Commentary: How obesity and physical activity contribute to colorectal cancer

In this issue of the journal Harriss and colleagues report a thorough meta-analysis of the association between excess body weight (expressed as BMI), leisure-time physical activity (LT-PA) and colorectal cancer in epidemiological studies. Meta-analysis is a popular means to summarise quantitatively evidence accumulating in repeated trials, that is to say in independent studies. The tool has been developed to pool results from randomised controlled trials where treatment is controlled by the investigator and randomisation minimises bias and confounding. In epidemiological studies that are typically non-experimental, the concern is with systematic sources of heterogeneity rather than random variation. These include primarily biases linked to study design and measurement error in the exposure and outcome variables. Meta-analyses of observational studies try to establish whether there is a genuine association and describe it through dose-response relationships and interactions with other factors, using information from studies that can be considered, under stated assumptions, independent assessments of the same phenomenon. Selection criteria allow one to exclude major causes of heterogeneity that cannot be controlled in the analysis. For this reason Harriss and colleagues used only longitudinal observations where current weight and height and leisure time exercise were assessed before the diagnosis of cancer and did not rely on retrospective recall of past habits. Indeed the 2nd WCRF report on nutrition and cancer (2007) showed through meta-analyses conflicting results for some diet-cancer associations when assessed in case-control compared with prospective studies. They also considered only studies that had incident cases as endpoint since obesity and sedentary life are associated with poorer survival in colorectal cancer cases leading to increased mortality. Pooling data from many studies increases the power of the analyses and, potentially, permits the detailed examination of interactions and reciprocal confounding that can help to clarify the role of individual factors in an exposure-outcome association. In our context, one important issue is, for example, whether physical activity (PA) protects from colorectal cancer independently of energy balance or whether that association is a component of the latter. In the first case one would expect to observe similar effects of PA within subgroups of lean and overweight people, while little residual effect of PA would be left once body fatness (indicating prolonged lack of energy balance) is accounted for. Several other complex associations need to be clarified if we want to reconcile epidemiological evidence with mechanistic hypotheses of colorectal carcinogenesis and ultimately identify the key factors to target with interventions. In their analyses Harriss and colleagues address specific questions, namely a possible different effect of excess body weight (measured by the body-mass index [BMI]) and LT-PA in men and women which would corroborate a role of sex hormones, the consistency of the associations by bowel sub-site, the strength of the associations in a dose-response fashion and the role of confounders. They found that over-weight was a stronger risk factor in men than in women for both colon and rectal cancers and, among men, excess body weight increased the risk of developing cancer in the colon more than in the rectum. In women the weaker relative risks were not significantly heterogeneous by sub-site. The summary relative risks associated with BMI were slightly increased if adjusted for PA while the protective effect of PA was reduced if adjusted for BMI, at least in men. Moreover, PA reduced the risk of colon but not rectal cancer with no evidence of heterogeneity by sex. Confounding/interaction analyses confirmed small effects on the main associations by adjustment for family history, smoking and alcohol drinking. Indicators of measurement accuracy suggested a weaker association for BMI (measured vs. self-reported weight and height), but stronger for LT-PA (validated vs. non-validated questionnaire). What use can we make of these results? On the side of public health they confirm the recommendation to maintain lean body weight, better if with the help of physical exercise. Even if we do not understand how these two factors affect the risk of cancer, the benefits are numerous and well documented in reducing cardiovascular diseases, diabetes, all-causes mortality including cancer and in increasing life expectancy [1,18]. On the side of understanding the causes of bowl cancer these summary measures offer an opportunity to reflect on how wide is the gap between mechanistic biological hypotheses that try to explain the associations, and the evidence that these are weak. Even interactions with and confounding by other risk factors are weak. Measurement errors may well cause an observed weak association for LT-PA, but it is less the case for BMI. A direct role of adipose tissue is therefore unlikely; rather, obesity could be simply a marker of prolonged excess of energy intake, or could have a role as a co-factor that becomes relevant only under certain conditions. Efforts should therefore concentrate on identifying those conditions, keeping in the picture only associations that are highly consistent or strong, if any. Most important is to challenge interpretative hypothesis with evidence that does not fit in order to re-define them or look elsewhere. Several hypotheses provide an interpretation for the fatness-cancer relationship. The hyper-insulinaemia hypothesis states that chronic elevated insulin promotes the proliferation of colonic cells including cancerous clones. The effect can be direct, since insulin has that potential in vitro, or indirect by increasing circulating levels of free (bio-available) insulin-like growth factors (IGF). The same mechanism could explain the protective effect of physical activity which reduces circulating levels of insulin. Animal experiments strongly corroborate the hypothesis [19]. Notably, the mechanism is that of a classical promoter that does not cause cancer in the absence of a carcinogen (in epidemiological terms, a co-factor). Epidemiological observations that support the mechanism include a small excess risk of colorectal cancer in diabetic patients [14] particularly in those who are insulin-dependent [25]. It would also fit with increasing incidence rates of the disease paralleling increasing prevalence of over-nutrition and consequent overweight [2]. But there are examples suggesting that other factors are more relevant. In Mumbai, India, for example, time trends of colorectal cancer incidence rates have been remarkably stable and are among the lowest worldwide, despite a rapidly growing middle class among which type 2 diabetes and obesity have become common [3]. Conversely, incidence rates have declined in the US and Canada since the mid-1980s well before screening was widely available [2]. The confirmation that the association between BMI and colorectal cancer is stronger and more consistent in men than in women is unexplained. An interaction between glucose metabolism and circulating oestrogen levels seems plausible but is not supported by evidence. Research on sex steroid hormones and colorectal cancer explains the role of obesity as a source of oestrogens produced in fat tissue from androgens conversion by the aromatase enzyme. Interactions between insulin and estrogens have not been thoroughly studied. The role of oestrogens in colorectal cancer remains obscure: observational studies suggest a protective role, possibly mediated by oestrogen receptors [26]. In the Women Health Initiative trials however, a short-term reduction of colorectal cancer incidence with use of combined oestrogen plus progestagen as menopause replacement therapy (HT) was not confirmed on a longer follow-up, and oestrogen-only therapy was not associated with the disease [4,17,20]. Nonetheless, independent lines of research suggest that a reduction of circulating oestrogens by aromatase inhibitors could control colonic tumour growth [12]. Once more we face inconsistency between mechanisms that are plausible at the tissue level and in animal models and epidemiological data. Why do we need to fill that gap? Because decades of attempts to single out nutritional and biochemical factors that predict risk have produced numerous risk factors but no convinceing explanation of the key events that lead to cancer. Epidemiological studies of colorectal cancer involving thousands of individuals are admittedly unsuitable to control tens of parameters often correlated. The design and analysis need to be driven by mechanistic hypotheses, which can be developed in animal models. New laboratory experiments should combine initiating and promoting mechanisms that mirror situations naturally occurring in humans. Along this line of thinking it would be particularly interesting to clarify whether and to what extent insulin, caloric restriction and treatment with metformin, a glucose-lowering drug widely used to treat type 2 diabetes, modulate innate immune responses to pathogenic entero-hepatic bacterial infections, e.g. Helicobacter species that colonise the gut [6,15]. Chronic elevated glycaemia could create the condition for normal commensal intestinal bacteria to overcome protective host response and cause pathologic effects locally [7,13]. Importantly, the immune response to pathogenic bacteria is strongly involved in cholelitiasis in mouse models [16], and cholelithiasis and gallbladder disease leading to cholecystectomy are disorders more common in women than men, are associated with excess body weight and with a modest increased in the risk of colon cancer [8,22,23]. There is a growing body of evidence involving inflammatory cytokines, particularly interleukins, in metabolic disorders associated with obesity [5] but it is possible that the relevant source of inflammation is other than the low-grade systemic state observed in obese patients; after all, most patients with colorectal cancer are not obese. Moreover, T-lymphocyte and NK cell infiltration is a common feature of solid tumours including the colon, and certain components of the haematopoietic system have been shown to be essential for polyp development [9]. Having recognised the promoting effect of excess body weight, efforts should concentrate on identifying relatively common conditions that have the potential to induce mutations in the colonic mucosal while at the same time, taking advantage of chronic elevated glycaemia.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.