br M Touillaud et al site was obtained
M. Touillaud, et al. site was obtained from the national network of cancer registries (FRANCIM). Age-, sex- and site-specific incidence rates were computed and applied to the 2015 national French population to obtain an esti-mate of the national number of new cases in that year. 2.4. Statistical analysis The lag time between the cumulative insufficient physical activity and the occurrence of cancer is uncertain and may potentially vary by cancer site (AICR-WCRF 2007). In our analyses, we used prevalence data from 2006/2007 and cancer incidence estimates from 2015, thereby assuming a lag time of approximately 10 years. This is based on scientific evidence that insufficient physical activity is not an initiator of cancer but rather a promotor of cancer to clinical presentation over several years. To account for population ageing with time since ex-posure and lag time, we mapped prevalence data to the cancer in-cidence age group that was 10 years older (e.g., cancer incidence in the 40–49 age group in 2015 was attributed to insufficient physical activity in the 30–39 age group in 2006/7). The following formula was used to compute age-, sex- and cancer-specific population attributable fractions (PAFs):
the total colon cancer diagnoses in 2015).
These estimated PAFs are lower than those estimated in France for the year 2000, which were 4.1% (0.5% for men and 4.7% for women) . These results are not directly comparable because the 2000 esti-mates for France only included breast and colon cancers, applied dif-ferent RRs and used vigorous recreational physical activity as a re-ference. These PAFs are also lower than those estimated in the United Kingdom in 2010, where they were 1.0% (0.4% for men and 1.7% for women) , and those estimated in Australia being 1.6% (0.5% for men and 2.9% for women) . However, these results are not directly comparable because differences in the RRs used for PAF estimation and in the reference levels used to define the lack of physical activity (15 MET-hours per week in the UK study and 30 MET-hours per week in the Australian study). When we used 15 MET-hours per week as reference, the French estimates were still lower than the UK estimates (0.5% vs 1%, respectively).
The preventive effects of physical activity on cancer incidence might be due to several mechanisms. Direct effects of physical activity on circulating levels of various MDMB-CHMINACA and growth factors, including decreased plasma levels of insulin and IGF-1, which promote cell pro-liferation and increase with overweight and obesity, could be one plausible causal pathway [1,7]. Physical activity also indirectly con-tributes to reducing cancer risk by reducing the risk of overweight or
(3) obesity, by limiting body fat and by promoting lean body mass. It might specifically reduce colon cancer risk by accelerating gut transit, thus reducing the exposure time of the gastrointestinal mucosa to foodborne carcinogens [2,7]. Physical activity might reduce the risk of post-menopausal breast cancer and endometrial cancer notably by de-creasing oestrogen levels, stimulating immunity and decreasing in-flammation (increase in numbers and/or activity of macrophage and lymphocytes) . In their 2015 update of the evidence, the National cancer institute in France additionally rated the level of evidence between physical ac-tivity and lung and premenopausal breast cancers as probable . Si-milarly, a recent meta-analysis showed that leisure-time physical ac-tivity is associated with lower risks of 13 cancer sites . Yet, these sites were not included in the analysis due to the lack of robust RRs that match with the exposure (measured in MET/hours), a prerequisite to perform PAF calculations. Nationally representative data on physical activity were used in this study , together with robust estimates of the relationship between insufficient physical activity and cancer. As such, the results of this study give an indication of the potential impact of prevention measures on the reduction of the cancer burden. Even if half of the recommended minimum of physical activity (10.5 MET-hours per week) would be attained, more than 1000 cancer cases could be avoided each year.
We acknowledge that the present study has a number of limitations. First, it was affected by inherent limitations to measuring exposure. Physical activity data were obtained from a population survey and are therefore subject to several biases, such as coverage and participation bias, but also to recall error and social desirability bias [19,20]. In-formation on physical activity was available from close to 3000 parti-cipants, and although this number is relatively large, for the analysis this group was categorised into subgroups according to sex, age and physical activity. The number of persons in each category may thus be small with less stable estimates. On the other hand, the PAFs here are reported for the whole group, i.e. based on a more stable average. Furthermore, data on physical activity prevalence were obtained at a given time and may not reflect the level of physical activity throughout the relevant biological exposure period. Second, although RRs used in the PAF calculations were taken from published meta-analyses and were adjusted for a range of other related exposures, most importantly for body fatness, residual confounding cannot be ruled out. Moreover, while RR estimates are based on Caucasian populations, they might not be fully applicable to France, especially if the prevalence of effect modifiers differs between settings . Third, this analysis did not