOR95% CIOR95% CIOR95% CIOR95% CIOR95% CIOR95% CI
  • NECSS, National Enhanced Cancer Surveillance System; OR, odds ratio; CI, confidence interval.

  • 1

    See Material and Method for calculation of MET units/week.

  • 2

    OR adjusted for 10-year age group, province of residence, education, alcohol consumption, cigarette pack-years, body mass index, total calorie intake, total vegetable consumption, number of live births and menopause status.

  • 3

    Reference category.

Moderate
 0–7.493712251.00 1.00 871.00 1.00 261.00 1.00 301.00 1.00 
 7.50–24.16712271.040.60–1.841.040.59–1.85860.900.64–1.250.880.62–1.23160.620.32–1.160.640.33–1.21310.850.49–1.450.770.44–1.33
 ≥24.17711180.660.35–1.230.700.33–1.17530.600.42–0.890.580.40–0.87150.620.29–1.150.520.23–0.99280.650.36–1.090.580.31–0.98
 p for trend  0.190.22 0.0150.011 0.140.049 0.100.041
Vigorous
 03740191.00 1.00 701.00 1.00 191.00 1.00 231.00 1.00 
 >0–4.15698301.760.96–3.251.770.96–3.30891.000.70–1.430.990.69–1.43231.150.61–2.201.140.59–2.25351.951.12–3.421.981.11–3.52
 ≥4.16697211.210.62–2.341.170.60–2.29670.790.55–1.140.790.54–1.15150.790.39–1.600.760.36–1.60311.550.88–2.751.650.91–2.96
 p for trend  0.600.67 0.220.22 0.540.49 0.150.11
Total
 0–11.613711291.00 1.00 861.00 1.00 251.00 1.00 311.00 1.00 
 11.62–34.63712210.700.39–1.260.710.40–1.29750.930.66–1.310.910.64–1.31180.780.42–1.470.820.43–1.55280.700.41–1.190.680.39–1.18
 ≥34.64712200.670.37–1.210.640.35–1.17650.750.53–1.070.730.50–1.01140.620.30–1.200.500.24–1.05300.690.39–1.150.640.33–1.03
 p for trend  0.180.14 0.1200.0580.170.0710.160.064

When we assessed the effect of recreational physical activity stratified by body mass index (<25, 25–29 and ≥30 kg/m2), we observed a risk reduction in relation to moderate activity among all 3 subgroups, with a somewhat larger effect among the overweight (25–29 kg/m2) and obese groups (≥30 kg/m2) (Table IV).

Table IV. Odds Ratios of Ovarian Cancer Associated with Moderate and Vigorous Recreational Physical Activity, by Body Mass Index (kg/m2), NECSS Study, Canada, 1994–1997
Recreational physical activity(MET units1/week)Low BMI (<25) (n = 1,444)Medium BMI (25–29) (n = 742)High BMI (≥80) (n = 378)
Cases (n)Controls (n)Age- and province- adjustedMultivariable-adjusted2Cases (n)Controls (n)Age- and province- adjustedMultivariable-adjusted2Cases (n)Controls (n)Age- and province- adjustedMultivariable-adjusted2
OR95% CIOR95% CIOR95% CIOR95% CIOR95% CIOR95% CI
  • NECSS, National Enhanced Cancer Surveillance System; BMI, body mass index; OR, odds ration; CI, confidence interval.

  • 1

    See Material and Methods for calculation of MET units/week.

  • 2

    OR adjusted for 10-year age group, province of residence, education, alcohol consumption, cigarette pack-years, total calorie intake, total vegetable consumption, number of live births and menopause status.

  • 3

    Reference category.

Moderate (tertiles)
 0–7.493753731.00 1.00 491991.00 1.00 441331.00 1.00 
 7.50–24.16754250.820.57–1.180.800.55–1.16522101.020.64–1.610.870.54–1.4032770.940.53–1.670.960.54–1.72
 ≥24.17654310.710.47–1.070.660.44–1.02312010.640.33–1.040.550.31–0.9816760.510.24–1.000.480.22–0.96
 p for trend  0.1000.065  0.1020.035  0.0610.022
Vigorous
 03603801.00 1.00 462251.00 1.00 241251.00 1.00 
 >0–4.15784210.970.66–1.430.930.62–1.39552001.160.72–1.871.170.72–1.9142771.770.93–3.401.840.94–3.60
 ≥4.16774280.960.65–1.430.970.64–1.44311850.720.42–1.230.680.39–1.1726841.200.60–2.421.260.62–2.58
 p for trend  0.860.89  0.250.17  0.660.60
Total (tertiles)
 0–11.613733721.00 1.00 532161.00 1.00 451211.00 1.00 
 11.62–34.63634310.770.52–1.120.720.49–1.07511981.100.69–1.750.980.61–1.5728811.040.57–1.901.100.59–2.03
 ≥34.64794260.850.61–1.250.790.53–1.11281960.590.35–1.000.510.30–0.8819840.550.29–1.050.520.27–1.01
 p for trend  0.340.21  0.0610.019  0.0940.079

We also examined whether the association between ovarian cancer and recreational physical activity varied by parity (nulliparous or parous) and by total energy intake (low, medium and high, kcal per week). These 2 variables did not significantly modify the association between physical activity and ovarian cancer (data not shown).

We also assessed the impact of occupational physical activity on ovarian cancer risk in the province of Ontario, by lifetime activity and by life periods, further controlling for oral contraceptive use, HRT use and family history of cancer. Women with jobs that required moderate or strenuous activity experienced a consistent reduction in ovarian cancer risk for all life periods compared to those who worked in sedentary occupations, with the effect being more evident for moderate than for strenuous activity (Table V).

Table V. Odds Ratios for Ovarian Cancer Associated with Occupational Physical Activity in Ontario by Life Period and by Cumulative and Averaged Lifetime Activity, NECSS Study, Canada, 1994–1997
Life period and activity intensityCases (n)Controls (n)Age-adjustedMultivariable-adjusted1
OR95% CIOR95% CI
  • NECSS, National Enhanced Cancer Surveillance System; OR, odds ratio; CI, confidence interval.

  • 1

    OR adjusted for 10-year age group, education, alcohol consumption, cigarette pack-years, body mass index, total vegetable consumption, number of live births, menopause status, cancer in first-degree relative, oral contraceptive use hormone replacement; ORs for 2 years previous, early 30s, early 50s and lifetime activity were also adjusted for combined moderate and vigorous recreational activity at corresponding period.

  • 2

    Reference category.

2 years previous
 Sitting21514611.00 1.00 
 Light512370.660.46–0.940.700.48–1.04
 Moderate442110.640.44–0.930.600.39–0.92
 Strenuous10330.940.46–1.960.760.33–1.73
 p for trend  0.0280.021
Early 20s
 Sitting21234001.00 1.00 
 Light552130.850.59–1.220.940.63–1.40
 Moderate642740.750.54–1.050.700.47–1.03
 Strenuous14550.810.43–1.500.720.36–1.41
 p for trend  0.110.068
Early 30s
 Sitting2872821.00 1.00 
 Light612190.910.63–1.320.930.61–1.42
 Moderate923720.780.56–1.080.760.52–1.12
 Strenuous12510.740.38–1.450.600.28–1.27
 p for trend  0.120.085
Early 50s
 Sitting2862201.00 1.00 
 Light481540.780.52–1.180.780.50–1.20
 Moderate401700.610.40–0.930.610.38–0.96
 Strenuous9310.750.34–1.630.670.29–1.57
 p for trend  0.0350.041
Cumulative lifetime (quartiles)
 Q12752361.00 1.00 
 Q2682340.930.63–1.341.140.74–1.75
 Q3522330.650.42–0.960.640.41–0.98
 Q4612370.720.49–1.060.710.46–1.12
 p for trend  0.0410.048
Averaged lifetime (quartiles)
 Q12772351.00 1.00 
 Q2712350.970.66–1.391.160.75–1.75
 Q3482370.580.40–0.840.590.41–0.92
 Q4602350.700.46–1.040.680.42–1.08
 p for trend  0.0170.024

Discussion

Our study results indicated that higher levels of moderate recreational physical activity were associated with a reduced risk of ovarian cancer. We also observed a reduction in ovarian cancer risk associated with moderate recreational physical activity for serous, endometrioid and other but not mucinous types of tumors, but we saw similar reductions for premenopausal and postmenopausal women. Analyses for Ontario subjects only suggested that occupational activity reduced ovarian cancer risk across various life periods, including early 20s, early 30s, early 50s and 2 years before interview.

The decreased risk of ovarian cancer associated with physical activity observed in our study agrees with several previous investigations. A hospital-based case-control study3 reported that increased total physical activity, including both recreational and occupational, was associated with a significantly lower risk of ovarian cancer among Chinese women. A large Italian case-control study4 found that occupational activity alone and occupational combined with leisure-time activity were associated with a significant reduction in ovarian cancer risk, whereas leisure-time activity was nonsignificantly related. Three other studies assessed only occupational physical activity: one case-control study6 suggested a lower risk among women in jobs with higher energy expenditure, another case-control study of 49 ovarian cancer cases and 244 controls10 showed a nonsignificant increase in risk with decreased energy expenditure, and one cohort study of small sample size9 found no association.

Two population-based case-control studies5, 8 and 3 cohort studies7, 11, 12 investigated only recreational physical activity. The large study of 767 cases with ovarian cancer and 1,367 controls5 observed a significant reduction of ovarian cancer risk associated with higher levels of physical activity, by lifetime total and at various life periods (ages 14–17, 18–21, 22–29, 30–39, 40–49 and ≥50 years. The other case-control study8 reported that total and vigorous recreational activity were not associated with a substantial decrease in ovarian cancer risk at age 12, age 20 and 5 years before interview, and only women reporting 42 or more MET hours per week experienced a nonsignificant risk reduction compared to sedentary women at 5 years before interview. The results from the Breast Cancer Detection Demonstration Project follow-up cohort7 suggested an inverse association between higher level of recreational physical activity and ovarian cancer risk. The prospective study of 16-year follow-up12 found that woman with 7 hr or more per week of recreational physical activity had a relative risk of 0.80 (95% CI 0.49–1.32) compared to those with less than 1 hr per week; however, women engaged in frequent vigorous activity had an increased risk of ovarian cancer. In The Iowa Women's Health Study cohort,11, 18 no association with frequency of moderate leisure-time activity but a significant positive association with more than 4 times per week of vigorous leisure-time activity were observed.

The finding that moderate recreational activity decreases and vigorous activity increases the risk for only some histologic types of ovarian tumor found in our study might explain the discrepancy between our study results and others because different studies may have different proportions of histologic types of ovarian tumors. However, no previous studies examined the association by tumor histologic type. The positive association found in the 2 prospective studies is observed only among people with frequent vigorous recreational activity. Few women in our study reporting a high level of vigorous recreational activity could be another explanation for the difference between our study and theirs.

Several plausible biologic mechanisms have been proposed for the protective effect of physical activity on ovarian cancer. Physical activity may influence ovarian cancer by mediating the levels of endogenous sexual and metabolic hormone and growth factors. Physical activity has been associated with decreased levels of circulating estrogen and progesterone in premenopausal women and with decreased serum estradiol, estrone and androgens in postmenopausal women,17, 19 as well as increased circulating concentrations of sex hormone-binding globulin, which reduces the amount of free, active hormones in circulation.2 Exposure to higher levels of these sex steroid hormones has been suggested as a risk factor for ovarian cancer.20, 21, 22 Physical activity may also increase the metabolism of estrogen to less potent forms.23, 24 Regular physical activity improves insulin sensitivity and may lower levels of insulin and bioavailable insulin-like growth factors (IGFs).25 Studies in humans found that circulating IGF-I levels were positively associated with ovarian cancer before age 55,26, 27 whereas circulating insulin-like growth factor binding protein (IGFBP)-1 and IGFBP-2 were inversely associated. Some research showed a decreased level of serum IGFBP-3 in women with epithelial ovarian cancer compared to those who had benign ovarian tumors or were healthy controls;28 low IGFBP-3 levels in tumor tissues were significantly associated with unfavorable prognosis.29

Physical activity may decrease ovarian cancer risk through its influence on obesity. Obesity, especially central obesity, has been shown to increase ovarian cancer risk.13, 18 Physical activity seems to preferentially reduce visceral adiposity30, 31, 32 and increase insulin sensitivity independent of body size.33, 34, 35 The greater risk reduction among overweight or obese women observed in our study supports the role of physical activity in affecting ovarian cancer risk through its effect on obesity and also implicates that obese women would get more benefit from physical activity against ovarian cancer than lean women. However, risk decrease also observed among women with normal weight suggests other mechanisms are involved in the effect of physical activity.

Physical activity may also decrease ovarian cancer risk by reducing the number of lifetime ovulatory cycles and thus decreasing lifetime exposure to endogenous estrogens, which is supported by observations that intense physical activity in girls results in delayed menarche,36, 37, 38, 39 anovulation40, 41 and amenorrhea.38, 39 However, these effects may be achieved only through regular intensive physical activity.

Other possible mechanisms include enhancement of the immune system and improvement of antioxidant defense systems. Regular moderate physical activity seems to enhance proliferation of lymphocytes, increasing the number of natural killer cells and the activity of lymphokine-activated killer cells,42 which is especially beneficial in older people with aging-related decrease in immune function, whereas too much exercise may cause immune suppression.17 Exercise improves antioxidant enzyme repair capacity but also increases oxygen uptake and the production of oxygen free radicals.43, 44 Therefore, exercise could be beneficial or deleterious, depending on the intensity of exercise, subject characteristics and nutritional status.17 Extreme exercise could cause excessive increase of free radicals and disturbance of intracellular prooxidant-antioxidant homeostasis and thus damage to biomolecules may occur.44, 45, 46, 47, 48 However, moderate exercise does not cause oxidative damage to healthy subjects.49 The decreased antioxidant defense capacity with aging can be partly compensated by exercise.50 Thus, older population could get significant benefit from moderate exercise by slowing or stopping the loss of antioxidants that is necessary to deal with the daily production of reactive oxygen species.17 On the other hand, severe exercise may overcome the antioxidant defense system with the potential oxidative DNA damage.51

We do not have an explanation for the differential associations between physical activity and various histologic types of ovarian tumors, but previous studies suggested that risk factors for ovarian cancer differ by histologic type.

Some potential limitations of our study should be addressed. A major limitation is that 20.2% of cases could not be included in this analysis because they died before they could be sent questionnaires or their physicians denied contact or could not be located, and 24.0% of cases did not return questionnaires. This low response rate among cases was largely due to the poor prognosis of ovarian cancer, and it could affect the generalization of our results. Recall bias is possible because cases might bias their responses to questions on physical activity after a few months of cancer diagnosis. The possibility that such a bias was introduced into our study was reduced by including many questions on other exposures, such as diet, employment history and residential history and by not placing any particular emphasis on physical activity in the questionnaire. In addition, we did not collect information on duration of each activity; thus, we could not assess the combined effect of all 3 parameters of physical activity (frequency, duration and intensity). The association seen with moderate rather than vigorous recreational activity could be because of a greater reliability of self-reported activities falling into the moderate spectrum than of activities classified as strenuous, i.e., greater nondifferential misclassification of vigorous exercise resulting in risk estimates closer to unity. Because the histologic classifications were done by pathologists in respective provinces and validated by respective provincial cancer registries other than reviewed and validated by a single expert pathologist, misclassification in histologic types was possible; however, the distribution of histologic subtypes of ovarian cancer cases in our study was similar to other studies.52, 53, 54

Our questionnaire was not validated. Self-reported physical activity assessment is commonly used in large-scale epidemiologic studies due to its economy and easy administration. However, extensive validation of self-reported physical activity questionnaires has not been conducted, in part because of a lack of acceptable criterion to determine the validity. Two studies suggested that self-administered physical activity questionnaires are reproducible and provide useful measure of physical activity with reasonable validity.55, 56

In conclusion, our population-based study observed a reduced risk of ovarian cancer among women engaged in higher levels of moderate but not vigorous recreational physical activity. The reduction associated with recreational activity was similar among premenopausal and postmenopausal women; however, a significant decrease in risk was observed for serous, endometrioid and other but not mucinous histologic types of ovarian tumors. The results from one province also suggested that occupational activity afforded protection again ovarian cancer across various life periods.

Ancillary