Does body size predict the risk of many different cancers?
Among millions of Spanish adults, higher BMI was linked to a higher risk of many cancers, especially of the womb and kidney, and waist size predicted risk about as well as BMI.
Source
Body mass index and waist circumference in relation to the risk of 26 types of cancer: a prospective cohort study of 3.5 million adults in Spain
Study at a glance
- Design
- Cohort — Prospective cohort built from Catalan primary-care electronic health records (SIDIAP), Cox models with age as time scale, followed up to end of 2018
- N
- N=3658417 · 3,658,417 adults with a valid BMI in the main dataset; a subsample of 291,305 also had waist circumference measured
- Population
- Adults aged 18-100 in Catalonia, Spain, with a BMI recorded by their GP or nurse between 2006 and 2017 and no previous cancer
- Outcome
- First incident diagnosis of each of 26 cancer types
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The researchers used routine primary-care records from Catalonia, where GPs and nurses measure weight, height and waist size directly. They followed 3,658,417 adults without cancer for a median of 8.3 years and linked their BMI to first diagnoses of 26 cancer types, adjusting for smoking, alcohol, type 2 diabetes, deprivation and nationality. They tested for curved (non-linear) relationships, repeated analyses in never smokers to strip out confounding by smoking, and in a subsample with waist measurements compared BMI and waist circumference head to head.
What they found
Each 5 kg/m² increase in BMI was linked to higher risk of nine cancers, most strongly cancer of the corpus uteri (hazard ratio 1.49), followed by kidney (1.16) and gallbladder/biliary cancers. For smoking-related cancers such as lung, larynx and oesophagus, low BMI looked risky, but these curves flattened out among never smokers, suggesting thinness was standing in for heavy smoking. Among never smokers, higher BMI was also linked to head and neck, brain and Hodgkin lymphoma. Waist circumference and BMI gave overlapping risk estimates for every cancer site.
The limits
What it doesn't show
This is observational, so residual confounding remains, and the authors had no data on diet, physical activity, reproductive history, or cancer subtype and stage. Only 62% of the adult database had a recorded BMI, raising possible selection bias, and only about 10% of participants had a waist measurement, so the BMI-versus-waist comparison applies to an older, heavier subgroup. With millions of people, even tiny hazard ratios reach significance, and cancer diagnoses in the records had only modest positive predictive value, which could bias results toward the null.
Key terms
- Hazard ratio
- How much faster an event (here a cancer diagnosis) occurs in one group compared with another over time; 1.49 means about 49% higher risk at any given moment.
- Residual confounding
- Bias left over after adjustment because a confounder, like smoking, was measured imperfectly or incompletely.
- Restricted cubic spline
- A flexible curve used in regression to let risk change non-linearly with an exposure such as BMI.
- Waist circumference
- A measure of central (abdominal) fat, as opposed to BMI, which reflects overall body size.
- Reverse causality
- When the outcome causes the exposure, e.g. an undiagnosed cancer causing weight loss before diagnosis.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What was the main exposure in the primary analysis?
Common questions
Why did low BMI seem to increase lung cancer risk?
Smokers tend to be thinner, and smoking causes lung cancer. When the analysis was limited to never smokers, the extra risk at low BMI largely disappeared, so the pattern mostly reflected smoking rather than thinness itself.
Is waist circumference better than BMI for predicting cancer?
In this study, no: for every cancer type the confidence intervals for waist and BMI overlapped, so the authors suggest BMI may be sufficient in primary care.
Why did they exclude the first year of follow-up?
To reduce reverse causality, since cancers that are growing but not yet diagnosed can cause weight loss and distort the BMI measured just before diagnosis.
More on Oncology outcomes