Roughly speaking, the second kind of models had higher


Roughly speaking, the second kind of models had higher

AUROCs than the first, especially in identifying significant fibrosis. In both kinds, models constructed from CHB patients (The S index, Hui model and SLFG model) were superior to the others. Se, Spe, PPV and NPV of the predictive models were calculated using the cut-off values previously described in the original publications (Table 5). The S index had the highest predictive value in predicting significant fibrosis in the validation cohort. For patients with S index lower than 0.1, 33 of 37 (89.2%) would not have significant fibrosis. Only four of 68 (5.9%) with significant fibrosis would have S index lower than 0.1 and be classified incorrectly (the fibrosis stages of three patients in S2 and one patient GSK1120212 in vitro in S3). For patients with S index higher than 0.5, 29 of 33 (87.9%) would have significant fibrosis, and only four of 78 (5.1%) without significant fibrosis would be classified incorrectly. Together, 62 (42.5%) selleckchem of the total 146 patients could be identified correctly, only eight (5.5%) were misidentified. Seventy six (52.1%) remained uncertain with S index between 0.1 and 0.5. In the prediction of cirrhosis, some models did not provide recommended cut-off values. The presence of cirrhosis could be excluded with high certainty applying the low cut-off

of the other three models. But APRI and Hepascore were at risk to provide false positive results because of very low PPV. Only nine of 32 (28.1%) patients

with APRI higher than 2.0 and only 15 of 103 (14.6%) patients with Hepascore higher than 0.84 would have cirrhosis, while the S index could accurately predict the absence (S index < 0.3) or presence (S index ≥ 1.5) of cirrhosis in 103 (70.5%) of the total 146 patients, with NPV of 96.9% and PPV of 80.0%. Farnesyltransferase Many studies on noninvasive diagnostic models of liver fibrosis in chronic liver diseases have been published in the past few years. Most of them were conducted in CHC and few data are available on the applicability to CHB patients. Although two recent reports applied Fibrotest in CHB showing 0.77 and 0.78 AUROC for detection of significant fibrosis,18,19 it comprises markers routinely unavailable such as haptoglobulin, A2M and apolipoprotein A1. The need of complex tests and extra cost in calculation obviously reduce its practical utility. A few predictive models designed especially for CHB patients have already been proposed,13–15 but our study has several unique features. First, the SLFG model was built and validated in HBeAg positive CHB patients with ALT between 2 and 10 times the upper limit of normal (ULN), while Mohamadnejad et al. offered formulas only suitable for HBeAg negative patients. Hui et al. recruited only patients with HBV-DNA > 105 copied/mL and ALT between 1.5 and 10 times ULN. In the current study, we consecutively enrolled untreated chronic HBV carriers regardless of HBeAg, ALT and HBV DNA level.

Comments are closed.