Obesity and risk of esophageal adenocarcinoma and Barrett's esophagus: a mendelian randomization study.

Thrift AP1, Shaheen NJ2, Gammon MD2, Bernstein L2, Reid BJ2, Onstad L2, Risch HA2, Liu G2, Bird NC2, Wu AH2, Corley DA2, Romero Y2, Chanock SJ2,Chow WH2, Casson AG2, Levine DM2, Zhang R2, Ek WE2, MacGregor S2, Ye W2, Hardie LJ2, Vaughan TL2, Whiteman DC2.

Author information

Abstract

BACKGROUND:

Data from observational studies suggest that body mass index (BMI) is causally related to esophageal adenocarcinoma (EAC) and its precursor, Barrett's esophagus (BE). However, the relationships may be affected by bias and confounding.

METHODS:

We used data from the Barrett's and Esophageal Adenocarcinoma Genetic Susceptibility Study: 999 patients with EAC, 2061 patients with BE, and 2169 population controls. We applied the two-stage control function instrumental variable method of the Mendelian randomization approach to estimate the unbiased, unconfounded effect of BMI on risk of EAC and BE. This was performed using a genetic risk score, derived from 29 genetic variants shown to be associated with BMI, as an instrument for lifetime BMI. A higher score indicates propensity to obesity. All tests were two-sided.

RESULTS:

The genetic risk score was not associated with potential confounders, including gastroesophageal reflux symptoms and smoking. In the instrumental variable analyses (IV), EAC risk increased by 16% (IV-odds ratio [OR] = 1.16, 95% confidence interval [CI] = 1.01 to 1.33) and BE risk increased by 12% (IV-OR = 1.12, 95% CI = 1.00 to 1.25) per 1kg/m(2) increase in BMI. BMI was statistically significantly associated with EAC and BE in conventional epidemiologic analyses.

CONCLUSIONS:

People with a high genetic propensity to obesity have higher risks of esophageal metaplasia and neoplasia than people with low genetic propensity. These analyses provide the strongest evidence to date that obesity is independently associated with BE and EAC, and is not due to confounding or bias inherent in conventional epidemiologic analyses.