We read with interest the Koda et al. report on the description of a new panel of biomarkers for predicting significant fibrosis in patients with chronic hepatitis C (CHC) called FibroIndex.1 This panel was derived from the platelet count, AST, and gamma globulin measurements. They built and validated their model on a cohort of 360 CHC patients with an estimation set (n = 240) and validation set (n = 120). The areas under the receiver operating characteristic (ROC) curves of FibroIndex for predicting significant fibrosis were 0.83 and 0.82 for the validation set, better than those of the Forns index and the aminotransferase-to-platelet ratio index (APRI). The ultimate utility of any noninvasive model for prediction of hepatic fibrosis depends on its practicality and validation by other investigators in a wide range of patients.2 We studied the diagnostic accuracy of the FibroIndex and compared it to APRI, Forns index, and Fibrotest in a cohort of 125 CHC patients prospectively enrolled in Fibroscore, a French national multicenter study. Of the 125, 85 (68%) were men, and mean age was 47 ± 9 years. Signed informed consent was obtained from all patients before their inclusion. Liver biopsy and biochemical markers were done the same day. Liver biopsy was performed in each center and analyzed by the resident pathologist. For all patients, ultrasound examination was performed before liver biopsy. Information relating to the patient demographic data, risk factors, virological status, clinical examinations, biological data (platelets, prothrombin time ratio, serum albumin level) was prospectively recorded in each center on the day of biopsy. All data were anonymously recorded in the database. Diagnostic accuracies were measured using area under ROC curves (AUCs), sensitivities, specificities, and positive and negative predictive values. The Hanley-McNeil test3 was used to compare AUCs; χ2 or Fisher exact test were used to compare proportions. Serum samples were taken, on the day of biopsy, for determination of biochemical markers to assess FibroIndex, APRI, Forns index, and Fibrotest. Biochemical marker analysis was performed in accredited laboratories following the guidelines recommended for Fibrotest assessment by the authors of the initial publication.4
In the hospital-based cohort, gamma glutamyl transpeptidase and total bilirubin levels were measured using a Hitachi 917 Analyzer and Roche Diagnostics reagents (both from Mannheim, Germany). Alpha2-macroglobulin, apolipoprotein A1, and haptoglobin were measured using a Modular analyzer (BNII, Dade Behring; Marburg, Germany). Platelets were measured with a Beckman Coulter LH 750 hematology analyzer. Biochemical assays were performed on fresh serum, decanted, and stored for 72 hours maximum at 2°C/8°C while protected from light. Cholesterol was measured on a Beckman Coulter Synchron CX 7 analyzer (Beckman Coulter France SA, Roissy, France). The assays of the specific proteins (alpha2-macroglobulin, haptoglobin, and apolipoprotein A1) were carried out on serum stored at 2°C/8°C for 5 days.
According to Metavir fibrosis staging, 6 of 125 (5%) patients were F0, 39 of 125 (31%) were F1, 36 of 125 (29%) were F2, 25 of 125 (21%) were F3, and 18 of 125 (14%) were F4. Note that the fibrosis distribution of our cohort is comparable to the one of the validation set of Koda et al.
The AUC results are presented in Table 1. For significant fibrosis, the AUC found in our cohort was lower than the one reported by Koda et al. (0.76 versus 0.86, P not significant). Our results observed for APRI and Forns index are similar to the corresponding results reported.
|Test||F01 vs. F234 AUC(95% CI)||F01 vs. F34 AUC(95% CI)||F01 vs. F23 AUC(95% CI)||F012 vs. F3 AUC(95% CI)|
|Fibroscore||Koda et al.||Fibroscore||Koda et al.||Fibroscore||Koda et al.||Fibroscore||Koda et al.|
|FibroIndex||0.76 (0.68-0.83)||0.86 (0.81-0.92)||0.72 (0.64-0.80)||0.85 (0.79-0.91)||0.72 (0.63-0.80)||0.83 (0.75-0.90)||0.62 (0.53-0.72)||0.81 (0.73-0.89)|
|FT||0.78 (0.70-0.85)||–||0.80 (0.72-0.87)||–||0.74 (0.64-0.82)||–||0.73 (0.63-0.81)||–|
|APRI||0.81 (0.73-0.82)||0.82 (0.76-0.89)||0.78 (0.69-0.85)||0.81 (0.74-0.88)||0.77 (0.68-0.84)||0.78 (0.69-0.86)||0.71 (0.61-0.79)||0.77 (0.69-0.86)|
|Forns||0.80 (0.72-0.86)||0.84 (0.77-0.90)||0.78 (0.70-0.85)||0.83 (0.77-0.89)||0.76 (0.67-0.84)||0.78 (0.70-0.86)||0.73 (0.64-0.81)||0.76 (0.68-0.85)|
The distribution of our population cannot explain the discordance observed in the diagnostic accuracy. Among the 3 variables included in the FibroIndex (AST, platelet count, and serum gamma globulin), only the 2 former can be assessed with precision. Despite the fact that serum gamma globulin was higher in patients with scores of F2 or F3 than in those with scores of F0 or F1,4 this parameter cannot be correctly assessed by electrophoresis. Gel electrophoresis is especially useful as an analytical method. Its advantage is that proteins can be visualized as well as separated, permitting a researcher to estimate quickly the number of proteins in a mixture or the degree of purity of a particular protein preparation.5 Electrophoresis provided only a semiquantitative assessment of serum gamma globulin. Calculation was done by the formula: (% of the area under the pic) × (concentration of total proteins). This calculation gives only an estimation of the serum gamma globulin with expected fluctuations.5 In our study, the same method that Koda et al. used was done to assess the serum gamma globulin.
In conclusion, robustness of the analysis should be assessed before a validation in clinical scores. The FibroIndex does not constitute for us a practical index for predicting significant fibrosis in patients with CHC. Moreover, FibroIndex has lower diagnostic accuracies than APRI, Forns index, or Fibrotest.