Using a case-control analysis, Chaiteerakij et al. revealed that diabetes mellitus (DM) was associated with a 3.6-fold risk of developing intrahepatic cholangiocarcinoma (ICC) and that metformin use for DM reduced the risk of ICC by 60%. Furthermore, hyperlipidemia was found to be a protective factor against ICC. These findings are impressive, but may not be translated into the general population. Chaiteerakij et al. recruited 594 participants as controls from the Mayo Clinic Biobank to compare to 612 patients with ICC. The case-control study, which is, in a way, analogous to the prospective cohort study, has been used for over 60 years to examine the association between disease and potential risk factors, but this method has some limitations, one of which is susceptibility to selection bias. Case-control sampling is carried out in the context of an actual cohort study, but sampling fractions for controls are much smaller than those for cases as noted in the study by Chaiteerakij et al. In such situations, selection bias does occur if exposed controls are more or less likely to be sampled than nonexposed controls. For instance, if the frequency of sampled exposed persons as controls was twice the frequency as nonexposed persons, the estimated odds ratio would be twice the correct value. In Table 1 of the article, the incidence of hyperlipidemia in controls is 43.1%, which appears to be higher than that in the U.S. general population, suggesting that the protective effect of hyperlipidemia was overestimated. There are some concerns with regard to the analysis of the relationship between metformin use and risk of ICC. To eliminate such concerns, further analysis would be warranted.
Tetsuji Fujita, M.D.
Department of Surgery, Jikei University School of Medicine, Tokyo, Japan