Potential conflict of interest: Nothing to report.
Article first published online: 30 OCT 2013
© 2013 by the American Association for the Study of Liver Diseases
Volume 58, Issue 5, page 1863, November 2013
How to Cite
Chaiteerakij, R., Harmsen, W. S., Ryu, E., Roberts, R. O., Olson, J. E., Therneau, T. M. and Roberts, L. R. (2013), Reply. Hepatology, 58: 1863. doi: 10.1002/hep.26441
- Issue published online: 30 OCT 2013
- Article first published online: 30 OCT 2013
- Accepted manuscript online: 8 APR 2013 04:50AM EST
- Manuscript Accepted: 4 APR 2013
- Manuscript Received: 29 MAR 2013
We thank Dr. Fujita for his interest in our study and his important comments. His general point is well taken that there is always a possibility of unknown factors causing differences between cases and controls in such a study. There are three possible sources of biases potentially influencing our finding of a protective effect of hyperlipidemia. First, Dr. Fujita expresses the concern that the frequency of hyperlipidemia in our controls (43.1%) is higher than that in the U.S. general population, i.e., 33.2%-35.0% in the reference he quotes. Given that the mean age of our controls was 62 and 90% of our controls were aged over 45, the frequency of hyperlipidemia in this age group in the U.S. general population is reported to be closer to 50% than to 33.2%-35.0%, i.e., 44.7%-46.1% among subjects aged between 45 and 64 years and 52.0%-54.4% among subjects aged = 65 years. Accordingly, the frequency of hyperlipidemia in our controls is lower than that in the age-matched general population. Second, our data source for determining the presence of hyperlipidemia was different from that used for the general population estimates. We used a self-report questionnaire to capture information on hyperlipidemia of both cases and controls, whereas the prevalence of hyperlipidemia in the general population was directly from blood test results. It is therefore possible that the frequency of hyperlipidemia from a self-reported questionnaire might be different from that obtained from blood testing. Lastly, since the recruitment process for cases and controls is different, there is a possibility that the differences in recruitment processes may be responsible for the differences detected between cases and controls. This basic principle is always true for a case-control study, as cases and controls are never identified by the same process. Because our source of cases was all cholangiocarcinoma patients seen at the Mayo Clinic, whereas the source of controls was subjects seen in the Department of Family Medicine, the Division of Community Internal Medicine, and the Division of General Medicine at Mayo Clinic, this difference in recruitment sources could be a bias that led to the finding of a protective effect of hyperlipidemia against cholangiocarcinoma. We recognized this potential bias and tried our best to minimize it by matching cases and controls by age, gender, and residence, as hyperlipidemia is highly associated with age and also with food and lifestyle patterns, which vary among different regions in the U.S. To determine whether hyperlipidemia is truly associated with a decreased risk of cholangiocarcinoma, a prospective study must be performed. However, it will be challenging to conduct a prospective study to answer this question because cholangiocarcinoma is relatively rare in the U.S. In support of our finding, a large 12-year prospective Japanese population-based study suggested an inverse relationship between baseline serum cholesterol level and risk for a number of cancers, in particular liver cancer. The mechanism by which hyperlipidemia may play a role against cancer needs to be further elucidated. For metformin, it is even more difficult to determine the difference in frequency of metformin use between our controls and the general population. An equally important question is whether there are differences in the characteristics of diabetic subjects who took or did not take metformin that could potentially bias or confound our study results.
Roongruedee Chaiteerakij, M.D.1,2
William S. Harmsen, M.S.3
Euijung Ryu, Ph.D.4
Rosebud O. Roberts, M.B., Ch.B.4
Janet E. Olson, Ph.D.4
Terry M. Therneau, Ph.D.3
Lewis R. Roberts, M.B., Ch.B. Ph.D.1
1Division of Gastroenterology and Hepatology, College of Medicine, Mayo Clinic and Mayo Clinic Cancer Center, Rochester, MN
2Department of Medicine, Faculty of Medicine, Chulalongkorn University and King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Bangkok, Thailand
3Department of Biomedical Statistics and Informatics, Mayo Clinic College of Medicine, Mayo Clinic Cancer Center, Rochester, MN
4Division of Epidemiology, Department of Health Sciences Research, Mayo Clinic, Rochester, MN
- 1Centers for Disease Control and Prevention (CDC). Prevalence of cholesterol screening and high blood cholesterol among adults—United States, 2005, 2007, and 2009. MMWR Morb Mortal Wkly Rep 2012;61:697-702.