Abstract
ObjectiveTo evaluate the magnitude of the potential underestimation of the proportion of cancer cases attributable to excess weight, known as population attributable fraction (PAF), due to potential bias from prediagnostic weight loss already present at baseline of cohort studies and to overcome it as much as possible.MethodsData from the UK Biobank cohort participants aged 40-69 without prior cancer diagnosis were analyzed. We assessed the magnitude of associations of excess weight with the incidence of obesity-related cancers combined, and separately for gastrointestinal (GI) and other cancers. Using multivariable Cox proportional hazards models, hazard ratios (HR) and their 95% confidence intervals (CI), and PAFs for excess weight at baseline were estimated for various periods of time after weight measurements.FindingsOf 458,660 participants, 20,218 individuals developed obesity-related cancers during a median 11.0-year follow-up, comprising 8,460 GI, and 11,765 non-GI cancers. PAFs were much higher for cancers occurring more than four years after recruitment than for cancers occurring within the initial four years: 17.7% versus 7.2%, 21.4% versus 11.7% for GI, non-GI and all obesity-related cancers combined, respectively. With respect to total cancer (including cancers with no established relationship with excess weight), PAFs were estimated as 5.1% and 8.8% for the 0-4 and 4-14-year periods of follow-up.ConclusionThe proportion of cancers attributable to excess weight is likely substantially larger than previously estimated based on cohort studies with short follow-up time or no or only limited exclusion of the early years of follow-up from the analyses.