OBJECTIVETo evaluate the association and its strength between body mass index (BMI, kg/m(2)) and cancer incidence in a large-scale population-based cohort study.

METHODSA dynamic cohort was established on May 1, 2006. Baseline information on demography, lifestyle, anthropometry such as body height and weight, were collected during the first interview, and cancer incidence, mortality and other related outcome information were obtained through active follow-up every two years and passive follow-up every year. Cancer cases diagnosed within 1 year follow-up period were excluded. Multivariable Cox proportional-hazards regression model was used to calculate the hazard ratios (HRs)and 95% confidence interval (CI) between BMI and cancer incidence after adjusted for age group, education level, tobacco smoking (smokers and non-smokers), alcohol consumption (drinkers or non-drinkers) and HBsAg status (positive or negative, for liver cancer only) when appropriate. Repeated analysis was carried out on male lung cancer, male liver cancer and female breast cancer, stratified by tobacco smoking, HBsAg status and menopausal status respectively.

RESULTSBy December 31, 2011, a total of 133 273 subjects, including 106 630 (80.01%)males and 26 643(19.99%)females were enrolled in the cohort. There were 570 531.02 person-years of follow-up and 4.28-year of average follow-up period. According to the Guidelines for Prevention and Control of Overweight and Obesity in Chinese Adults, study subjects were divided into groups as: underweight (BMI<18.5 kg/m(2)), normal weight (BMI 18.5 kg/m(2)-23.9 kg/m(2)), overweight (BMI 24.0 kg/m(2)-27.9 kg/m(2)) and obese(BMI≥28.0 kg/m(2)). In males, 2 387 (2.24%) were underweight, 45 090(42.29%)were normal weight, 43 774 (41.05%) were overweight and 15 379 (14.42%) were obese. Meanwhile, in females, 858 (3.22%)were underweight, 14 037 (52.69%) were normal weight, 8 507 (31.93%) were overweight and 3 241 (12.16%) were obese. A total of 1 647 incident cancer cases among different cancers were collected during the follow-up, including 1 348 in men and 299 in women. Results from Multivariate Cox proportional-hazards regression model showed that 'underweight' increased the risk on both gastric cancer incidence (adjusted HR = 3.82, 95% CI: 1.97-7.38) and liver cancer incidence (adjusted HR = 3.00, 95% CI:1.36-6.65) in males, while both 'overweight' (adjusted HR = 1.98, 95% CI:1.03-3.82) and 'obesity' (adjusted HR = 2.75, 95% CI: 1.25-6.06) increased the risk of colon cancer incidence in males. But for bladder cancer incidence in males, overweight seemed being protective (adjusted HR = 0.44, 95%CI:0.23-0.84). And for lung cancer incidence in males, both overweight and obesity were protective as well (adjusted overweight vs. normal weight, HR = 0.59, 95%CI: 0.46-0.76;adjusted obese vs. normal weight, HR = 0.64, 95%CI: 0.44-0.92). In females, obesity increased the risk of breast cancer incidence(adjusted HR = 1.86, 95%CI:1.05-3.31). Further analysis for lung cancer cases stratified by tobacco smoking, data showed that overweight decreased the risk of lung cancer in both male non-smokers (adjusted HR = 0.50, 95%CI:0.35-0.72) and male smokers (adjusted HR = 0.70, 95%CI:0.50-0.98) while obesity decreased the risk of lung cancer in male non-smokers(adjusted HR = 0.57, 95% CI:0.33-0.97), but not in smokers (adjusted HR = 0.72, 95%CI:0.43-1.21). Stratified analysis in females by menopausal status, data showed that obesity increased the incidence of breast cancer in postmenopausal subjects (adjusted HR = 1.97, 95% CI:1.01-3.82), but not in premenopausal subjects.

CONCLUSIONThe association between BMI and cancer incidence varied by cancer site. Underweight increased the risk of gastric cancer and liver cancer in males, and obesity increased the risk of colon cancer in males, breast cancer and ovarian cancer in females. However, overweight might played a protective role in lung cancer incidence and bladder cancer incidence in males and obesity might play a protective role in lung cancer incidence in male non-smokers.