LAUSR

We were interested to read the article by Cheng et al. published in HEPATOLOGY in September 2016. The researchers aimed to evaluate the effect of various metabolic factors on the risks of hepatic events, cardiovascular events, and death in chronic hepatitis B patients. The results have demonstrated that metabolic syndrome (MetS) increased the risk of cardiovascular events, but not hepatic events and death. The researchers did not assess the bivariate correlation among the explanatory variables. It seems that the results of the study are biased by multicollinearity among MetS components. Actually, studies suggest that systolic blood pressure (SBP), diastolic blood pressure, and lipid profiles are highly correlated. The researchers assessed the effect of baseline MetS components and baseline liver stiffness measurement on the studied outcomes, but they only assumed a linear association between the exposures and the outcomes. Many studies found that there is a doseresponse or nonlinear association between each component of MetS with cardiovascular disease (CVD) events and death. For example, some studies found U-shape or J-shape associations between MetS components and CVD events. Full-range associations between each MetS components and outcomes risk can be studied using advanced methods such as Restricted Cubic Splines in Cox models. Moreover, it seems that the estimated associations between MetS components and the studied outcomes are diluted. For example, it was found that SBP and adiposity ratios are subject to within-person variability, and using a single measurement, such as baseline measurements, in the analysis is a source of bias. This bias is introduced as a regression dilution bias. Etiological associations between MetS components and CVD events and death should be appropriately checked using linear and nonlinear functions. Regression models with baseline measurements may be subject to regression dilution bias.