Dr. Petta and colleagues1 are to be congratulated on the first suggestion that retinol-binding protein 4 (RBP4) is a host protein that possibly interacts with viral proteins, resulting in hepatic steatosis in patients with genotype 1 chronic hepatitis C. It is currently suspected that the interaction between viral core proteins and host liver proteins is involved in the pathogenesis of hepatic steatosis not only in patients with genotype 3 chronic hepatitis C but also in those with genotype 1, although responsible proteins in the host have not been identified.2 In their study, multiple linear regression analysis and multiple logistic regression analysis are used to test the relationship between the blood levels of RBP4 and the degree of hepatic steatosis in 143 patients with genotype 1 chronic hepatitis C. In a multivariate linear regression model when the outcome variable is RBP4 concentration and explanatory variables are steatosis, insulin resistance, blood levels of liver enzymes, and other possibly contributing factors, only steatosis is independently correlated with RBP4 concentration. In the same analytic method when the outcome variable is steatosis, elevated RBP4 levels and increased insulin resistance are both independently correlated with the severity of steatosis. In logistic regression models when the determinant outcome is the severity of steatosis, univariate analysis identified four possible factors contributing to steatosis, such as body mass index (P = 0.09), waist circumference (P = 0.05), insulin resistance (P = 0.07), and RBP4 concentration (P = 0.0009). When the latter three variables are entered into a multiple logistic regression model, elevated RBP4 concentration is the only variable independently associated with the progression of steatosis.
I have two questions, one of which is on statistical methods used for the study. If we have continuous variables to be included in multiple logistic regression analysis, these variables must be converted into the categorical variables that are usually coded as 0 or 1. First, I would like to ask the authors how the cut-off values of RBP4 and the other two variables were determined in order to convert continuous data into categorical ones, because the results of multiple logistic regression analysis are strongly influenced by the number of covariables and these cut points.
As discussed in the article, RBP4 is secreted mainly by hepatocytes (80%) but also by adipose tissue (20%), and RBP4 secretion from visceral adipose tissue is increased in patients with obesity and type 2 diabetes,3 which is the reason why this protein is regarded as an adipokine. Circulating RBP4 levels in patients with chronic liver disease are probably dependent on liver protein synthesis capacity and effective hepatic blood flow. In fact, in a recent study that measured hepatic RBP4 production in vivo in patients with cirrhosis, blood concentrations of RBP4 were associated with hepatic production of RBP4 and correlated with circulating levels of albumin, cholinesterase, and coagulation factors.4 I wonder why mean blood levels of RBP4 were significantly higher in patients with chronic hepatitis C than those in healthy controls despite the lack of association between RBP4 and the grade of hepatitis activity or fibrosis among patients. Second, I would like to ask the authors whether a causal link between RBP4 and steatosis is restricted to patients having a normal hepatic capacity to secrete RBP4.