SEARCH

SEARCH BY CITATION

Keywords:

  • physical activity;
  • cancer;
  • mortality

Abstract

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Higher physical activity levels have been associated with a lower risk of developing various cancers and all-cancer mortality, but the impact of pre-diagnosis physical activity on cancer-specific death has not been fully characterized. In the prospective National Institutes of Health-AARP Diet and Health Study with 293,511 men and women, we studied prediagnosis moderate to vigorous intensity leisure time physical activity (MVPA) in the past 10 years and cancer-specific mortality. Over a median 12.1 years, we observed 15,001 cancer deaths. Using Cox proportional hazards regression, we estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for MVPA with cancer mortality overall and by 20 specific cancer sites, adjusting for relevant risk factors. Compared to participants reporting never/rare MVPA, those reporting >7 hr/week MVPA had a lower risk of total cancer mortality (HR = 0.89, 95% CI 0.84–0.94; p-trend <0.001). When analyzed by cancer site-specific deaths, comparing those reporting >7 hr/week of MVPA to those reporting never/rare MVPA, we observed a lower risk of death from colon (HR = 0.70; 95% CI 0.57–0.85; p-trend <0.001), liver (0.71; 0.52–0.98; p-trend = 0.012) and lung cancer (0.84; 0.77–0.92; p-trend <0.001) and a significant p-trend for non-Hodgkins lymphoma (0.80; 0.62–1.04; p-trend = 0.017). An unexpected increased mortality p-trend with increasing MVPA was observed for death from kidney cancer (1.42; 0.98–2.03; p-trend = 0.016). Our findings suggest that higher prediagnosis leisure time physical activity is associated with lower risk of overall cancer mortality and mortality from multiple cancer sites. Future studies should confirm observed associations and further explore timing of physical activity and underlying biological mechanisms.

Abbreviations
BMI

body mass index

CI

confidence interval

HEI

healthy eating index

HR

hazard ratio

MVPA

moderate to vigorous physical activity

NIH

National Institutes of Health

RFQ

risk factor questionnaire

A recent annual report to the nation reported that more than one-third of U.S. men and women were considered to be physically inactive and that over half did not meet physical activity recommendations of 150 minutes of moderate to vigorous intensity activity each week.[1] Physical activity has been related to lower cancer incidence for the breast, colon and endometrium[2] and has also been associated with disease severity, recurrence and survival, particularly for breast and colorectal cancer.[3, 4] Existing studies on physical activity and mortality have shown an association between physical activity and lower risk of all cause[5, 6] and overall cancer mortality,[7-9] but the effect of leisure time physical activity on mortality from individual cancer sites has not been fully characterized.

We hypothesized that moderate to vigorous intensity leisure time physical activity (MVPA) would be inversely related to overall cancer death and that examination of physical activity in relation to specific cancer sites would generate hypotheses for future research. While physical activity and overall mortality have previously been examined in the National Institutes of Health (NIH)-AARP Diet and Health Study,[10, 11] the large, prospective nature and extended follow-up time in this cohort of men and women allowed for additional exploration of individual cancer sites. We thus analyzed pre-diagnosis physical activity and cancer-specific deaths in the NIH-AARP cohort to augment the paucity of evidence on physical activity and cancer mortality.

Methods

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Study population

The NIH-AARP Diet and Health Study has been previously described.[12] Briefly, the NIH-AARP cohort included 566,398 AARP members (aged 50–71 years) who completed a mailed baseline questionnaire in 1995–1996. In 1996–1997 an additional risk factor questionnaire (RFQ) including additional questions about participation in physical activity was mailed to participants who did not have self-reported cancer of the colon, breast or prostate at the time of the baseline questionnaire (response rate = 67%). Participants resided in six states (California, Florida, Pennsylvania, New Jersey, North Carolina, or Louisiana) or two metropolitan areas (Atlanta, Georgia or Detroit, Michigan). Of the 334,905 men and women who completed the baseline and risk factor questionnaires we excluded those whose questionnaires were completed by proxy (n = 10,383), those diagnosed with cancer before collection of physical activity data (n = 18,810), those with self-reported poor health (n = 4,382), those who moved out of the study area or died at or before processing of the questionnaires (n = 23) and individuals with missing information on physical activity (n = 7,796). After exclusions, 293,511 participants were included in this analysis. The NIH-AARP Diet and Health study was approved by the Special Studies Institutional Review Board of the U.S. National Cancer Institute, and all participants gave informed consent by virtue of completing and returning the questionnaire.

Mortality ascertainment

Participants were followed for address changes using the U.S. Postal Service's National Change of Address database, and vital status was ascertained periodically by linkage to the Social Security Administration Death Master File and the National Death Index Plus through 12/31/2008. To classify deaths due to specific cancers, we used ICD-9 (1995–1998) and ICD-10 (1999–2008) codes and considered individual cancer sites where there were more than 50 deaths. Mortality follow-up in this study is estimated to be more than 93% complete.[13, 14]

Exposure assessment

The baseline questionnaire (1995–1996) queried about demographic characteristics, diet, reproductive and medical history and lifestyle. The physical activity questions on the subsequent RFQ (1996–1997) predominantly assessed recreational and leisure time activities performed in the last 10 years. Examples of activities include tennis, golf, biking, swimming, heavy gardening, fast walking or dancing, aerobics and jogging. Participants reported categorical duration of activities (never, rarely, <1 hr/week, 1–3 hr/week, 4–7 hr/week and >7 hr/week). The NIH-AARP physical activity questionnaire has not been validated directly but has demonstrated expected inverse associations with all-cause mortality[10] and for incident cases of breast, colon and endometrial cancers in this cohort.[15-17]

Statistical analysis

Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards regression models with age as the underlying time metric in SAS version 9.2 (SAS Institute Inc, Cary, NC). Follow-up started at physical activity questionnaire completion and ended at death or the end of follow-up, whichever occurred first. MVPA was categorized into never/rarely (reference), <1 hr/week, 1–3 hr/week, 4–7 hr/week and >7 hr/week.

We explored all variables in Table 1 as potential confounders and retained factors that changed modeled estimates of overall cancer mortality by >10% or that are supported in previous literature as associated with cancer mortality. BMI was calculated as kg/m2 using baseline self-reported height and weight. Physical activity HR estimates were modeled with and without adjustment for BMI to address the possibility that BMI is in the causal pathway between physical activity and mortality. Although including BMI in models did not significantly change estimates, we included BMI in final models to eliminate concerns that the association between MVPA and cancer mortality is explained by BMI. We adjusted for cancer screening in the three years prior to questionnaire in analyses of breast and colon cancer-specific deaths. We did not adjust for prostate cancer screening because the U.S. randomized clinical trial found no association between screening and mortality for this cancer.[18] In analyses of female cancers (breast, endometrial, ovarian), we also adjusted for menopausal hormone therapy use (ever/never) and age at menarche (<12 years old, 12+ years old). We also performed analyses using a 31-level detailed smoking variable to see if our estimates were altered. MVPA estimates did not change substantially with the detailed smoking categories.

Table 1. Baseline characteristics in the NIH-AARP Diet and Health study population by level of moderate to vigorous-intensity physical activity (MVPA) (N = 293,511)
Frequency of MVPA (hr/wk)Never/rare<11–34–7>7
  1. a

    Measured only among women.

Cancer Deaths/Total participants2,524/41,1651,659/31,0953,756/74,5023,636/76,0493,426/70,700
Age at MVPA assessment, years62.862.462.662.863.2
Sex, %     
Women41.741.841.841.341.3
Body mass index, kg/m228.427.927.126.425.9
Education, %     
<High school/high school graduate28.523.621.020.524.1
Post high school/some college32.934.433.832.533.0
College or graduate degree36.139.742.944.740.3
Race/ethnicity, %     
Non-Hispanic White90.491.492.893.693.7
African American4.74.13.32.62.4
Other3.73.42.82.82.9
Married or living as married, %63.467.268.369.970.1
Television or video watching ≤2 hr/d, %28.333.235.838.637.9
Self-reported diabetes, %12.69.68.26.86.1
Family history of cancer, %52.452.853.353.454.3
Daily aspirin use, %26.024.424.524.925.4
Daily ibuprofen use, %11.810.110.09.59.8
Smoke, %     
Never33.735.636.436.536.6
Former48.048.349.350.850.8
Current14.812.911.29.69.2
Total energy intake, kcal/d18451810180518231931
Healthy Eating Index 2010 score62.864.766.267.768.1
Red meat intake, grams/1000 kcal/d37.036.634.832.531.4
Alcohol intake, g/d13.712.912.512.914.1
Coffee intake, g/d769.0774.9766.1767.3767.6
Multivitamin use, %51.554.756.457.657.3
Age at menarche <12 yearsa, %51.550.049.047.746.8
Ever MHT usea, %21.022.423.023.522.5

The final models thus included adjustment for sex, race (non-Hispanic white, African American, other), body mass index (BMI) (15 to <18.5 kg/m2, 18.5 to <25, 25 to <30, 30 to <35, 35+), smoking (never, quit >10 years before, quit 5–9 years before, quit 1–4 years before, quit <1 year prior or current smoker ≤20 cigarettes per day, quit <1 year prior or current smoker >20 cigarettes per day), alcohol intake (0, 0< to <3 and 3+ drinks/day), Healthy Eating Index 2010 (HEI-2010) score, energy intake, married or living as married (yes/no) and history of diabetes (yes/no). Tests for linear trend were performed coding MVPA categories in an ordinal fashion.

To maintain power in stratified analyses, we created a continuous MVPA measure by assigning the average number of hours per week to each category of physical activity and treated that parameter as a continuous variable. To assess effect modification and to detect potential residual confounding by smoking, we performed stratified analyses of the association between physical activity and cancer death. We also created an interaction term as the product of conttinuous MVPA and smoking level and entered the interaction term into the fully adjusted models, using the Wald test to evaluate the statistical significance of multiplicative interaction.

Because previous studies have shown joint associations of physical activity and BMI on both cancer risk and mortality, we also created models to examine a joint measure of physical activity and obesity. In a dichotomous measure “inactive” individuals included those who reported never/rare MVPA and “active” individuals included those who reported <1–7+ hr per week. Individuals with a BMI <30 kg/m2 were characterized as “non-obese” and individuals with a BMI ≥30 were characterized as “obese.”

In additional analyses, we stratified by sex and age at questionnaire completion (<55, 55 to <60, 60 to <65 and 65+ years old) and performed Wald tests for interaction. HR estimates are presented by cancer site for men and women combined because stratification by sex did not show substantial differences and marginal multiplicative interaction by sex as indicated by the Wald test was observed only for oral cancer deaths (p = 0.05). We also assessed robustness of estimates by restricting analyses to individuals without heart disease, diabetes and emphysema at baseline. Additionally, we performed analyses limiting the population to those with a BMI >15 and ≤50 kg/m2 to test whether estimates were influenced by outliers. We also ran models excluding individuals who were diagnosed within two years of completing the physical activity questionnaire to test whether latent disease affected our mortality risk estimates. The proportional hazards assumption, which was evaluated by modeling interaction terms of the main exposure trend with follow-up time, was met using the Wald test. All statistical tests were two-sided, with p-values <0.05 considered significant. We report results as significant where the p-trend test shows statistical significance and present the comparison of highest to lowest categories to show the estimated magnitude of effect at the highest reported activity levels.

Results

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Our analytic cohort included 293,511 participants (171,666 men and 121,845 women). The median time from MVPA assessment to cancer diagnosis was 5.2 years (range 0–13.2). Over a median 12.1 years, we observed 15,001 cancer deaths. Population characteristics are presented in Table 1. Compared to those reporting never/rare physical activity, those who reported higher levels of physical activity had a lower BMI and a healthier diet as scored by the HEI-2010. Active individuals were also more likely to be married or living as married, report regular multivitamin use and were less likely to report diabetes and smoking at baseline. Active women were less likely to report menarche at age <12 years.

We found a reduction in risk for overall cancer mortality with increasing MVPA (p-trend = <0.001), which reached statistical significance for those reporting greater than 1 hr per week of MVPA. Compared to those reporting never/rare MVPA, we observed a 7% lower risk for those reporting 1–3 MVPA hr/wk (HR = 0.93; 95% CI 0.88–0.98), a 10% lower risk for those reporting 4–7 hr/wk (0.90; 0.85–0.95) and a 11% lower risk for those reporting >7 hr/wk of MVPA (0.89; 0.84–0.94) (Table 2). Estimates for MVPA and cancer-specific mortality are presented in Table 2 by strongest inverse to strongest positive point estimates in the highest category of MVPA. We observed statistically significant inverse trends with higher MVPA for deaths due to cancers of the colon (HR = 0.70; 95% CI 0.57–0.85, p-trend = <0.001), liver (0.71; 0.52–0.98, p-trend = 0.012), lung (0.84; 0.77–0.92, p-trend = <0.001) and non-Hodgkins lymphoma (0.80; 0.62–1.04, p-trend = 0.017). In sensitivity analysis, excluding those diagnosed within two years of the MVPA questionnaire, the trend became nonsignificant for non-Hodgkins lymphoma (0.86; 0.62–1.19, p-trend = 0.176) (Supporting Information Table 1). No statistically significant trends were observed for lymphocytic leukemia, oral cavity and pharynx, esophagus, myeloma, myeloid/monocytic leukemia, stomach, ovarian, prostate, bladder, breast, brain, pancreas or rectum cancer deaths. We observed an unexpected statistically significant positive trend between MVPA and mortality from kidney cancer (1.42; 0.98–2.03, p-trend = 0.016), which remained significant in the analyses excluding individuals diagnosed within two years of MVPA assessment (1.67; 1.04–2.70, p-trend = 0.013) (Supporting Information Table 1).

Table 2. Associations between moderate to vigorous intensity physical activity (MVPA) and cancer mortality in the NIH-AARP study population (N = 293,511)a
 Never/rare <1 1–3 4–7 >7 
MVPA (hr/wk) Cancer deathNo. of deathsHR (ref.)No. of deathsHR (95% CI)No. of deathsHR (95% CI)No. of deathsHR (95% CI)No. of deathsHR (95% CI)P trend
  1. a

    Models are adjusted for sex, BMI (categories), education, race, alcohol (categories of 0, 0< to <3, and 3+ drinks/day), Healthy Eating Index 2010 score, calories, marriage status, diabetes and smoke level (six categories describing status and dose).

  2. b

    Additionally adjusted for screening in the three years prior to questionnaire.

  3. c

    Additionally adjusted for hormone therapy use and age at menarche.

  4. d

    Excluded women who reported a hysterectomy.

All cancers25241.0016590.95 (0.89–1.01)37560.93 (0.88–0.98)36360.90 (0.85–0.95)34260.89 (0.84–0.94)<0.001
Individual cancer sites by inverse magnitude of point estimate comparing MVPA >7 hr/wk to never/rare MVPA
Lymphocytic leukemia201.00140.96 (0.48–1.89)461.30 (0.76–2.21)240.65 (0.35–1.19)240.68 (0.37–1.25)0.058
Colonb1981.001090.80 (0.63–1.01)2680.85 (0.70–1.02)2500.79 (0.65–0.96)2110.70 (0.57–0.85)<0.001
Liver821.00430.79 (0.54–1.14)1120.90 (0.68–1.21)780.64 (0.47–0.88)820.71 (0.52–0.98)0.012
Oral cavity and pharynx381.00200.83 (0.48–1.44)410.79 (0.51–1.24)380.76 (0.48–1.21)360.75 (0.47–1.20)0.217
Non-Hodgkins lymphoma1041.00901.19 (0.90–1.58)1370.76 (0.58–0.98)1540.83 (0.64–1.06)1430.80 (0.62–1.04)0.017
Esophagus881.00550.92 (0.65–1.29)1230. 91 (0.69–1.20)1270.96 (0.73–1.27)980.80 (0.60–1.08)0.251
Myeloma611.00340.75 (0.49–1.14)630.56 (0.40–0.81)900.77 (0.55–1.07)930.82 (0.59–1.15)0.579
Lung9231.005220.85 (0.76–0.95)12580.92 (0.84–1.00)10950.82 (0.75–0.90)10730.84 (0.77–0.92)<0.001
Myeloid/monocytic leukemia541.00501.27 (0.86–1.86)810.85 (0.60–1.21)1091.10 (0.79–1.54)820.86 (0.60–1.22)0.346
Stomach491.00341.00 (0.65–1.56)770.99 (0.69–1.42)760.97 (0.67–1.40)680.90 (0.61–1.31)0.541
Ovarianc611.00410.92 (0.62–1.36)900.83 (0.59–1.15)980.87 (0.63–1.21)970.91 (0.65–1.26)0.623
Prostate801.00550.97 (0.69–1.37)1070.79 (0.59–1.06)1451.03 (0.78–1.37)1260.93 (0.69–1.24)0.968
Bladder521.00451.25 (0.84–1.86)820.97 (0.68–1.38)820.95 (0.67–1.36)851.03 (0.72–1.46)0.698
Breastb, c701.00551.21 (0.82–1.80)1030.92 (0.65–1.29)990.97 (0.68–1.37)1091.08 (0.76–1.53)0.973
Brain571.00511.14 (0.78–1.66)1151.03 (0.75–1.42)1070.91 (0.65–1.26)1281.14 (0.82–1.56)0.791
Endometrialc, d231.00241.52 (0.85–2.69)270.79 (0.45–1.38)351.13 (0.66–1.93)341.21 (0.70–2.08)0.745
Pancreas1491.001451.35 (1.07–1.70)2921.14 (0.93–1.39)3381.28 (1.05–1.56)3151.25 (1.03–1.53)0.065
Kidney471.00361.10 (0.71–1.70)851.14 (0.80–1.64)1061.47 (1.03–2.09)931.42 (0.98–2.03)0.016
Rectum181.00161.26 (0.64–2.48)471.57 (0.90–2.71)391.27 (0.72–2.25)481.63 (0.93–2.84)0.150

To explore the possibility of effect modification by smoking, in Table 3 we present analyses stratified by smoking status for MVPA and cancer death overall and by cancer site. Multiplicative interaction was not observed for any of the cancer sites (all p-interactions <0.05) except for ovarian cancer death (p = 0.022). Still, stratified ovarian cancer death hazard ratios were not statistically significant for never (1.04; 0.99–1.09) or ever (0.96; 0.92–1.01) smokers.

Table 3. Associations for a 1 hr increase in moderate to vigorous intensity physical activity (MVPA) per week and cancer mortality in the NIH-AARP study population, stratified by smoking status (N = 293,511)a
 Total populationNever smokersEver Smokers 
Type of cancer deathNo. of eventsHR (95% CI)No. of eventsHR (95% CI)No. of eventsHR (95% CI)p-interaction
  1. a

    Models are adjusted for sex, body mass index, education, race, alcohol intake, Healthy Eating Index 2010 score, calories, marriage status, diabetes and smoking intensity (where appropriate).

  2. b

    Significant at a p < 0.05 level.

  3. c

    Additionally adjusted for hormone therapy use and age at menarche.

All cancers150010.99 (0.98–0.99)b32000.99 (0.98–1.00)11,8010.99 (0.98–0.99)b0.782
Lymphocytic leukemia1280.93 (0.88–0.99)b390.90 (0.81–1.00)b890.95 (0.88–1.02)0.744
Colon10360.97 (0.95–0.99)b3090.98 (0.94–1.01)7270.96 (0.94–0.98)b0.507
Liver3970.96 (0.93–0.99)b1180.92 (0.87–0.98)b2790.98 (0.94–1.02)0.112
Oral cavity and pharynx1730.98 (0.93–1.02)250.93 (0.81–1.06)1480.98 (0.93–1.04)0.597
Non-Hodgkins lymphoma6280.97 (0.95–1.00)b2180.96 (0.92–1.01)4100.98 (0.95–1.01)0.606
Esophagus4910.98 (0.96–1.01)620.96 (0.89–1.05)4290.99 (0.96–1.02)0.635
Myeloma3411.01 (0.97–1.04)1211.02 (0.96–1.08)2201.00 (0.96–1.05)0.840
Lung48710.98 (0.98–0.99)b2521.00 (0.96–1.04)46190.98 (0.97–0.99)b0.454
Myeloid/monocytic leukemia3760.99 (0.95–1.02)1000.99 (0.93–1.06)2760.98 (0.95–1.02)0.884
Stomach3040.99 (0.95–1.02)740.98 (0.91–1.05)2300.99 (0.95–1.04)0.745
Ovarianc3871.00 (0.97–1.03)1841.04 (0.99–1.09)2030.96 (0.92–1.01)0.022
Prostate5131.01 (0.98–1.03)1320.98 (0.93–1.04)3811.01 (0.98–1.05)0.258
Bladder3460.99 (0.96–1.03)600.94 (0.87–1.03)2861.00 (0.97–1.04)0.433
Breastc4361.01 (0.98–1.04)1890.99 (0.95–1.04)2471.02 (0.98–1.06)0.653
Brain4581.01 (0.98–1.04)1661.02 (0.97–1.07)2921.00 (0.96–1.03)0.335
Endometrialc1461.02 (0.97–1.07)731.02 (0.94–1.09)731.02 (0.94–1.09)0.791
Pancreas12391.02 (1.00–1.03)3701.00 (0.96–1.03)8691.02 (1.00–1.05)b0.314
Kidney3671.04 (1.01–1.08)b991.07 (1.00–1.14)b2681.03 (0.99–1.08)0.598
Rectum1681.03 (0.98–1.08)490.99 (0.91–1.09)1191.04 (0.99–1.11)0.385

Joint associations for physical activity and BMI with all cancer mortality are presented in Table 4. Compared to the reference group of those who were active (reported more than never/rare MVPA) and non-obese (BMI<30 kg/m2), we found an increased risk of cancer mortality for those who were inactive and non-obese, (HR = 1.12; 95% CI 1.06–1.18) active and obese (1.16; 1.11–1.21) and inactive and obese (1.22; 1.13–1.32). Individuals who were obese, whether active or inactive, showed a 39% to over twofold statistically significant increased risk of death due cancers of the colon, liver, breast and endometrium compared to those who were active and non-obese.

Table 4. Joint associations between obesity (body mass index 30+ kg/mb) and moderate to vigorous intensity physical activity (MVPA, <1–7+ hr/wk) and cancer mortality in the NIH-AARP study population (N = 293,511)a
 Active/non-obese Inactive/non-obese Active/obese Inactive/ obese
Type of cancer deathNo. of eventsHR (ref.)No. of eventsHR (95% CI)No. of eventsHR (95% CI)No. of eventsHR (95% CI)
  1. a

    Models are adjusted for sex, body mass index (categories), education, race, alcohol (categories of 0, 0< to <3 and 3+ drinks/day), Healthy Eating Index 2010 score, calories, marriage status, diabetes and smoke level (six categories describing status and dose).

  2. b

    Additionally adjusted for hormone therapy use and age at menarche.

  3. c

    Excluded women who had a hysterectomy.

  4. Abbreviations: HR: hazard ratio, CI: confidence interval

All cancers10,0011.001,7581.12 (1.06–1.18)2,4761.16 (1.11–1.21)7661.22 (1.13–1.32)
Lymphocytic leukemia901.00120.95 (0.52–1.75)180.97 (0.58–1.63)81.60 (0.77–3.33)
Colon6391.001161.22 (1.00–1.49)1991.39 (1.18–1.64)822.01 (1.59–2.54)
Liver2311.00411.12 (0.80–1.56)841.62 (1.25–2.09)412.64 (1.88–3.72)
Oral cavity and pharynx1121.00321.43 (0.96–2.14)230.97 (0.62–1.53)60.73 (0.32–1.67)
Non–Hodgkin's lymphoma4061.00711.28 (0.99–1.65)1181.35 (1.09–1.66)331.38 (0.96–1.98)
Esophagus2971.00631.30 (0.99–1.72)1061.63 (1.30–2.04)251.28 (0.85–1.94)
Myeloma2261.00401.36 (0.97–1.91)541.14 (0.84–1.54)211.67 (1.06–2.64)
Lung33471.007011.16 (1.06–1.26)6010.87 (0.80–0.95)2221.01 (0.88–1.16)
Myeloid/monocytic leukemia2621.00310.87 (0.60–1.26)601.08 (0.81–1.43)231.55 (1.00–2.38)
Stomach1971.00321.05 (0.72–1.53)581.37 (1.01–1.84)171.40 (0.85–2.31)
Ovarianb2641.00371.05 (0.74–1.48)621.05 (0.79–1.39)241.45 (0.95–2.22)
Prostate3401.00571.17 (0.88–1.56)931.45 (1.15–1.82)231.33 (0.87–2.04)
Bladder2351.00350.98 (0.68–1.40)591.19 (0.89–1.59)171.20 (0.73–1.98)
Breastb2591.00411.09 (0.78–1.52)1071.73 (1.37–2.19)291.58 (1.07–2.35)
Brain3201.00411.04 (0.75–1.45)811.22 (0.95–1.56)160.95 (0.57–1.58)
Endometrialb, c731.00111.03 (0.55–1.96)502.48 (1.70–3.61)122.04 (1.09–3.82)
Pancreas8801.001150.91 (0.74–1.10)2101.12 (0.96–1.30)340.64 (0.45–0.91)
Kidney2401.00290.81 (0.55–1.19)801.50 (1.16–1.94)181.18 (0.72–1.91)
Rectum1231.00120.66 (0.36–1.19)270.99 (0.65–1.50)60.77 (0.34–1.77)

Restricting analyses to individuals without heart disease, diabetes and emphysema did not change results (data not shown). Excluding individuals with a BMI <15 kg/m2 or >50 also did not affect associations (data not shown). Stratification by age at questionnaire completion suggested stronger associations among older participants, but multiplicative interaction terms were not significant (data not shown).

Discussion

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

In our study population, higher MVPA before diagnosis was independently associated with a reduced risk of total cancer mortality. The observed inverse association was primarily driven by lower risks for death from cancers of the colon, liver and lung and non-Hodgkins lymphoma. Analyses of the joint effects of BMI and MVPA suggested lower risks of cancer death for active, non-obese individuals for various cancer sites.

Few studies report on physical activity and cancer mortality. A previous study in this cohort with less follow-up and fewer deaths found that those reporting >7 hr of MVPA each week had a 17% lower risk of cancer mortality (0.83; 0.74–0.93) compared to individuals reporting no MVPA.[19] In a large Taiwanese cohort, researchers found a lower overall cancer mortality risk (0.83; 0.77–0.90) comparing those with higher leisure time physical activity (∼20 metabolic equivalent hours per week) to inactive individuals.[20] In this Taiwanese study, significant inverse associations were observed for deaths due to colon and rectum cancer, liver cancer and lung cancer, while the association was not significant for breast cancer mortality. Another study using a treadmill test to assess fitness and a questionnaire to report physical activity levels showed that fitness and reported physical activity each were associated with lower overall cancer mortality among men, while among women there was a suggested but not significant association.[21] However, that study[21] had a small number of female deaths due to cancer (n = 44), possibly contributing to the lack of statistically significant findings.

In this cohort, previous physical activity analyses focused largely on cancer incidence, reporting inverse associations for liver, renal cell, breast, endometrial and colon cancers.[17, 22-25] Another study in this cohort on the joint effects of physical activity and adiposity on all-cause mortality also showed independent effects of each on mortality risk.[11] Our finding on the joint association between physical activity and body mass index with all-cancer mortality adds to the published literature in this cohort and is consistent with a previous publication in a large cohort of women reporting independent and joint effects of physical activity and BMI on reduced all-cancer mortality risk.[8]

To our knowledge, previous studies have not comprehensively reported on physical activity and risk of death from individual cancers. A growing body of energy balance survivorship literature was summarized in a recent review of 27 observational studies on physical activity among cancer survivors, reporting that physical activity was associated with lower rates of all-cause mortality and cancer-specific mortality for breast and colon cancers.[4] A recent forum on obesity, energy balance and cancer survival also highlighted protective effects among those who were normal weight and active for breast and colon cancer-specific mortality among survivors.[26] Our observed inverse association between physical activity and colon cancer mortality is consistent with these findings. A study in the Women's Health Initiative reported an association between pre-diagnosis moderate intensity activity and breast cancer death,[27] but similar to our findings, other studies have shown no association between pre-diagnosis physical activity and breast cancer death.[20, 28, 29] These conflicting findings could be due to differences in physical activity questionnaires, differences in time periods from measurement to diagnosis and follow up time or differences in study populations, such as the age at diagnosis or the percentage who reported menopausal hormone therapy usage. Literature on physical activity and mortality from other specific cancer sites is limited, although the above-mentioned review summarized limited evidence of an inverse association for prostate cancer mortality and no association for ovarian cancer mortality.[4] Also, a previous study on pre-diagnosis physical activity and endometrial cancer mortality in this cohort also showed no association.[30] Our observed positive association between MVPA and kidney cancer death contrasts with previous findings on MVPA cancer incidence in this cohort, which showed an inverse association between physical activity and renal cancer, which accounts for about 80% of kidney cancers.[25] However, other prospective cohort studies have shown no association between physical activity and renal cancer risk[31-33] and both risk factors and potential mechanisms related to physical activity may be different for incidence than for cancer mortality. A recent meta-analysis showed protective associations between leisure-time physical activity and lung cancer risk,[34] and the high fatality of lung cancer may translate into similarities between risk factors for incidence and mortality. Published findings on physical activity and liver cancer are limited, although a previous analysis in this cohort showed an inverse association between physical activity and risk of developing liver cancer.[22]

Mechanisms explaining the association between physical activity and mortality have been more widely studied in relation to cancer incidence. Review papers on mechanisms by which physical activity may affect cancer risk cite metabolic effects of high physical activity including lower body mass index, lower circulating sex hormones and insulin levels and possibly effects on inflammation or the immune system.[35, 36] These proposed mechanisms likely differ by specific cancer site. Some exercise intervention studies among survivors show improvements in circulating insulin levels and biomarkers related to cancer progression and recurrence.[37, 38] It is also possible that individuals who are active before diagnosis might have healthful habits (e.g., exercise) that are easier to maintain after diagnosis.

Strengths of our study include the large, prospective nature of the cohort and sufficient follow-up time. Other strengths include the objective endpoint (mortality) as well as a comprehensive range of cancer deaths, allowing examination of site-specific cancers. Detailed covariate information allowed us to examine effect modification by other risk factors such as BMI, smoking and diabetes.

Limitations of our study include measurement errors associated with self-report of physical activity. Although we had only a single measure of physical activity, we asked about physical activity in the past 10 years, attempting to reflect individuals' usual activity levels over time. Also, we lacked information on physical activity levels after cancer diagnosis, which may differ from pre-diagnosis levels. We adjusted for known confounders but we cannot eliminate the possibility that results were influenced by other lifestyle factors associated with physical activity, including better health maintenance and health insurance coverage. Lastly, we cannot rule out a possibility of observing associations by chance due to multiple comparisons.

Notably, the inverse association between MVPA and cancer mortality risk became significant with 1 hr or more of physical activity per week, and the magnitude of the protective association increased only slightly with more MVPA, as shown by similar point estimates in the higher MVPA categories. This exploratory analysis was the first to examine MVPA and site-specific cancer mortality and was intended to generate new hypotheses. In particular, inverse associations observed with MVPA and deaths from colon, liver and lung cancers should be further explored. The observed positive association with kidney cancer death also merits investigation. In addition, future research should isolate associations with moderate versus vigorous activities, post-diagnosis physical activity and whether changes in activity levels from pre to post-diagnosis affect cancer-specific mortality.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Cancer incidence data from the Atlanta metropolitan area were collected by the Georgia Center for Cancer Statistics, Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, Georgia. Cancer incidence data from California were collected by the California Cancer Registry, California Department of Public Health's Cancer Surveillance and Research Branch, Sacramento, California. Cancer incidence data from the Detroit metropolitan area were collected by the Michigan Cancer Surveillance Program, Community Health Administration, Lansing, Michigan. The Florida cancer incidence data used in this report were collected by the Florida Cancer Data System (Miami, Florida) under contract with the Florida Department of Health, Tallahassee, Florida. The views expressed herein are solely those of the authors and do not necessarily reflect those of the FCDC or FDOH. Cancer incidence data from Louisiana were collected by the Louisiana Tumor Registry, Louisiana State University Health Sciences Center School of Public Health, New Orleans, Louisiana. Cancer incidence data from New Jersey were collected by the New Jersey State Cancer Registry, Cancer Epidemiology Services, New Jersey State Department of Health, Trenton, New Jersey. Cancer incidence data from North Carolina were collected by the North Carolina Central Cancer Registry, Raleigh, North Carolina. Cancer incidence data from Pennsylvania were supplied by the Division of Health Statistics and Research, Pennsylvania Department of Health, Harrisburg, Pennsylvania. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations or conclusions. Cancer incidence data from Arizona were collected by the Arizona Cancer Registry, Division of Public Health Services, Arizona Department of Health Services, Phoenix, Arizona. Cancer incidence data from Texas were collected by the Texas Cancer Registry, Cancer Epidemiology and Surveillance Branch, Texas Department of State Health Services, Austin, Texas. Cancer incidence data from Nevada were collected by the Nevada Central Cancer Registry, State Health Division, State of Nevada Department of Health and Human Services, Las Vegas, Nevada. The authors are indebted to the participants in the NIH-AARP Diet and Health Study for their outstanding cooperation. The authors also thank Sigurd Hermansen and Kerry Grace Morrissey from Westat for study outcomes ascertainment and management and Leslie Carroll at Information Management Services for data support and analysis.

References

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ijc28659-sup-0001-suppinfo01.doc58KSupplemental Table 1. Associations between moderate to vigorous intensity physical activity (MVPA) and cancer mortality in the NIH-AARP study population, excluding those diagnosed within 2 years of MVPA assessment (N=279,498)a

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.