Article Text

This article has a correction. Please see:

Download PDFPDF

Leisure time physical activity and cancer risk: evaluation of the WHO's recommendation based on 126 high-quality epidemiological studies
  1. Li Liu,
  2. Yun Shi,
  3. Tingting Li,
  4. Qin Qin,
  5. Jieyun Yin,
  6. Shuo Pang,
  7. Shaofa Nie,
  8. Sheng Wei
  1. Department of Epidemiology and Biostatistics, The Ministry of Education Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China
  1. Correspondence to Professor Sheng Wei, Department of Epidemiology and Biostatistics, The Ministry of Education Key Lab of Environment and Health, School of Public Health, Tongji Medical College, Huazhong University of Science and Technology, No.13 of Hangkong Road, Wuhan, Hubei 430030, China; ws2008cn{at}

Statistics from


Physical activity has a significant influence on the risk of developing non-communicable diseases (NCDs). The WHO states that, compared with inactive adults, active individuals have a lower frequency of coronary heart disease, high blood pressure, stroke, type 2 diabetes and so on. Accordingly, the WHO recommends at least 2.5 h of moderate-intensity, 1.25 h of high-intensity or an equivalent combination of moderate-intensity and high-intensity aerobic physical activity throughout the week to reduce the risk of NCDs.1 However, the ability of physical activity to reduce the risk of NCDs has only been satisfactorily validated and quantified for a minority of diseases.2 ,3 Importantly, quantitative evidence concerning the amount of physical activity necessary to be protective against other important NCDs, such as cancer, is still absent. Public health guidelines concerning the amount of physical activity required for health benefits rely on individual comparisons, rather than a systematic assessment of the available evidence. However, many of the published studies have grouped participants into quantitatively designated categories of special types of physical activity (eg, quartiles of leisure time physical activity (LTPA) or metabolic equivalents of energy hours per week (MET-h/wk)), making it possible to quantify the relationship between physical activity and cancer risk in a dose–response manner.

To this end, we conducted a meta-analysis of high-quality prospective epidemiological studies to summarise the current knowledge about the relationship between LTPA, the most common and adjustable domain of physical activity, and cancer risk in order to provide more in-depth and quantitative evidence on which to base public health guidelines.


Literature search

The meta-analysis was performed according to the Preferred Reporting Items for MOOSE (Meta-analysis Of Observational Studies in Epidemiology) guidelines.4 MEDLINE and Web of Science were searched for eligible studies published until 30 December 2014 with the following search strategy: (exercise [title/abstract] OR physical activity [title/abstract] OR walking [title/abstract] OR motor activity [title/abstract]) AND (cancer [title/abstract] OR neoplasm [title/abstract] OR carcinoma [title/abstract] OR tumor [title/abstract]). The references listed in any applicable publications and reviews were further screened for potential publications of interest.

Selection criteria

The selected studies had to: (1) be an original study, (2) be a cohort study design, (3) have measured cancer risk based on LTPA and (4) have provided relative risks (RRs) with the 95% CIs or sufficiently detailed frequency data to derive RRs and their corresponding 95% CIs. Publications were excluded if they (1) were from a case report, review, editorial, commentaries, meta-analysis or conference proceeding, and/or (2) included cohorts of patients with an original chronic disease. Duplicate studies reporting on the cancer risk and LTPA were carefully perused and the one with the largest sample size was included in analyses.

Data extraction

Two authors extracted relative data from each comparison and then added the information into a database independently. Any differences were resolved by discussion. For each study, we found the estimated effect on cancer risk (reported as a RR) and the corresponding 95% CI for the association of LTPA with cancer risk. Given the long incubation periods of cancer initiation and unsteady physical activity at different stages of life, for a study reporting the effect of LTPA at multiple time points or ages and over an individual's lifetime, we preferred the risk estimate for lifetime physical activity. If a study reported the results separately for males and females, both risk estimates were included in our primary analysis. For all comparisons derived from the studies, we compared the RR and the corresponding 95% CI of the most active group and the least active group. The resulting size of the effect and 95% CIs were inverted for comparisons in which the most active group was used as the reference group. For studies that reported LTPA in units of MET-h/wk or hours per week (h/wk), the LTPA range, effect size and 95% CIs were noted for each activity-level group. Other data recorded were the first author's name, publication year, age at recruitment, number of cases and participants, median follow-up time, type of cancer studied, gender of the participants, the country where the study took place and adjustment for confounders.

Quality assessment

Two reviewers completed the quality assessment independently using the Newcastle-Ottawa Scale.5 Briefly, this scale awards a maximum of nine points to each study based on the quality of the study group selection (maximum 4 points), study population comparability (maximum 2 points) and measurement of the outcome of interest (maximum 3 points). For each criterion, a higher numerical score indicates higher quality, while a score closer to 0 indicates lower quality. Studies that scored a total of 0–3, 4–6, and 7–9 were categorised as low, moderate and high quality, respectively.

Statistical analyses

Fixed-effects or random-effects meta-analyses were used to estimate the summarised relative cancer risk for highest versus lowest LTPA when appropriate.6 Heterogeneity was assessed using I2 indexes and Q statistics.7 Publication bias was measured visually by assessing funnel plots and statistically using Begg's test.8 A two-stage random-effects dose–response meta-analysis was performed to compute the correlation of the log RR estimates across levels of LTPA. Only comparisons with three or more quantitative physical activity levels were included in these analyses. LTPA reported in the form of time was transferred into metabolic equivalents of energy. Considering the different intensities of activity components, the reported weekly hours were multiplied by 8 MET for vigorous activity, by 4 MET for moderate activity, and by 6 MET for moderate-to-vigorous activity.9 ,10 For each study, the median or mean level of LTPA was assigned to the corresponding RR. The midpoints of the upper and lower boundaries of each category were used if the median or mean were not reported. For studies that reported the open upper boundaries or uppermost boundaries closed with extreme value, we assumed the width of the interval to be the same as in the closest category.11 A restricted cubic spline model with four knots set at the 5th, 35th, 65th and 95th centile levels of LTPA was then produced using the generalised least square regression and taking into account the correlation between each set of published RRs.12 Study-specific estimates were combined using the restricted maximum likelihood method in a multivariate random effects meta-analysis.13 A p value for non-linearity was calculated by testing the null hypothesis that the coefficient of the second spline was equal to 0. Subanalyses were performed only when at least four original comparisons were available. All statistical tests were two-sided, and statistical significance was considered at p<0.05. All analyses were conducted using Stata software (V.12.0; StataCorp, College Station, Texas, USA).


Study selection

The initial search yielded 17 002 studies for initial consideration. After removing duplicated publications from two databases and title/abstract screening of each article, 4312 studies warranted further evaluation. In total, 4182 of the 4312 studies were excluded after careful evaluation of the study design and data sufficiency, whether the study focused on LTPA and cancer risk. Among the 130 studies remaining, we excluded four publications which provided data from the same population with a smaller sample size or shorter follow-up period. This resulted in a final 126 studies that were included in the meta-analysis (figure 1).

Figure 1

Selection of studies for inclusion in this meta-analysis, 1989–2014.

Study characteristics

The main characteristics of the 126 studies included in the meta-analysis are presented in online supplementary table S1.10 ,14–138 Among these studies, 119 publications were included in primary binary meta-analysis, which totalled 169 comparisons. The follow-up periods of these studies ranged from 2 to 30 years, during which time 199 820 cancer incidences were identified in 7 304 954 cohort populations. The numbers of comparisons to estimate the cancer risk for different genders, study locations and cancer types are listed in table 1, among which 92, 72, 67, 38, 40 and 39 studies had been adjusted for age, body mass index, smoking, family history of cancer, alcohol use and education levels during the calculation of RRs, respectively. Sixty-nine comparisons from 54 studies provided quantitative estimates of LTPA categories in units of MET-h/wk or h/wk (table 1 and online supplementary table S1). Totally, 88 853 cancer incidences were found among 2 919 977 cohort populations in these 54 studies. Among the 126 studies, 7 were of moderate quality and 119 of high quality.

Table 1

Pooled measures on the relation of LTPA to cancer risks, 1989–2014

Binary analysis

Compared with the lowest level of LTPA, the highest level had pooled cancer risk of 0.90 (95% CI 0.88 to 0.92, I2for heterogeneity=69.7%). Subanalyses indicated a similar cancer risk reduction for LTPA in males and females. Geographically, active LTPA resulted in the largest cancer risk reduction in America (RR=0.86, 95% CI 0.82 to 0.90), followed by Europe (RR=0.94, 95% CI 0.91 to 0.97) and then Asia (RR=0.95, 95% CI 0.91 to 1.00). There was a significant cancer prevention effect by LTPA in individuals with normal weight (RR=0.89, 95% CI 0.81 to 0.98) but not in overweight ones. LTPA in older individuals but not in younger ones caused a significant cancer risk reduction, with an RR of 0.95 (95% CI 0.92 to 0.97). LTPA displayed a cancer prevention role in ever smokers (RR=0.77, 95% CI 0.61 to 0.98) but not in never smokers. In addition, studies of high quality presented a more conservative effect than those of moderate quality, with RRs of 0.92 (95% CI 0.90 to 0.94) and 0.66 (95% CI 0.59 to 0.74), respectively. Studies with a follow-up period of at least 10 years showed a 6% increase in cancer prevention than those with a follow-up period of less than 10 years. LTPA exhibited protective effects against breast cancer and colorectal cancer with RRs of 0.88 (95% CI 0.84 to 0.91) and 0.84 (95% CI 0.77 to 0.93), respectively. LTPA showed a higher protection rate against breast cancer in premenopausal women than in postmenopausal women, with RRs of 0.79 and 0.89, respectively. Further analyses showed that the cancer prevention role of LTPA was restricted to invasive breast cancer (RRfor invasive breast cancer=0.85, 95% CI 0.78 to 0.92) and colon cancer (RRfor colon cancer=0.81, 95% CI 0.75 to 0.88). We did not find a relationship between LTPA and other cancer types. Funnel plots and Begg's tests did not provide significant evidence to suggest publication bias except for studies focusing on prostate cancer, American patients with breast cancer and normal weight individuals (table 1 and online supplementary figure S1).

Dose–response analysis

Pooled results from 69 comparisons indicated the expected inverse non-linear relationship between LTPA and cancer risk (Pfor non-linearity<0.001; figure 2 and table 2). As compared with individuals who failed to undertake LTPA, the RR of cancer for 10 MET-h/wk of LTPA equal to the average recommendation from the WHO was 0.93 (95% CI 091 to 0.95). Total cancer risk quickly decreased to levels of LTPA below the recommended amount, and then slowly decreased with increasing levels of LTPA above the recommended amounts. A 2% cancer risk reduction occurred with each additional 3 MET-h/wk below 10 MET-h/wk of LTPA compared with a 1% cancer risk reduction by every additional 20 MET-h/wk above 20 MET-h/wk (figure 2). In females, we found an approximate 1% reduction in cancer risk for every 2 MET-h/wk below 10 MET-h/wk and an approximate 3% reduction in cancer risk for every additional 20 MET-h/wk above 20 MET-h/wk. Meanwhile, the protective role of LTPA against cancer in males started at 20 MET-h/wk. The 40 MET-h/wk of LTPA presented a saturated protective role against cancer in males (figure 3A, B). Furthermore, the influence of activity on preventing cancer was consistently found in studies performed in America, Europe and Asia (figure 3C–E). In a subgroup analysis by cancer types, inverse associations between LTPA and incidence of breast cancer and colorectal cancer were found. The cancer reduction role of LTPA against breast cancer dropped dramatically below the recommended 10 MET-h/wk and became subdued above the recommendation. The recommended 10 MET-h/wk caused a 4% reduction in breast cancer risk. Specifically, the protective role of LTPA against colorectal cancer started at 10 MET-h/wk, with an RR of 0.92 (95% CI 0.85 to 1.00) (figure 3F, G). Consistent with the binary analysis result, LTPA did not present a role in cancer prevention among other cancer types. Consistent with binary analysis results, significant cancer prevention roles were found in studies reposted in later life LTPA and studies of high quality (figure 3H, I). The inverse association between LTPA and the cancer prevention role was more conservative in studies with a follow-up period of at least 10 years than in those with a follow-up period of less than 10 years (figure 3J, K). Moderate-to-vigorous LTPA showed a similar cancer prevention role with total LTPA. For energy expenditure above 20 MET-h/wk, the role of moderate-to-vigorous LTPA in cancer prevention was stronger than that of total LTPA at the same energy expenditure (figure 3L).

Table 2

Dose–response analysis of relative risk for cancer in relation to LTPA, 1989–2014

Figure 2

The dose–response analysis between cancer risk and leisure time physical activity in the form of energy expenditure (metabolic equivalents of energy hours per week (MET-h/wk)), 1989–2014. The solid line and the long dashed lines represent the estimated relative risk and corresponding 95%CI, respectively. The short dashed line represents the linear relationship.

Figure 3

The dose–response analysis between cancer risk and leisure time physical activity (LTPA) in the form of energy expenditure (metabolic equivalents of energy hours per week (MET-h/wk)) from 1989 to 2014 for female (A), male (B), America (C), Europe (D), Asia (E), breast cancer (F), colorectal cancer (G), age at measurement at least 50 years (H), quality score of study more than 6 (I), follow-up at least 10 years (J), follow-up less than 10 years (K) and moderate-to-vigorous LTPA (L). The solid line and the long dashed lines represent the estimated relative risk and corresponding 95% CI, respectively. The short dashed line represents the linear relationship.


Principal findings and comparison with other studies

This meta-analysis is the largest to date summarising the contribution of LTPA to reducing cancer risk. Furthermore, it is the first study to quantify the dose–response relationship between LTPA and risk of all cancers with regard to the amount of LTPA and the magnitude of the reduction in cancer risk. In line with previous meta-analyses and reviews, we identified a protective role for LTPA against cancer.139–141 A 10% reduction in total cancer risk was revealed by comparing the individuals participating in the highest levels of LTPA with those that do none. Specifically, this meta-analysis revealed the average recommendation of physical activity by the WHO produced a 7% reduction in risk of cancer. Besides, prior to reaching the recommended amount of physical activity per week, a 2% reduction in cancer risk occurs with every additional 3 MET-h/wk of activity, whereas on reaching two-fold of the recommended amount, a 1% reduction in cancer risk needs an additional 20 MET-h/wk. This means that the cancer preventive role of physical activity should not be simplified as a negative linear relationship.

From this huge meta-analysis, a significant protective role for LTPA was observed against breast cancer and colorectal cancer. Further analyses showed that LTPA mainly displayed a protective role against invasive breast cancer and colon cancer, which was also supported by previous studies.142 ,143 A number of plausible candidate mechanisms have been proposed. For colon cancer, the major hypothesis is that some hormones, such as insulin, may be biological mechanisms through which physical activity influences colon cancer risk. Giovannucci144 proposed the insulin–colon cancer hypothesis, in which they suggest that insulin resistance leads to colon cancer through the growth promoting effect of insulin, glucose or triglycerides. This hypothesis has been widely supported by other researchers.145 Besides, LTPA could reduce colon cancer risk by lowering faecal bile acid concentrations and decreasing the gastrointestinal transit time of food to reduce their stimulation to intestinal epithelia.146 For invasive breast cancer, it is hypothesised that physical activity alters the endogenous production of sex steroid hormones by altering menstrual cycle patterns and controlling body weight among postmenopausal women.147–149 In addition, other mechanisms including the impact of physical activity on adiposity, insulin resistance, adipokines and inflammatory markers may also contribute to the protective effect.150

From binary meta-analysis and dose–response meta-analysis, we found that the protective role of LTPA against cancer was stronger in studies conducted in America than in those conducted in Europe and Asia. Recently, a research investigating the physical activity level among college students of different populations found that Asians had the lowest physical activity, and Caucasians had the highest physical activity, which may reflect the different physical activity patterns in different populations.151 Therefore, we could infer that the baseline physical activity in ‘inactive’ individuals is higher in Europeans than in Americans and Asians, which results in the underestimation of the protective role of LTPA against cancer in Europeans. Besides, Asians are more physically inactive than Americans and the physical activity in ‘active’ individuals is lower in Asians than in Americans, which may cause a weaker protective role of physical activity in Asians.

The binary analysis results showed that physical activity exhibited a role in cancer prevention in the normal weight population but not in the overweight population. Given the common carcinogenesis mechanisms of obesity and physical inactivity, such as abnormalities of hormone secretion, insulin resistance and inflammatory response, obesity may offset the protection of physical activity against cancer.144 Furthermore, we found that studies with a longer follow-up period presented more conservative effects than those with a shorter follow-up period, which is consistent with previous meta-analyses.152 We further found that LTPA presented the role of cancer prevention in ever smokers but not in never smokers. Since the included original studies are limited, the results should be understood properly.

Strengths and weakness of the review

Overall, this meta-analysis is a comprehensive evaluation of the evidence examining the association of LTPA levels with cancer risk. To the best of our knowledge, this is the first meta-analysis to investigate the dose–response relationship between LTPA and total cancer risk, as well as several subgroup analyses. Reassuringly, the results from the binary comparisons and dose–response analyses were comparable. However, there were also several limitations to this study that should be acknowledged. First, significant heterogeneity existed for several outcomes that could not be explained by our prespecified subgroups. Heterogeneity was used to reflect the concordance of the original results and somehow the uncertainty of the association indirectly. Therefore, for analyses with significant heterogeneity, the pooled results most likely reflect the uncertainty of whether there is an inverse relation. The existing heterogeneity limits our understanding of the association between physical activity and cancer risk in various settings and the generalisability of our findings. Second, although we used high-quality cohort studies, our results still could have been biased by residual confounding, which could occur in either direction depending on its nature. Third, in some of the studies included in this analysis, LTPA was assessed by self-reporting. This makes some misclassification of activity levels probable, and therefore quantitative characterisations should be considered approximate.

Conclusions and recommendations

Our meta-analysis indicates that the current WHO recommendation about physical activity can result in a 7% reduction in cancer risk, which is mainly attributed to its protective role against invasive breast cancer and colon cancer. Furthermore, the protective role of physical activity against cancer becomes saturated at two-fold of current recommendation. Therefore, besides the benefits previously stated by the WHO, our meta-analysis suggests that compared with less active adult men and women, individuals who are more active have a lower risk of cancer, especially invasive breast cancer and colon cancer. The current recommendation of physical activity caused a 7% reduction in total cancer risk. Increase of recommended level to two-fold of the current level (20 MET-h/wk) may give nearly the saturated effect against cancer.

What are the findings

  • The total cancer risk was reduced by 10% in people who undertook the most leisure time physical activity as compared with those who did the least.

  • Pooled results indicate the expected inverse non-linear relationship between leisure time physical activity and cancer risk.

  • According to a recommendation from the WHO, energy expenditure equivalent to 2.5 h/wk of moderate-intensity activity lowered the risk of cancer by 7%. For physical activity less than the recommended amount, a 2% cancer risk reduction occurred with each additional 3 MET-h/wk, while for more than two-fold of the recommendation, every additional 20 MET-h/wk caused a 1% cancer risk reduction.

How might it impact on clinical practice in the future?

  • It brings a quantised conception about the role of the current recommendation of physical activity in cancer reduction. Physical activity may be emphasised in the future clinical practice for cancer prevention.

  • It may also simulate the clinical researches about the intervention of physical activity in cancer survivors, especially in breast cancer and colorectal cancer survivors.

  • The public health guidelines of physical activity may be changed due to the nearly saturated effect against cancer by two-fold of current recommendation (20 MET-h/wk).


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Correction
    BMJ Publishing Group Ltd and British Association of Sport and Exercise Medicine