|
Objective To evaluate the clinical value of non-invasive serum markers for fibrosis assessment in chronic hepatitis B virus infection (CHB), and to establish logistic regression equations and artificial Neural Network(ANN) prediction model for quantitative estimation of liver fibrosis.Methods serum samples from patients with CHB, and 53 control serum samples (23 from patients with non-HBV infectious diseases, and 30 from healthy individuals ) were collected. In the CHB group, 16 patients had stage 1 inflammation (G1); 21 patients had stage 2 inflammation (G2); 51 patients had stage 3 inflammation (G3); 20 patients had stage 4 inflammation (G4) ,while 12 patients had stage 0 fibrosis (S0); 18 patients had stage 1 fibrosis (S1); 21 patients had stage 2 fibrosis (S2); 20 patients had stage 3 fibrosis (S3) ; and 37 patients had stage 4 fibrosis (S4), according to the histological assessment. The levels of HA, LN, PⅢNP and CⅣin serum were measured by chemiluminescence immunoassay, the levels ofα2M and HPT were measured via laser immunonephelometry(BN-II ,Dade Behring),and GGT, TBIL and ApoA1 were analyzed by automated biochemical analyzer(AU2700, Olympus). The Fibrotest score was got according to the classic formula. ROC curve was drawed based on the results, which was used to evaluate the diagnostic value of HA, LN, PⅢNP, CⅣand Fibrotest. In addition, correlation between several laboratory markers and fibrosis stages was assessed by logistic regression and regression equations and ANN prediction model was subsequently established. Result The levels of serum HA(63.95 ng/ml), LN(78.39 ng/ml), PⅢNP(9.26 ng/ml), CⅣ(67.32 ng/ml)and Fibrotest(0.41±1.55) in CHB were higher than those in healthy controls(HA 19.73 ng/ml, LN 26.84 ng/ml, PⅢNP 2.58 ng/ml, CⅣ15.30 ng/ml, Fibrotest -1.03±0.63, all P﹤0.01). The levels of serum LN, PⅢNP, CⅣ and Fibrotest in CHB were higher than those in non-HBV infectious disease(LN 30.69 ng/ml, PⅢNP 6.10 ng/ml, CⅣ35.27 ng/ml , Fibrotest -0.59±1.58, P<0.05),but the difference of HA between the two goups was not significant (63.95 ng/ml vs 68.23 ng/ml, P﹥0.05). The levels of serum HA(34.72 ng/ml), LN(30.67 ng/ml), PⅢNP ( 3.90 ng/ml ) , CⅣ( 23.31 ng/ml ) and Fibrotest ( -0.84±1.15 ) in non-inflammatory group were lower than those in G3 group(HA 76.63 ng/ml, LN 120.38 ng/ml, PⅢNP 10.01 ng/ml, CⅣ74.25 ng/ml , Fibrotest 0.34±1.20)and G4 group(HA 104.68 ng/ml, LN 174.20 ng/ml, PⅢNP 15.95 ng/ml, CⅣ144.35 ng/ml , Fibrotest 1.74±1.53,all P<0.001). All the markers in G4 group were higher than those in G1 group(HA 40.55 ng/ml, LN 32.03 ng/ml, PⅢNP 5.95 ng/ml, CⅣ40.66 ng/ml , Fibrotest -0.24±1.39)and G2 group(HA 35.77 ng/ml, LN 22.72 ng/ml, PⅢNP 6.13 ng/ml, CⅣ35.49 ng/ml , Fibrotest 0.23±1.28, all P﹤0.01). The levels of serum HA(34.72 ng/ml), LN(30.67 ng/ml), PⅢNP(3.90 ng/ml), CⅣ(23.31 ng/ml)and Fibrotest(-0.84±1.15) in non-fibrosis group were lower than those in S3 group(HA 66.31 ng/ml, LN 130.27 ng/ml, PⅢNP 8.73 ng/ml, CⅣ86.86 ng/ml , Fibrotest 0.62±1.29)and S4 group (HA 88.78 ng/ml, LN 154.84 ng/ml, PⅢNP 12.36 ng/ml, CⅣ112.74 ng/ml , Fibrotest 1.15±1.63, all P﹤0.01). The level of Fibrotest in S4 group was higher than that in S1 group(-0.24±1.84)and S2 group(-0.27±1.26,P<0.05).The higher the severity of fibrosis, the higher the levels of all the markers. The area under the receiver operating characteristic curve(AUC) for diagnosing inflammation were PⅢNP(0.793)﹥CⅣ(0.787)﹥Fibrotest(0.753)﹥LN(0.715)﹥HA(0.631). LN can be used to diagnose inflammation with positive likelihood ratio (+LR)﹥10 , while combination of all the five markers can be used to exclude inflammation with negative likelihood ratio (-LR)﹤0.1. The AUC for diagnosing fibrosis were Fibrotest(0.80)﹥CⅣ(0.771)﹥PⅢNP (0.755)﹥LN(0.738)﹥HA(0.626),and LN can be used to diagnose fibrosis with +LR﹥10 , while combination of all the five markers can be used to exclude fibrosis with -LR﹤0.1. The AUC for diagnosing cirrhosis were Fibrotest(0.776)﹥CⅣ(0.752)﹥LN(0.745)﹥PⅢNP (0.725)﹥HA(0.625),and combination of all the five markers can be used to exclude cirrhosis with negative predictive value(NPV) of 98.1%. The logistic regression equation to diagnose fibrosis in CHB was:logitP = -2.739 + 3.270*gender + 0.003*CⅣ+ 1.959*α2M - 0.032*PLT and to diagnose cirrhosis was: LogitP = 16.388 + 2.85*gender + 0.004*CⅣ- 0.037*PLT - 1.694*FIB. In 23 validation cases for evaluating the established ANN prediction model, 17 matched the pathological diagnosis. In diagnosing fibrosis with ANN , the sensitivity was 95.0%,specificity was 66.67% and diagnostic accuracy was 91.30%.Conclusions The levels of serum markers are of certain value for fibrosis assessment in CHB, and combination of the markers can increase the diagnostic value and reduce unnecessary liver biospy. Chemiluminescence immunoassay was better than traditional RIA and ELISA methods in detecting LN, PⅢNP and CⅣ. The logistic regression equations and ANN model would be helpful in the quantitative assessment of CHB liver fibrosis.
|