Intended for healthcare professionals


Fresh evidence links adiposity with multiple cancers

BMJ 2017; 356 doi: (Published 28 February 2017) Cite this as: BMJ 2017;356:j908

Chinese translation


  1. Yikyung Park, associate professor,
  2. Graham A Colditz, professor
  1. Division of Public Health Sciences, Washington University School of Medicine, St Louis, MO 63110, USA
  1. Correspondence to: G A Colditz colditzg{at}

The association is now clear; it’s time to get serious about prevention

The International Agency for Research on Cancer (IARC) working group recently reviewed epidemiological data, studies in experimental animals, and mechanistic data and concluded that excess body fatness causes cancer of the colon and rectum, liver, gallbladder, pancreas, kidney, thyroid, breast (postmenopausal), endometrium, ovary, oesophagus (adenocarcinoma), and gastric cardia, as well as meningioma and multiple myeloma.1 This potentially makes excess body fat the second most important modifiable cancer risk factor after tobacco use.

The study by Kyrgiou and colleagues2 took up the challenge of evaluating the robustness of multiple, sometimes overlapping, meta-analyses that reported an association between body adiposity measures (such as body mass index, weight gain, and waist circumference) and cancer. The authors conducted an umbrella review, also known as a “review of reviews” or “meta-review,”34 and initially identified a total of 204 individual meta-analyses from 49 papers. They further examined the 95 meta-analyses that reported the association between body fatness measured on a continuous scale (mostly body mass index in 5 kg/m2 increase) and cancer in cohort studies. After a rigorous evaluation for strength and validity of reported associations, 13% (12 of 95) of meta-analyses were judged to provide strong evidence on the basis of their statistical criteria. The rest of the meta-analyses were deemed to be highly suggestive (18%), suggestive (25%), and weak (20%). Twenty four per cent of meta-analyses found no association between body fatness and cancer.

Nine obesity related cancers were supported by strong evidence: oesophageal adenocarcinoma, colon and rectal cancer (in men), biliary tract system, pancreatic, and kidney cancer, endometrial cancer (premenopausal women), breast cancer (postmenopausal), and multiple myeloma. A positive association between body mass index and liver, ovarian, or thyroid cancer was highly suggestive or suggestive; a negative association with oesophageal squamous cell carcinoma or lung cancer was highly suggestive. In additional analyses using obesity categories (obesity versus normal weight), strong evidence also supported increased risks of gastric cardia and ovarian cancer in obese individuals.

Both the IARC report and Kyrgiou and colleagues’ umbrella review consistently and strongly concluded that excess body fat increases the risk of most digestive system cancers as well as endometrial and postmenopausal breast cancer. However, for gastric cardia, and cancers of the liver, ovary, or thyroid, the strength of evidence differed between the two approaches, which can be explained by differences in the method used to summarise the evidence.

Firstly, unlike the IARC report, Kyrgiou and colleagues’ review did not evaluate the quality of the original meta-analyses. Although it reviews meta-analyses that may be susceptible to publication bias, the IARC report clearly demonstrated the importance of assessing the quality of each meta-analysis, including search strategy, inclusion and exclusion criteria, and data extraction,5 which is often outside the scope of an umbrella review.

Secondly, inclusion of suboptimal studies in a review, such as those that inadequately control for smoking in analyses exploring body mass index and lung cancer, may still provide a precise estimate, but biased in the wrong direction. This also limits the quality of the evidence. In contrast to the study by Kyrgiou and colleagues, the IACR report judged the evidence linking higher body mass index with lower risk of oesophageal squamous cell carcinoma and lung cancer as inadequate and inconsistent, partly owing to inadequate control for confounding by smoking and potential bias in the published literature.

Lastly, Kyrgiou and colleagues did not appraise meta-analyses of individual participant data but reviewed meta-analyses that combined results from published studies. Pooled analyses of individual participant data can be an efficient way to examine associations between obesity and cancer, especially for rare cancers,67 cancer subtypes (different histologies),8 and subgroups such as never smokers9 where individual studies and meta-analyses often do not have sufficient power. Associations deemed to be less convincing in Kyrgiou and colleagues’ review, such as those linking adiposity with cancers of the liver, ovary, and thyroid, were mostly downgraded because of heterogeneity between studies or the small number of cancer cases in each meta-analysis. Both these shortcomings can be overcome by including large pooled analyses. Although an umbrella review can be useful in providing a snapshot of evidence to explore a broad research question, the findings from umbrella reviews should be interpreted with caution as they are less comprehensive than reviews based on all available data.

Though some specifics remain to be worked out, the unavoidable conclusion from these data is that preventing excess adult weight gain can reduce the risk of cancer. Furthermore, emerging evidence suggests that excess body fat in early life also has an adverse effect on risk of cancer in adulthood.6710 Given the critical role of healthcare providers in obesity screening and prevention,1112 clinicians, particularly those in primary care, can be a powerful force to lower the burden of obesity related cancers, as well as the many other chronic diseases linked to obesity such as diabetes, heart disease, and stroke. The data are clear. The time for action is now.


  • Research: doi: 10.1136/bmj.j477
  • Competing interests: We have read and understood BMJ policy on declaration of interests and have no interests to declare.

  • Provenance and peer review: Commissioned; not peer reviewed.


View Abstract

Log in

Log in through your institution


* For online subscription