A meta-analysis of the association of physical activity with reduced risk of colorectal cancer
Mr M. Chapman, Consultant Colorectal Surgeon, Good Hope Hospital, Rectory Road, Sutton Coldfield, West Midlands B75 7RR, UK.
Background Physical activity may be associated with reduced risk of colorectal cancer. The main aim of this paper is to review the available evidence for a link between exercise and large bowel cancer.
Methods A Cochrane-type methodology was performed. Data extracted included, type of study, type of physical activity measured and the numerical results. The risk ratios (RR) of the studies have been pooled according to the type of study, type of exercise, type of cancer and sex. Pooling was undertaken using fixed effect meta-analysis. A random effect meta-analysis was used where substantial heterogeneity existed.
Result Data from 19 cohort studies showed a statistically significant reduction in the risk of colon cancer in physically active males, RR being 0.79 (95% CI 0.72–0.87) and 0.78 (95% CI 0.68–0.91) for occupational and recreational activities, respectively. In women only recreational activities are protective against colon cancer (RR = 0.71, 95%CI 0.57–0.88). Case-control studies showed significantly reduced risks of colon cancer in both sexes irrespective of the type of activity. No protection against rectal cancer is seen in either sex.
Conclusion There is considerable evidence that physical activity is associated with reduced risk of colon cancer in both males and females.
Colorectal cancer is a leading cause of morbidity and mortality with about 300 000 new cases and 200 000 deaths in Europe and the USA each year . The rates of colon cancer vary considerably among countries with high rates in industrialized nations and low prevalence in developing countries. Incidence of colorectal cancer increases among groups migrating from low to high incidence areas  within one or two generations, suggesting exogenous influences, and a number of lifestyle factors have been suggested to explain this international variation [3,4]. Established risk factors for colon cancer include a low consumption of fibre and folate, high consumption of animal fat, red meat and a family history of colon cancer . The association of physical activity with reduced risk of various types of cancers has generated much interest recently. The literature published, mainly in epidemiological journals, suggests that there is a causal association between physical activity and cancer of the colon. A sedentary life style is attributed to influence changes in hormone and growth factor levels, increased body fat content and impaired immune function possibly promoting the development of cancer.
Reviews looking at the effect of physical activity in reducing colon cancer have mainly been observational and there are no published data in the surgical literature. Colditz et al.  in their review found that in general, studies showed an inverse relation between physical activity and risk of colon cancer and a consistent ‘dose–response relation’ with those exhibiting the highest level of activity being at significantly lower risk of developing colon cancer compared with the least active group.
The aim of this paper is to systematically review the available evidence linking physical activity (both exercise and occupational) with large bowel cancer.
Materials and methods
A search was conducted on Medline, Embase and the Cochrane Library database for all publications in the English language up to January 2002. A combination of text search terms was used: physical activity, exercise, colon and colorectal cancer. In addition the reference lists of all included studies were checked for additional studies.
Studies were included based on the following criteria:
- • study design: either case-control or cohort;
- • exposure: to any form of physical activity that could include occupational activity, recreational or both;
- • control: the inclusion of a no physical activity group;
- • outcomes: were development of colon, rectal or colorectal cancer.
To be included in this review, the studies needed to focus on some type of physical activity assessment. Studies have defined physical activity as occupational activity, ranging from a job title to detailed lifetime history of actual occupational activities, or as recreational activity ranging from simple lists of activities to detailed personalized records. Physical activity data have been extracted from existing records and databases or have been measured directly from individuals by interviews or questionnaires.
Data extracted for each study included details on type of study, type of physical activity measured and the numerical results. Given the high level of heterogeneity in the study designs and methods used for measuring physical activity and for analysing the result, no attempt at estimating overall quantitative synthesis combining data across studies was made.
A single investigator (AKAS) performed study selection and data extraction.
The results of each study were expressed as risk ratios (RR), i.e. relative risks or odds ratios (OR) together with their confidence intervals (CI) for cohort and case control studies, respectively. The risk ratios of the studies were pooled according to the type of study (case-control or cohort); type of exercise (occupational, leisure, or both); type of cancer (colon or rectal); and sex.
Pooling was undertaken using fixed effect meta-analysis, except where substantial heterogeneity existed according to the χ2 statistic a random effect meta-analysis was used. All analyses were performed using Stata v.6 software.
A total of 47 studies on physical activity and colon cancer were performed in different parts of the world including China, Japan, New Zealand, Spain, Sweden, Turkey and US, were included in the analysis of which 19 cohort studies [7–25] and 28 are case-control studies [26–53](Tables 1 and 2). Four studies were excluded, one that did not report details of classification of physical activity  and three others, as they were preliminary reports, which were later revised in subsequent manuscripts [55–57]. Of the 47 colon and colorectal cancer and physical activity studies 40 demonstrated reduction of risk in cancer in males, while only 18 showed reduced risk in females. In studies, which considered rectal cancer separately, only 8 showed any reduction in risk of cancer.
Table 1. Cohort studies included in analyses. C = colon cancer; R = rectal cancer.
|Gerhardsson et al.*||1986||1100000 (M)||5100 C (M)||O||RR = 1.3(1.2–1.5)|
|Paffenbarger et al.†||1987||56683 |
|201 C |
21 CR (M)
|RR = 0.91 (P = 0.604) |
RR = 0.5 (P = 0.038)
RR = 0.80 (P = 0.168)
|Wu et al.†||1987||11888||58 CR (M) |
68 CR (F)
|L||RR = 0.4 (0.2–0.8) M; P for trend 0.008 |
RR = 0.9 (0.5–1.6) F
|Gerhardsson et al.*||1988||16477||99 C (M) |
92 C (F)
|RR = 1.5 (0.7–3.4) M |
RR = 1.7 (0.6–4.4) F
RR = 1.4 (0.8–2.6) M
RR = 2.1 (0.9–5.00) F
|Lynge et al.*||1988||12148 (M) |
|39C (M) |
|O||RR = 1.38 (1.0–1.9) CC |
RR = 0.96 (0.62–1.42) R
RR = 1.73 (1.1–2.7)
RR = 0.61 (0.17–1.57) R
|Severson et al.†||1989||8006 (M)||192C (M) |
|RR = 0.71 (0.51-.099); P for trend 0.201 |
RR = 0.66 (0.49–0.88); P for trend 0.136
|Ballard-Barbash et al.*||1990||1906 (M) |
|73CR (M) |
|O & L||RR = 1.8 (1.0–3.2) M; P for trend 0.06 |
RR = 1.1 (0.6–1.8) F; P for trend 0.89
|Lee et al.†||1991||17148 (M)||225C (M) |
|L||RR = 0.50 (0.27–0.93) |
RR = 1.72 (0.38–7.71)
|Chow et al.*||1993||1500000 (M) |
|1291C (M) |
|O||SIR = 121(108–135) M; P for trend = 0.001 SIR = 99(83–118)|
|Chow et al.*||1994|| ||13940C (M) |
4892 C (F)
|O||SIR = 1.2 (P < 0. 01) |
SIR = 1.1 (P < 0.5)
|Bostick et al.†||1994||35215 (F)||212C (F)||L||RR = 0.95 (0.7–1.39)|
|Giovannucci et al.†||1995||47723 (M)||203C (M)||L||RR = 0.53 (0.3–0.9); P for trend 0.03|
|Steeland et al.*||1995||14407||94CR (M) |
|O||RR = 1.02 (0.48–2.15) |
RR = 0.94 (0.4–2.21)
|Thune et al.†||1996||53242 (M) |
|236C (M) |
|O & L |
O & L
|RR = 0.97 (0.63–1.50)C; P for trend 0.49 |
RR = 1.20 (0.72–2.02)R; P for trend 0.63
RR = 0.63 (0.39–1.04)C; P for trend 0.04
RR = 1.27 (0.59–2.72)R; P for trend 0.45
|Lee et al.†||1997||21 807 (M)||217C (M)||L||RR = 1.1 (.7–16); P for trend 0.6|
|Martinez et al.†||1997|| ||763CR (F)||L||RR = 0.56 (0.36–0.89); P for trend 0.01|
|Hsing et al.*||1998||17633 (M)||125CR (M)||O||RR = 1.4 (0.9–2.3)|
|Will et al.†||1998||349631 (M) |
|3218CR (M) |
|O & L||RR = 0.74 (0.64–0.86) |
RR = 0.90 (.77–1.06)
|Colbert et al.†||2001||29133 (M)||152C (M) |
|RR = 0.45 (.26-.78) C; P for trend = 0.003 |
RR = 0.5 (0.26–0.97) R; P for trend = 0.04
RR = 0.82 (.59–1.13) C
RR = 0.93 (0.63–1.37) R
Table 2. Case-control studies.
|Garabrant et al.*||1984||2950 C (M) |
1213 R (M)
|31724 (M)||O||RR = 1.6 (1.3–1.8) C; P < 0.0001 |
RR = 0.9 (0.7–1.1) R
|Vena et al.*||1985||210 C (M) |
276 R (M)
|1431 (M)||O||OR = 1.97 C |
OR = 1.13 R
|Vena et al.†||1987||64589 C (M) |
604 C (F)
|430000 (M) |
|O||PMR = 0.9 M |
PMR = 0.8 F
|Slattery et al.†||1988||110 C (M) |
119 C (F)
|180 (M) |
|O & L||RR = 0.7 (0.38–1.29) M |
RR = 0.48 (0.27–0.87) F
|Brownson et al.*||1989||1211 C (M)||5693 (M)||O||RR = 1.4 (1.0–1); P for trend = 0.02|
|Fredeiksson et al.†||1989||156 C (M) |
|306 (M) |
|O||OR = 0.82 M (χ2 = 27.3) |
OR = 0.68 F (χ2 = 141.3)
OR = 0.49 (0.25–0.93) for left colon
|Peters et al.*||1989||147 CR (M)||147 (M)||O||OR = 3.0 (1.2–7.2); P < 0.05 in transverse colon|
|Benito et al.†||1990||151 CR (M) |
135 CR (F)
|261 (M) |
|O||RR = 0.5 P < 0.01|
|Gerhardsson et al.*||1990||163 C (M) |
107 R (M)
189 C (F)
110 R (F)
|512 (M & F)||O & L||RR = 3.3 (1.0–10.9) M |
RR = 0.9 (0.3–2.4) M
RR = 4.2 (1.2–14.0) F
RR = 1.4 (0.5–1.3) F
|Kato et al.*||1990||1716 C (M) |
1611 R (M)
|16600 (M)||O||RR = 1.52(1.19–1.94) C |
RR = 1.38(1.17–1.62) R
|Kato et al.†||1990||79 C (M) |
60 R (M)
53 C (F)
31 R (F)
|377 (M) |
|RR = 0.51(0.3–0.87) C |
RR = 0.7(0.36–1.38) R
RR = 0.55(0.33–0.89) C
RR = 0.54(0.30–0.97) R; Gender adjusted
|Kune et al.†||1990||388 CR (M) |
327 CR (F)
|398 (M) |
|O & L||RR = 1.50(.8–2.7) M |
RR = 0.89(.3–2.8) F
|Whittemore et al.*||1990||61 C (M)||255 (M)||O||OR = 2.5 (1.1–5.9) C; P < 0.01|
|Chinese in America|| ||50 R (M) |
42 R (F)
|198 (F)||L |
|OR = 1.6 (0.55–4.7) R |
OR = 1.6 (1.1–2.4) C
OR = 1.5 (0.93–2.5) R
OR = 1.2 (0.43–3.2) C
OR = 0.84 (0.32–2.2) R
OR = 2.0 (1.2–3.3) C; P < 0.01
OR = 1.9 (1.0−3.6) R
|Chinese in China|| ||95 C (M) |
131 R (M)
78 C (F)
128 R (F)
|678 (M) |
|OR = 1.4 (0.6–3.5) C |
OR = 0.86 (0.40–1.8) R
OR = 0.85 (0.39–1.9) C
OR = 0.71 (0.32–1.6) R
OR = 1.7 (0.56–5.2) C
OR = 0.58 (0.20–1.7) R
OR = 2.5 (1.0–6.3) C
OR = 0.69 (0.34–1.4) R
|Slattery et al.†||1990||290 C (M) |
|O & L||RR = 0.7(0.4–1.4) M |
RR = 0.5(0.3–0.9) F
|Brownson et al.*||1991||1838 (M)||1539 (M)||O||OR = 1.2(1.0–1.5); Trend P = 0.05|
|Markowitz et al.†||1992||307 C (M) |
133 R (M)
|1164 (M)||O |
|RR = 0.5 (0.3–0.8) C |
RR = 0.6 (0.3–1.1) R
RR = 0.4 (0.1–1.3)
|Thun et al.†||1992||611C (M) |
539 C (F)
|5376 (M & F)||O & L||RR = 0.6 (0.3–1.3) M |
RR = 0.9 (0.4–2.0) F
|Arbman et al.†||1993||51 C (M) |
48 R (M)
47 C (F)
31 R (F)
|512 (M) |
|O||OR = 0.4 (0.1–0.9) C |
OR = 1.7 (1.0–3.0) R
|Dosemeci et al.*||1993||93 C (M)||486 (M)||O||OR = 1.5 (0.8–2.0) C; P trend = 0.03|
|120 R (M)|| || ||OR = 1.3 (0.8–2.3) R; P = 0.21|
|Fraser et al.*||1993||1651 C (M) |
1046 R (M)
|Males aged 15–64 |
years in New
|O||RR = 1.45 (1.12–1.88) |
RR = 1.47 (1.04–2.08)
|Marcus et al.†||1994||536 (F)||2315 (F)||O & L||OR = 1.02 (0.82–1.27); P for trend 0.84|
|Longnecker et al.†||1995||163 C (M) |
242 R (M)
|703 (M)||O |
|RR = 0.7 (0.3–1.5) C; P for trend 0.38 |
RR = 1.09 (0.57–2.05) R; P for trend = 0.26
RR = 0.60 (0.3–1.0) C; P for trend = 0.03
RR = 1.18 (0.78–1.80) R; P for trend = 0.47
|White et al.†||1996||251 C (M) |
193 C (F)
|233 (M) |
|RR = 0.66 (0.41–1.05) M; P for trend 0.07 |
RR = 0.74 (042–1.29) F; P for trend 0.37
Occupational activity not related to colon cancer
except in males < 55 years RR = 0.29 (0.12–0.69)
|Lemarchand et al.†||1997||698 C (M) |
494 C (F)
|698 (M) |
|L||RR = 0.6 (0.4–0.8) M; P for trend 0.007 |
RR = 0.7 (0.5–1.1) F; P for trend 0.32
|Slattery et al.*||1997||1099 C (M) |
889 C (F)
|1290 (M) |
|L||OR = 1.63 (1.26–2.12) |
OR = 1.59 (1.21–2.10)
|Tang et al.†||1999||92CR (M) |
71 CR (F)
|92 (M) |
|RR = 0.19 (0.05–0.77) M; P for trend = 0.03 |
RR = 0.77 (0.31–1.70) F; P for trend = 0.48
Effect of occupational activity not significant
|Tavani et al.†||1999||688 C (M) |
537 C (F)
286 R (F)
|2073 (M) |
|RR = 0.64 (0.44–0.93) C R |
RR = 1.32 (0.86–2.03) R
RR = 0.49 (0.33–0.72) C
RR = 0.88 (0.48–1.60)
Effect of leisure activity NS
|Steindorf et al.†||2000||95CR (M) |
|95 (M) |
|RR = 0.61 (0.29–1.2) |
RR = 0.45 (0.24–0.84); Gender adjusted
Analyses were undertaken for cohort studies in Table 3. A statistically significant reduction in the risk of colon cancer was seen in physically active males, irrespective of the type of activity measured, RR being 0.79 (95% CI 0.72–0.87) and 0.78 (95% CI 0.68–0.91) for occupational and recreational activities, respectively. Women who are exposed to high levels of recreational activities are protected against colon cancer (RR = 0.71, 95% CI 0.57–0.88). However in women high levels of occupational activity failed to offer any protection against colon cancer (Table 4).
Table 3. Pooled estimates of cohort studies with colon and rectal cancer in males. Both random and fixed analyses is given in groups where there is significant heterogeneity between studies.
Table 4. Pooled estimates of cohort studies with colon and rectal cancer in females. Both random and fixed analyses is given in groups where there is significant heterogeneity between studies.
Tables 5 and 6 show similar analyses with case-control studies. Significantly reduced risks are seen in both males and females, with all three types of activities measured and in both forms of analyses. The analyses failed to show any protection against rectal cancer in both males and females.
Table 5. Pooled estimate of case-control studies with colon and rectal cancer in males. Both random and fixed analyses is given in groups where there is significant heterogeneity between studies.
|Case Control||Occupational |
Table 6. Pooled estimate of case-control studies with colon and rectal cancer in females. Both random and fixed analyses is given in groups where there is significant heterogeneity between studies.
|Case Control||Occupational |
Several biological mechanisms have been hypothesized to explain this association between physical activity and colorectal cancer. Possible mechanisms are hyperinsulinemia, obesity, decreased gut transit time, change in prostaglandin ratio, lowered bile acid secretion and altered gut flora. Furthermore physical activity may affect cancer risk indirectly through other correlated or confounding factors like diet, smoking, alcohol consumption and life style habits.
The insulin-colon cancer hypothesis
There is increasing evidence that insulin is a tumour growth factor. McKeown-Eyssen and Giovannucci have proposed the insulin-colon cancer hypothesis, in which they suggest that insulin resistance leads to colorectal cancer through growth promoting effect of insulin, glucose or triglycerides [58,59]. Giovannucci pointed out that central obesity and physical inactivity, which are major determinants of insulin resistance and hyperinsulinaemia, appear to be related to colon cancer risk. Insulin is an important growth factor for colonic mucosal cells and is mitogenic to colonic carcinoma cell groups in vitro[60–62]. Furthermore hyperinsulinaemia can reduce apoptosis, an attribute favouring tumour development. . Experimental studies have shown an increase in the number of aberrant crypt foci  and colon cancers  when insulin is given to carcinogen treated rats.
A consequence of insulin resistance is that higher levels of insulin are needed to normalize plasma glucose resulting in increased circulating levels of insulin, glucose, triglycerides and nonesterified fatty acids (NEFA). These elevated levels of hormones and energy substrates have effects on cell cycle control, cell survival and cell mutations that affect colon carcinogenesis.
Colon cancer tissue has both insulin and IGF-1 receptors [66,67] and elevated levels of these molecules may have mitogenic properties. Recent studies have shown an association between plasma IGF-I and IGF-binding protein (IGF-BP) levels with risk of colon cancer [68,69]. Interestingly there is an increased risk of colonic malignancy in acromegalic individuals, in whom there is chronic growth hormone and IGF-1 hypersecretion .
Increased concentration of circulating energy substrates, triglycerides and NEFA together with metabolic effects of insulin lead to increase intracellular energy substrates affecting important cell signal pathways. This may stimulate cell growth  increasing the risk of carcinogenesis. Evidence supporting this hypothesis is that mitogenic signal transduction induced by insulin involves ras, a proto-oncogene in colon carcinogenesis. Increased energy availability could also affect colon carcinogenesis through the formation of reactive oxygen intermediates (ROI) activating mitogen activated protein (MAP) kinase and increase the expression of oncogenes such as c-fos and c-jun. There is evidence that high levels of insulin leading to accumulation of cellular energy will result in increased utilization of antioxidants , which are protective against carcinogenesis. The protective effect of physical activity may be because physical activity directly increases insulin sensitivity  and reduces plasma insulin levels .
Caan et al.  has shown that body mass index (BMI) was consistently associated with an increased risk of colon cancer. However Slattery et al.  have found that physical activity is independently related to colon cancer risk after controlling BMI and this may be due to reduction in the highly metabolic visceral (intra-abdominal) fat as changes in visceral fat may have significant effects on glucose tolerance and insulin and lipid levels .
Individuals with colorectal polyps or cancer have higher levels of colonic mucosal prostaglandin E2 (PG E2) than control individuals . This has been the basis of the use of aspirin in reducing the risk of colorectal cancer and adenoma [80,81]. High levels of physical activity have an inverse association, lowering PG E2 and this action is possibly mediated through insulin or IGF-1 . Furthermore physical activity produces an increase in prostaglandin F2α that increases gut motility possibly reducing the exposure of colonic mucosa to carcinogens.
It has been postulated that regular exercise increases number and activity of macrophages, natural killer and lymphokine-activated killer cells and cytokines [83,84] which destroy cancer cells.
Gut transit time
Physical activity decreases the gut transit time, possibly through increased vagal tone and subsequent increased peristalsis [85,86]. It is suggested this will reduce the contact time of colonic mucosa with the carcinogens like bile acids [87,88] and may explain the protective effect of exercise on colorectal cancer.
The high incidence of colon cancer in affluent societies has often been attributed to a high fat and low fibre diet and individuals who exercise regularly may have diets low fat and high fibre content [89,90]. However studies determining associations between cancer and exercise, which have been able to control for diet have shown an independent effect of physical activity and colon cancer [6,91].
This study shows a statistically significant protective effect against colon cancer among physically active males and females. However the analysis shows that (i) exercise exerts no protective effect against rectal cancer; (ii) the results for males and females are very similar, except for cohort studies of occupational exercise.
Of the different mechanisms by which exercise exerts its protective effect against colorectal cancer, the insulin-colon cancer hypothesis has generated the most interest. It is evident from the epidemiological studies that there is a close association between insulin resistance and risk of colon carcinogenesis. The hypothesis associating insulin resistance and hyperinsulinemia with colorectal cancer risk provides a unifying mechanism by which physical activity, dietary and other life style factors have a causal effect on colorectal cancer. It is now possible to quantify end point biomarkers for colon cancer like ACF in the human colon and readily measure putative risk factors like intravascular energy substrates, insulin and IGF-I. Future studies could be focused on the effect of exercise on these biomarkers.
No randomised control trials of physical activity as a means for the primary prevention of colon cancer have been published. Cancer incidence cannot be used as primary end point because it would mean a lengthy follow up, however, intermediate end points that are biological markers of the disease may be more feasible. McTiernan et al.  recommend that a series of small clinical trials of exercise interventions be conducted to measure exercise change effects on biomarkers for cancer risk, which serve as feasibility studies for larger randomised controlled trials of cancer and precursor end points and for community intervention studies. One such trial known as the Colon Polyp and exercise study is already under way, in which, 200 men and women of ages 40–75 years, who are sedentary and who have a diagnosis of one or more adenomatous polyps in the past 18 months, are randomised to either a moderate exercise intervention or the stretching control group . The aim is perform colon and rectal biopsies and to measure numerous biomarkers including apoptosis related proteins, rectal prostaglandin levels, and serum glucose/insulin and insulin growth factor profiles. The results of these trials are eagerly awaited.