Body-mass index (BMI) is a simple and commonly used measurement in clinical medicine and population health—a ratio of weight (kilogrammes) and height (metres squared). Adults with a BMI of 25—29·9 kg/m2 are considered overweight and adults with a BMI of 30 kg/m2 or higher are considered obese. The causes and consequences of overweight and obesity, however, are anything but simple. Determinants of BMI include genetic and epigenetic factors; individual behaviours (eg, physical activity, sedentary time, and caloric intake); sociocultural factors; and the physical, economic, and policy environments. Excess adipose tissue is not a benign storage depot for lipids.1 Indeed, overweight and obesity were estimated to cause 3·4 million deaths worldwide in 2010.2 The proportion of adults who were overweight or obese worldwide increased markedly from 1980 to 2013, with parallel increases in children and adolescents.3 Although the causes of these disturbing trends are multifactorial, they almost certainly include the wide propagation of cheap, calorically dense, and nutritionally poor food and drink.4, 5 The well-established associations between high BMI and increased morbidity include coronary heart disease, type 2 diabetes, high blood pressure, stroke, dyslipidaemia, and osteoarthritis. Studies published during the last two decades have added certain cancers to this list.6, 7

In The Lancet, Krishnan Bhaskaran and colleagues8 add further supportive evidence that high BMI is associated with risk of many specific types of cancer. Although numerous studies have examined associations between BMI and risk of individual cancers, this broad and comprehensive study illustrates the full cancer incidence burden that is imposed by high BMI. Using data from the UK's Clinical Practice Research Datalink (CPRD), the authors identified 5·24 million persons, from an initial cohort of 10 million, with sufficient BMI and follow-up time for analysis. All 5·24 million individuals were thought to be cancer-free at the onset of analysis. After a mean observation period of 7·5 years, over 166 000 people developed one of the 22 cancers of interest to the authors, which represent about 90% of all cancers diagnosed in the UK. Each 5 kg/m2 increase in BMI was associated with higher risks of cancers of the uterus (hazard ratio [HR] 1·62, 99% CI 1·56—1·69), gallbladder (1·31, 1·12—1·52), kidney (1·25, 1·17—1·33), liver (1·19, 1·12—1·27), colon (1·10, 1·07—1·13), cervix (1·10, 1·03—1·17), thyroid (1·09, 1·00—1·19), ovary (1·09, 1·04—1·14), postmenopausal breast (1·05, 1·03—1·07), pancreas (1·05, 1·00—1·10), and rectum (1·04, 1·00—1·08), and of leukaemia (1·09, 1·05—1·13). More detailed analyses revealed associations between BMI and cancers of the oesophagus, pancreas, and stomach in non-smokers; associations with liver, colon, ovarian, and breast cancers were modified by other factors (eg, sex and menopausal status). Higher BMI was associated with lower risk of prostate and premenopausal breast cancers. Inverse associations between BMI and cancers of the oral cavity and lung were null after restricting analyses to never smokers, suggesting confounding by smoking dose.

Importantly, beyond measuring relative risks, the authors assessed measures of population effect. Of the ten cancer sites associated with BMI in categorical models, overweight and obese BMI was estimated to account for between 2% (thyroid) and 41% (uterus) of these cancers in the UK. Additionally, the authors estimate that a population-wide 1 kg/m2 increase in BMI would cause an additional 3790 cancer diagnoses in the UK each year.

Confidence in these results is bolstered by the authors' many sensitivity and model checking exercises. The message from these additional analyses is that the main study results were not affected by obvious analytical decisions by the authors or confounding. Confidence in the study is further strengthened by the fact that nearly all of these findings have been reported in previous studies; a notable exception includes the study's finding of higher risk of cervical cancer with high BMI.

There are a few caveats to Bhaskaran and colleagues' study that should be mentioned. Selection bias might be a strong threat to validity in this study; in particular, only slightly more than 50% of the initial cohort had sufficient BMI data and follow-up time for analysis. If BMI is more likely to be measured in people with comorbidities or medical conditions that are also associated with cancer risk (eg, type 2 diabetes or gallstones) then some bias might be present. The lack of histological, molecular phenotype, and cancer sub-site specific data for many of the cancers in this study is a serious limitation, as acknowledged by the authors. This is especially true when interpreting the results for oesophageal cancer where a clear link is established between BMI and oesophageal adenocarcinomas, and a possible inverse association has been shown for oesophageal squamous cell carcinomas.9 The authors included in this analysis persons with exceptionally low BMI values (<15 kg/m2), which sometimes hinders interpretation of the data because exceedingly low BMI values probably represent data recording errors or extreme illness.

We have sufficient evidence that obesity is an important cause of unnecessary suffering and death from many forms of cancer,6, 7, 10 in addition to the well recognised increased risks of mortality and morbidity from many other causes. More research is not needed to justify, or even demand, policy changes aimed at curbing overweight and obesity. Some of these policy strategies have been enumerated recently,11 all of which focus on reducing caloric intake or increasing physical activity, and include taxes on calorically dense, nutritionally sparse foods (eg, sugar-sweetened beverages); subsidies for healthier foods, especially in economically disadvantaged groups; agricultural policy changes; and urban planning aimed at encouraging walking and other modes of physical activity. Research strategies that identify population-wide or community-based interventions and policies that effectively reduce overweight and obesity should be particularly encouraged and supported. Moreover, we need a political environment, and politicians with sufficient courage, to implement such policies effectively.

I declare no competing interests.