Reply: To PMID 24443059.

Abstract

Potential conflict of interest: Nothing to report. Author names in bold designate shared co‐first authorship. We appreciate the interest that Dr. Li and colleagues have shown in our article published in Hepatology.1 Using a prospective nested case‐control design within the large cohort of the European Prospective Investigation into Cancer and Nutrition, our study provides one of the first pieces of epidemiological evidence on the role of specific biomarkers of inflammation and hyperinsulinemia in the risk of hepatocellular carcinoma (HCC). The comment by Dr. Li and colleagues implies that by not matching incident cases and controls on hepatitis B and C virus (HBV/HCV) infection, we may have not sufficiently accounted for confounding by HBV/HCV status. We do not agree with this argument. As reviewed in detail elsewhere, the primary purpose of matching in nested case‐control studies is not to avoid confounding, but to enhance study efficiency.2 In fact, matching can introduce a selection bias, that must be accounted for in the analysis by control of the matching factors.4 Our goal was to provide an unbiased estimate of the relative risk of HCC that would be expected in a general population for the metabolic and inflammatory biomarkers. Matching on HBV/HCV would likely make the controls less representative of the person‐time experience of the overall cohort5 and result in biased relative risk estimates. Thus, matching on HBV/HCV infection would not increase the validity (or avoid confounding) of our study. One appropriate method to control for confounding in epidemiological studies is adjustment for potentially confounding variables in regression modeling, as has been done in our study. In our analysis, adjustment for HBV/HCV infection did not substantially affect the relative risk estimates for the inflammatory and metabolic biomarkers. Thus, the relative risks of HCC per doubling of biomarker concentrations in the final multivariable model for C‐reactive protein, interleukin‐6, C‐peptide, and non‐high‐molecular‐weight adiponectin were 1.10 (95% confidence interval [CI]: 0.95‐1.28), 1.79 (95% CI: 1.29‐2.46), 2.64 (95% CI: 1.71‐4.09), and 2.81 (95% CI: 1.68‐4.70), respectively, without adjustment for HBV/HCV infection, and 1.22 (95% CI: 1.02‐1.46), 1.90 (95% CI: 1.30‐2.77), 2.25 (95% CI: 1.43‐3.54), and 2.09 (95% CI: 1.19‐3.67), respectively, with adjustment for HBV/HCV infection. Furthermore, as reported in our article, relative risk estimates were not substantially different when persons with prevalent HBV/HCV infection were excluded from the analysis (Supporting Table 2 in our publication). The practical value of our study is supported by our results showing that these inflammatory and metabolic biomarkers significantly improved prediction for future HCC risk beyond HBV/HCV infection and other established HCC risk factors. In conclusion, we have no doubt in stating that our main inference on the role of inflammatory and metabolic biomarkers in HCC risk is unlikely to be influenced by underlying HBV/HCV infection. We hope that our work will prompt further research that may explore potential applications for cancer prevention.