High body mass index and cancer risk-a Mendelian randomisation study

High body mass index (BMI) has been associated with increased risk of some cancer. Whether these reflect causal associations is unknown. We examined this issue. Using a Mendelian randomisation approach, we studied 108,812 individuals from the general population. During a median of 4.7 years of follow-up (range 0-37), 8002 developed non-skin cancer, 3347 non-melanoma skin cancer, 1396 lung cancer, 637 other smoking related cancers, 1203 colon cancer, 159 kidney cancer, 1402 breast cancer, 1062 prostate cancer, and 2804 other cancers. Participants were genotyped for five genetic variants associated with BMI. Two Danish general population studies, the Copenhagen General Population and the Copenhagen City Heart Study. In observational analyses, overall risk of non-melanoma skin cancer was 35 % (95 % confidence interval 28-42 %) lower and risk of lung cancer 32 % (19-43 %) lower in individuals with a BMI ≥ 30 versus 18.5-24.9 kg/m(2). Corresponding risk of breast cancer was 20 % (0-44 %) higher in postmenopausal women. BMI was not associated with risk of colon, kidney, other smoking related cancers, prostate cancer, or other cancers. In genetic analyses, carrying 7-10 versus 0-4 BMI increasing alleles was associated with a 3 % higher BMI (P < 0.001), but not with risk of cancer. In instrumental variable analysis for a 10 kg/m(2) higher genetically determined BMI the odds ratio for any non-skin cancer was 1.16 (0.64-2.09), with a corresponding observational estimate of 0.94 (0.88-1.01). Using 108,812 individuals from the general population, we found that observationally high BMI was associated with lower risk of lung and skin cancer overall and with higher risk of breast cancer in postmenopausal women, but not with other types of cancer. BMI increasing alleles were not associated with risk of cancer, and results do not support causal associations. Power to test associations for some cancer sites was low.