We are very pleased about the interest and the debate elicited by our article,1 and we read with attention the letters to the editor from both Tacke et al. and from Fujita. In our article1 we showed an association between elevated retinol-binding protein 4 (RBP4) levels and steatosis only in genotype 1 chronic hepatitis C (G1 CHC) patients, largely independent of obesity and the degree of insulin resistance. In contrast, in nonalcoholic fatty liver disease (NAFLD) the elevated RBP4 levels were more closely associated with insulin resistance and obesity, in keeping with a general metabolic disorder. We did not find any correlation between RBP4 serum levels and liver fibrosis.
Tacke et al., considering that the liver is the major source of circulating RBP4, suggest that our results could be affected by the severity of liver disease. In particular, the authors showed that, in a heterogeneous cohort of 111 patients with chronic liver disease (CLD) of different etiology (viral, biliary/autoimmune, alcohol, other) and at different stage of disease (16% no cirrhosis, 32% Child A, 52% Child B and C), referred for liver transplantation, serum RBP4 significantly decreased with the stage of cirrhosis and correlated with liver function tests.2
In response to Tacke et al., we would like to emphasize that in our study we included only fully compensated patients with G1 CHC or NAFLD. Differences in etiology of liver disease and in the baseline severity of the illness among the two studies may explain these discrepancies. We agree with Tacke et al. that hepatic RBP4 production is clearly reduced in Child B and C cirrhosis.2, 3 On the other hand, conflicting data exist on RBP4 expression in compensated Child A cirrhosis. In this setting, Yagmur et al.2 showed a significant RBP4 reduction compared to controls, whereas a similar expression was identified by Bahr et al.3 Therefore, we think that the actual evidences are sufficient to conclude that liver function may have a significant impact on the assessment of RBP4 serum levels only in patients with decompensated liver disease, although more data are needed in compensated liver disease. We agree with Tacke et al. that a methodological issue arises in the potential limitation of the generalizability of our results to new population and setting. Our study included a cohort of nondiabetic European patients with fully compensated CLD due to hepatitis C virus (HCV) or NAFLD (only one patient with HCV infection had histological cirrhosis) limiting the broad application of the results. Further large-scale prospective studies conducted in populations with different etiologies and different liver functions may prove useful to clarify the role of RBP4 in CLD.
Tacke et al. performed an additional analysis on their published population2 and did not confirm that serum RBP4 levels increase according to the severity of the steatosis, as shown in our study.1 However, the lack of this association in a cohort of patients in which only 28% had postviral cirrhosis due to hepatitis B virus or HCV infection does not conflict with our results. In fact, we did not identify a link between RBP4 levels and steatosis in NAFLD, but only in G1 CHC, and hypothesized that: (1) genotype 1 HCV might directly induce the hepatic overexpression of RBP4, independently of necroinflammation and fibrosis, and (2) this protein may in turn have an important role in the control of hepatic fat infiltration.1 This hypothesis could be further confirmed by our preliminary and unpublished data showing an overexpression of RBP4 compared to controls in the liver of G1 CHC subjects according to degree of steatosis, independent of inflammation and fibrosis, and not observed in NAFLD.
Concerning the methodological question of Fujita, it is well known that in multiple logistic regression models it is possible to include as independent variables both qualitative and quantitative variables. We chose to include the Homeostasis Model Assessment as quantitative variable because there is no general agreement about which cutoff should be used. In fact, this cutoff depends on the distribution of insulin resistance in the referral general population. Finally, replacing in the logistic model all the continuous variables with categorical variables makes no material difference to the findings.
Regarding the other questions of Fujita, we think that: (1) genotype 1 HCV might directly induce the hepatic overexpression of RBP4, independent of necroinflammation and fibrosis, which explains the higher RBP4 values compared to controls, and (2) RBP4 liver production is reduced only in patients with decompensated liver disease.