Previous reviews of the association between physical activity and colon cancer have reported a risk reduction of approximately 30%, based on qualitative review, when comparing the most to the least active individuals (Lee and Oguma, 2006
). Our formal meta-analysis of the data generally supports this, showing a 24% risk reduction overall, and generally similar risk reductions when men and women were examined separately. Several mechanisms have been proposed for the role of physical activity in reducing colon cancer risk including reduced insulin resistance and hyperinsulinemia, anti-inflammatory action, direct immune action, decreased intestinal transit time or higher vitamin D levels (Wolin et al, 2007
). As future investigations explore these pathways, they will likely provide additional insights on the association between physical activity and colon cancer risk.
Although each study quantified activity differently limiting our ability to draw conclusions about the amount of physical activity necessary for the 24% risk reduction observed, a recent example provides some information. In the US Nurses' Health Study, Wolin et al (2007
) report a 23% risk reduction when comparing the most to the least active women. The most active women expended more than 21.5 MET hours per week in leisure-time physical activity, whereas the least active expended less than 2 MET hours per week. These levels are equivalent to brisk walking for some 5–6
h per week in the most active and 0.5
h per week in the least active.
We found that the magnitude of risk reduction reported in case–control studies were stronger than those reported in cohort studies (30 vs
15% risk reduction), as also observed in a previous qualitative review (Lee and Oguma, 2006
) and a previous meta-analysis (Samad et al, 2005
). There are several possible reasons for this difference. Case–control studies may be subject to greater recall bias. Overall, we found no difference in risk reduction between men and women. This supports one previous qualitative review that found similar effects in women and men (Lee and Oguma, 2006
), and contradicts suggestions that the beneficial effects of exercise may be attenuated in women (McTiernan et al, 2006
; Mai et al, 2007
). In their meta-analysis, Samad et al (2005
) found a stronger association in men than women in cohort studies, but no difference in effects by gender in case–control studies . Similarly, we observed that cohort study results among women were less pronounced than those among men. This may be partly due to the small number of studies within each stratum. It may also reflect differences in the absolute physical activity values being compared, in that women typically report lower levels (intensity and duration) of physical activity than men, and higher levels of physical activity may be needed for risk reductions.
We found significant heterogeneity in the effects used in our analyses. This is not surprising given the variations in time of exposure assessment, length of follow-up, method of exposure assessment, type of physical activity assessed, levels of physical activity compared and covariates included in the analysis. To account for this heterogeneity, we used random effects models in our analyses.
Most cohort studies formally evaluated the presence of a dose–response effect. Of the cohort studies, only four studies did not report a test of trend. Among the dose–response effects evaluated, less than half reported a significant trend. Of note, several studies that examined dose–response effects in men and women separately found significant trends only in men (Chao et al, 2004
; Lee et al, 2007
; Takahashi et al, 2007
; Howard et al, 2008
; Nilsen et al, 2008
), but one found significant trends only in women (Thune and Lund, 1996
) and two in both men and women (Wei et al, 2004
; Moradi et al, 2008
). In contrast, only 10 case–control studies reported a test for trend, and four found significant results for at least one group. Because the studies used different physical activity measures in their tests for trend (e.g., energy expended, intensity, frequency, or duration), as well as different categorisation schemes, we did not conduct a formal meta-analysis of trend across studies ().
Meta-analysis of physical activity and colon cancer: cohort studies.
Increasing interest has focused on the type, intensity or duration of physical activity necessary for a protective effect (Physical Activity Guidelines Advisory Committee, 2008
). We were able to examine the effect of physical activity domain (occupational vs
leisure-time) and found that the results were similar. Few studies have reported results in sufficient detail to allow a formal evaluation of the different effects of intensity or duration. Qualitative evaluations have suggested that vigorous physical activity may be necessary to reduce the risk (Slattery, 2004
) though others have concluded that sufficient durations of moderate or vigorous intensity physical activity are likely to reduce the risk of colon cancer (Lee and Oguma, 2006
). Recently, it has been suggested that walking alone may be sufficient to reduce risk (Takahashi et al, 2007
; Wolin et al, 2007
) though not all studies agree (Chao et al, 2004
). Sufficient number of studies have not reported on the benefits of walking to allow formal evaluation of the effect. As additional studies report on the separate effects of intensity, duration and physical activity type, analyses of this data should be undertaken as these details are important to inform public health recommendations. Additional studies may also examine modification of the effect by race/ethnicity, BMI, diet and tumour location as the quantity of data on those factors increases.
We have previously hypothesised (Wolin et al, 2007
) that the association between physical activity and colon cancer may be attenuating over time as screening decreases the number of colon tumours overall and distal colon tumours in particular. Fraser and Pearce (1993
) examined secular trends in the association between occupational physical activity and risk of colon cancer and found that the association was stronger in the earlier era examined. However, this should be interpreted cautiously as other factors, including changes in the quality of physical activity assessment, may also contribute. In addition, the later studies also tended to adjust for larger numbers of potential confounders, typically attenuating relative risks. We found little evidence for a difference over time when we stratified the meta-analysis by publication year.
In conclusion, this meta-analysis provides additional support for the inverse association between physical activity and colon cancer. It provides a formal estimate showing that individuals can likely reduce their risk of colon cancer, overall, by 24% through participation in physical activity. Additional research on the type, intensity, and duration of physical activity that may afford the greatest risk reduction will inform public health recommendations regarding quantification of specific physical activity details.