To the Editor:

Noninvasive assessment of hepatic fibrosis is gaining popularity.1, 2 Several noninvasive tests have been proposed, including indirect markers of liver fibrosis composed of routine laboratory parameters. We read with great interest the article by Lok et al.3 on the use of the regression formula log odds = −5.560.0089 × platelet109/L) + 1.26 × AST/ALT ratio + 5.27 × INR (“model 3”) for prediction or exclusion of cirrhosis (Ishak fibrosis score F5–F6) in patients with chronic hepatitis C (CHC). Although this model was developed in a cohort of nonresponders to previous antiviral treatment that comprised patients with more advanced stages of fibrosis (Ishak fibrosis score F3–F6), it was also applied to 265 treatment-naïve CHC patients previously studied,4 confirming its diagnostic accuracy across a broad range of fibrosis (Ishak fibrosis score F0–F6).

We recently evaluated several noninvasive tests based on standard laboratory parameters, including the aspartate aminotransferase to platelet ratio index (APRI) and platelet count per se, in a cohort of 194 treatment-naïve CHC patients.5 For external validation, we applied model 3 in our CHC study cohort and compared its diagnostic accuracy for prediction/exclusion of cirrhosis (Ishak fibrosis score F5–F6) with that of platelet count and APRI. Ishak fibrosis score correlated moderately with platelet count, APRI, and model 3 (Spearman's R −0.50, 0.62, and 0.53, respectively). Evaluation of the overall diagnostic accuracies for the prediction of cirrhosis by receiver-operating characteristic analysis revealed comparable results: areas under the receiver-operating characteristic curves for platelet count, APRI, and model 3 were 0.89, 0.90, and 0.90, respectively. Our findings on the diagnostic accuracy of model 3 for prediction/exclusion of cirrhosis are very similar to the results reported by Lok et al. (Table 1). Each of the noninvasive tests evaluated was more suited for exclusion of cirrhosis than for its prediction. Using the reported cutoff values of ≥150 × 109/L (platelet count), <1.0 (APRI), and <0.2 (model 3), each of the tests reliably excluded cirrhosis with a negative predictive value of 95% or more in 78%, 59%, and 49% of patients, respectively. Conversely, prediction of cirrhosis using a cutoff value of <150 × 109/L (platelet count), ≥2.0 (APRI), and ≥0.5 (model 3) was much less reliable (positive predictive value 56%, 59%, and 58%, respectively). Our data confirm the excellent diagnostic accuracy of model 3 for the exclusion of cirrhosis; however, we would like to point out that the platelet count per se allows the exclusion of cirrhosis with a fair degree of accuracy without the need for a calculator or computer.

Table 1. Diagnostic Accuracies of Noninvasive Fibrosis Tests for Prediction of Cirrhosis (Ishak Fibrosis Score F5–F6)
Noninvasive TestCutoffPatients Classified (%)Diagnostic Accuracy (%)Ishak Fibrosis Score F5–F6
  • NOTE. The cutoffs described in the original reports were used for all calculations. Values reported in the original studies are given in parentheses.

  • Abbreviations: PPV, positive predictive value; NPV, negative predictive value; APRI, aspartate aminotransferase to platelet ratio index.

  • *

    Model 3: log odds = −5.560.0089 × platelet109/L) + 1.26 × AST/ALT ratio + 5.27 × INR; predicted probability = exp(logodds)/(1 + exp(logodds)).

  • Values reported by Wai et al.4

  • Values reported by Lok et al.3

Platelet count (×109/L)  Sensitivity77
APRI  Sensitivity93 (89)
<1.059Specificity70 (75)
≥1.041PPV38 (38)
  NPV98 (98)
   Sensitivity55 (57)
<2.085Specificity93 (93)
≥2.015PPV59 (57)
  NPV91 (93)
Model 3*  Sensitivity100 (98)
<0.2049Specificity58 (53)
≥0.2051PPV30 (27)
  NPV100 (99)
   Sensitivity44 (53)
<0.5088Specificity94 (95)
≥0.5012PPV58 (64)
  NPV90 (92)

Exclusion of cirrhosis in patients with CHC is important, because costly diagnostic workup may be avoided, including (1) endoscopic screening for esophageal varices and (2) surveillance for hepatocellular carcinoma via regular ultrasound and α-fetoprotein examinations. In treatment-naïve patients with CHC typically seen in a referral center, cirrhosis can be reliably excluded in a considerable number of patients using simple noninvasive fibrosis tests based on routine laboratory parameters.


  1. Top of page

Carolin Lackner*, Gerd Struber*, Csilla Bankuti†, Bernd Bauer‡, Rudolf E. Stauber†, * Department of Pathology, Medical University Graz, Graz, Austria, † Department of Internal Medicine, Medical University Graz, Graz, Austria, ‡ Department of Internal Medicine, Landeskrankenhaus Hoergas, Gratwein, Austria.