• Andrew G Renehan, Isabelle Soerjomataram, and Michael F Leitzmann.
• Department of Surgery, The Christie NHS Foundation Trust, School of Cancer and Enabling Sciences, University of Manchester, UK. arenehan@picr.man.ac.uk
• Eur. J. Cancer. 2010 Sep 1; 46 (14): 2581-92.

AbstractStandard approaches to estimating population-attributable risk (PAR) include modelling estimates of exposure prevalence and relative risk. Here, we examine the associations between body mass index (BMI) and cancer risk and how effect modifications of these associations impact on PAR estimates. In 2008, sex- and population-specific risk estimates were determined for associations with BMI in a standardised meta-analysis for 20 cancer types. Since then, refinements of these estimates have emerged: (i) absence of menopausal hormonal therapy (MHT) is associated with elevated BMI associations in post-menopausal breast, endometrial and ovarian cancers; (ii) current smoking attenuates the BMI associations in oesophageal squamous cell carcinoma, lung and pancreatic cancers; (iii) prostate screening attenuates BMI associations when all prostate cancers are considered together; and (iv) BMI is differentially associated with different histological subtypes within the same cancer group. Using secondary analyses of the aforementioned meta-analysis, we show 2-3-fold shifts in PAR estimations for breast and endometrial cancers depending on the MHT usage in European countries. We also critically examine how to best handle exposures (in this example, BMI distributions) and relative risk estimates in PAR models, and argue in favour of a counterfactual approach based around BMI means. From these observations, we develop a research framework in which to optimally evaluate future trends in numbers of new cancers attributable to excess BMI. Overall, this framework gives conservative estimates for PAR - nonetheless, the numbers of avoidable cancers across Europe through avoidance of excess weight are substantial.Copyright © 2010 Elsevier Ltd. All rights reserved.

22 keywords...

hide…