Serum Liver Fibrosis Markers in the Prognosis of Liver Cirrhosis: A Prospective Observational Study.

BACKGROUND The prognostic role of serum liver fibrosis markers in cirrhotic patients remains unclear. We performed a prospective observational study to evaluate the effect of amino-terminal pro-peptide of type III pro-collagen (PIIINP), collagen IV (CIV), laminin (LN), and hyaluronic acid (HA) on the prognosis of liver cirrhosis. MATERIAL AND METHODS All patients who were diagnosed with liver cirrhosis and admitted to our department were prospectively enrolled. PIIINP, CIV, LN, and HA levels were tested. RESULTS Overall, 108 cirrhotic patients were included. Correlation analysis demonstrated that CIV (coefficient r: 0.658, p<0.001; coefficient r: 0.368, p<0.001), LN (coefficient r: 0.450, p<0.001; coefficient r: 0.343, p<0.001), and HA (coefficient r: 0.325, p=0.001; coefficient r: 0.282, p=0.004) levels, but not PIIINP level (coefficient r: 0.081, p=0.414; coefficient r: 0.090, p=0.363), significantly correlated with Child-Pugh and MELD scores. Logistic regression analysis demonstrated that HA (odds ratio=1.00003, 95% confidence interval [CI]=1.000004-1.000056, p=0.022) was significantly associated with the 6-month mortality. Receiver operating characteristics analysis demonstrated that the area under the curve (AUC) of HA for predicting the 6-month mortality was 0.612 (95%CI=0.508-0.709, p=0.1531). CONCLUSIONS CIV, LN, and HA levels were significantly associated with the severity of liver dysfunction, but might be inappropriate for the prognostic assessment of liver cirrhosis.

Background

Amino-terminal pro-peptide of type III pro-collagen (PIIINP), collagen IV (CIV), laminin (LN), and hyaluronic acid (HA) are 4 major serum markers for the non-invasive assessment of liver fibrosis [1-4]. Numerous studies have confirmed their diagnostic performance. Some examples have been shown as follows. In 1996, Murawaki et al. demonstrated a close relationship of elevated HA with the severity of liver fibrosis in patients with chronic viral liver diseases [5]. In 2000, Plevris et al. found that HA alone could reliably identify the presence of liver cirrhosis in patients with chronic liver diseases of mixed etiologies [6]. In 2000, the consensus interferon study group also found that HA alone may be effective in non-invasively assessing the degree of liver fibrosis and cirrhosis in patients with chronic hepatitis C virus infection [7]. In 2001, Murawaki et al. suggested the usefulness of PIIINP and HA for staging liver fibrosis in chronic hepatitis C [8]. In 2002, Xie et al. confirmed the relationship of HA and CIV with the degree of hepatic fibrosis in patients with chronic viral hepatitis [9]. In 2004, Patel et al. found that HA in combination with tissue inhibitor of matrix metalloproteinase 1 and alpha 2-macroglobulin can reliably differentiate between moderate/severe and no/mild fibrosis in patients with chronic hepatitis C infection [10]. In 2004, the European Liver Fibrosis Group reported that HA and PIIINP in combination with age and tissue inhibitor of matrix metalloproteinase 1 can accurately identify the absence of liver fibrosis [11]. In 2005, Cale et al. reported that HA in combination with platelet count, prothrombin index, aspartate aminotransferase, alpha 2-macroglobulin, urea, and age can predict the presence of clinically significant fibrosis in patients with viral hepatitis; HA in combination with prothrombin index, alpha 2-macroglobulin, and age could predict the presence of clinically significant fibrosis in patients with alcoholic liver diseases; HA in combination with gamma-glutamyltransferase, bilirubin, platelet count, and apolipoprotein A1 could predict the area of fibrosis in patients with viral hepatitis; and HA in combination with prothrombin index, alpha 2-macroglobulin, and platelets could predict the area of fibrosis in patients with alcoholic liver diseases [12]. In 2011, Seven et al. confirmed the correlation of PIIINP, CIV, LN, and HA with advanced fibrosis in patients with chronic hepatitis B and D [13]. In 2015, El-Mezayen et al. found that CIV and LN in combination with aspartate aminotransferase-to-platelet ratio index and albumin can be used to identify a very low risk of significant fibrosis in patients with chronic hepatitis C virus infection [14]. Theoretically, the grade of liver fibrosis is positively associated with the severity of liver dysfunction, thereby influencing the survival conditions. However, it remains unclear whether serum liver fibrosis markers can predict the prognosis of patients with liver cirrhosis. We conducted the present prospective observational study to explore this issue.

Material and Methods

This was a prospective observational study, which was registered with clinicaltrials.gov (NCT02335073). The study was conceived by 2 researchers (XSQ and XZG). The study protocol was written by XSQ, discussed with our study group, and approved by the Medical Ethics Committee of our hospital. The approval number was k(2014)28. The written informed consent was signed by every participant before liver fibrosis tests were performed. Inclusion criteria were: 1) patients were admitted to our department; 2) patients were diagnosed with liver cirrhosis; and 3) patients signed the informed consent and agreed to testing of serum liver fibrosis markers (PIIINP, CIV, LN, and HA). Major exclusion criteria were: 1) non-cirrhotic patients; 2) malignancy; and 3) repeated admission.

The participants were prospectively enrolled by our study group (ZMC n=1, YH n=17, XYL n=1, HL n=2, XL n=22, LNR n=14, DW n=6, HZW n=2, JLW n=2, CYW n=10, YLZ n=3, XZ n=3, YGZ n=18, and JJZ n=7). Three researchers (YP, HD, and FFH) recorded the regular clinical and laboratory data of participants from the electronic medical charts of our hospital in the printed case report forms. The primary data at admissions were: age, sex, hepatic encephalopathy, ascites, hydrothorax on chest X ray or CT scans, etiology of liver cirrhosis, red blood cell (RBC), hemoglobin (Hb), white blood cell (WBC), platelets count (PLT), total bilirubin (TBIL), albumin (ALB), alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), gamma-glutamine transferase (GGT), blood urea nitrogen (BUN), creatinine (Cr), prothrombin time (PT), international normalized ratio (INR), potassium (K), and sodium (Na). Child-Pugh and MELD scores were calculated [15,16]. Survival condition at the 6th month was obtained by collecting the re-admission and outpatient information in the electronic medical charts and telephone follow-up. Three researchers (XYW, FFH, and XSQ) were responsible for the telephone follow-up. The last telephone follow-up date was April 1, 2016.

As previously mentioned [17,18], a 3-ml fasting venous blood sample was obtained from every participant and then centrifuged. Two laboratory researchers (JC and CLX) tested the PIIINP, CIV, LN, and HA levels by using the chemiluminescent immunoassay in the LUmo Microplate Luminometer equipment at the lab of our department. The diagnostic kits for the PIIINP, CIV, LN, and HA were provided by the Autobio Diagnostics Co., Ltd. (Zhengzhou, Henan Province, China). According to the directions of diagnostic kits, the reference values were defined by the samples from 546 healthy volunteers. The reference values were: PIIINP <15 ng/mL, CIV <95 ng/mL, LN <130 ng/mL, and HA <120 ng/mL.

Statistical analyses were performed by using the SPSS and MedCalc software. Categorical data are expressed as frequencies. Continuous data are expressed as means with standard deviations and medians with ranges. Spearman non-parametric tests were employed to test the correlation of PIIINP, CIV, LN, and HA levels with clinical and laboratory data. Logistic regression analysis was used to test the statistically significant prognostic factors. Odds ratio (OR) with 95% confidence interval (CI) was calculated. Receiver operating characteristics (ROC) analysis was used to test the prognostic accuracy. The AUC was calculated and compared by the De Long test. Two-tailed p<0.05 was considered statistically significant.

Results

Patients characteristics

Between January and June 2015, 108 cirrhotic patients were included in this prospective observational study. Patient characteristics are shown in Table 1. A majority of included patients had hepatitis B virus infection and alcohol abuse as the major etiology of liver cirrhosis. Child-Pugh score was calculated for 104 of them. Mean Child-Pugh score was 7.5±1.9. Eighty-four percent of patients had Child-Pugh classes A and B. MELD score could be calculated in 104 of them. Mean MELD score was 7.1±5.6.

Table 1

Patient characteristics.

Variables	No. Pts available	Mean or frequency	Std. deviation	Median	Minimum	Maximum
Age (years)	108	59.030	11.498	59.065	26.74	83.16
Sex (Male/Female) – n.	108	67/41
Hepatic encephalopathy – n.	108	8
Ascites – n.	108	63
Hydrothorax on chest X ray or CT scans – n.	88	8
Etiology of liver cirrhosis – n.	108
Hepatitis B virus alone		22
Hepatitis C virus alone		7
Alcohol alone		27
Drug alone		4
Hepatitis B virus+Alcohol		8
Autoimmune		5
Cholestatic		2
Hepatitis B virus+Fatty Liver		1
Alcohol+Budd-Chiari Syndrome		1
Unknown		31
Red blood cell (1012/L)	108	3.239	0.856	3.225	1.38	5.36
Hemoglobin (g/L)	108	94.148	29.253	93	33	153
White blood cell (109/L)	108	4.257	2.315	4.1	0.9	15.7
Platelet (109/L)	108	96.185	62.805	78.5	22	316
Total bilirubin (umol/L)	108	33.418	36.361	22.65	5.2	234.8
Albumin (g/L)	107	31.665	6.276	31.5	16.8	46
Alanine aminotransferase (U/L)	108	35.306	33.980	25	5	249
Aspartate aminotransferase (U/L)	108	48.046	37.442	36	10	227
Alkaline phosphatase (U/L)	108	123.741	81.544	102.5	24	543
Gamma-glutamyl transpeptidase (U/L)	108	93.046	156.625	50.5	9	1377
Blood urea nitrogen (mmol/L)	106	6.676	3.645	5.62	1.47	20.46
Creatinine (umol/L)	106	86.243	133.729	62.45	34.5	1092
Potassium (mmol/L)	108	3.849	0.584	3.81	2.53	6.13
Sodium (mmol/L)	108	138.232	4.112	138.9	115.7	144.4
Prothrombin time (seconds)	105	14.663	3.648	14	10.4	38.8
International normalized ratio	105	1.262	0.313	1.2	0.9	3.37
Amino-terminal pro-peptide of type III pro-collagen (ng/mL)	108	31.373	37.246	13.14	2.18	192.35
IV-collagen (ng/mL)	108	225.882	333.062	149.69	28.79	2990.01
Laminin (ng/mL)	108	182.016	429.861	92.045	16.09	4184.99
Hyaluronic acid (ng/mL)	108	5275.794	21185.907	636.885	66.69	145053.94
Child-Pugh score	104	7.538	1.920	7	5	12
Child-Pugh class – n.	104
A		33
B		55
C		16
Model for end stage liver diseases (MELD) score	104	7.106	5.588	6.672	−3.16	31.4

Correlation of serum liver fibrosis markers with clinical and laboratory data

PIIINP level. No variables significantly correlated with PIIINP level (Table 2).

Table 2

Correlation of PIIINP with clinical and laboratory data by Spearman non-parametric tests.

Variables	No. Pts available	Correlation coefficient	Sig. (2-tailed)
Age	108	0.028	0.771
Sex	108	−0.047	0.628
Hepatic encephalopathy	108	0.079	0.414
Ascites	108	0.046	0.636
Hydrothorax on chest X ray or CT scans	88	0.114	0.292
Red blood cell	108	0.014	0.890
Hemoglobin	108	0.055	0.574
White blood cell	108	0.133	0.170
Platelet	108	−0.075	0.443
Total bilirubin	108	0.074	0.448
Albumin	107	−0.122	0.212
Alanine aminotransferase	108	−0.019	0.843
Aspartate aminotransferase	108	−0.016	0.871
Alkaline phosphatase	108	−0.113	0.246
Gamma-glutamyl transpeptidase	108	0.006	0.955
Blood urea nitrogen	106	−0.018	0.858
Creatinine	106	0.038	0.700
Potassium	108	−0.047	0.629
Sodium	108	−0.058	0.554
Prothrombin time	105	0.019	0.844
International normalized ratio	105	0.065	0.512
Child-Pugh score	104	0.081	0.414
Model for end stage liver diseases (MELD) score	104	0.090	0.363

CIV level. Ascites, RBC, WBC, TBIL, ALB, ALT, AST, GGT, Na, PT, INR, Child-Pugh score, and MELD score significantly correlated with CIV level (Table 3).

Table 3

Correlation of CIV with clinical and laboratory data by Spearman non-parametric tests.

Variables	No. Pts available	Correlation coefficient	Sig. (2-tailed)
Age	108	0.063	0.519
Sex	108	−0.126	0.192
Hepatic encephalopathy	108	0.027	0.78
Ascites	108	0.438	<0.001
Hydrothorax on chest X ray or CT scans	88	0.204	0.057
Red blood cell	108	−0.225	0.019
Hemoglobin	108	−0.073	0.453
White blood cell	108	0.225	0.019
Platelet	108	−0.131	0.176
Total bilirubin	108	0.434	<0.001
Albumin	107	−0.567	<0.001
Alanine aminotransferase	108	0.235	0.014
Aspartate aminotransferase	108	0.321	0.001
Alkaline phosphatase	108	0.174	0.072
Gamma-glutamyl transpeptidase	108	0.289	0.002
Blood urea nitrogen	106	−0.074	0.453
Creatinine	106	0.026	0.793
Potassium	108	−0.212	0.028
Sodium	108	−0.339	<0.001
Prothrombin time	105	0.356	<0.001
International normalized ratio	105	0.37	<0.001
Child-Pugh score	104	0.658	<0.001
Model for end stage liver diseases (MELD) score	104	0.368	<0.001

LN level. Ascites, WBC, TBIL, ALB, ALT, AST, ALP, GGT, Na, INR, Child-Pugh score, and MELD score significantly correlated with LN level (Table 4).

Table 4

Correlation of LN with clinical and laboratory data by Spearman non-parametric tests.

Variables	No. Pts available	Correlation coefficient	Sig. (2-tailed)
Age	108	0.059	0.544
Sex	108	0.022	0.821
Hepatic encephalopathy	108	0.100	0.301
Ascites	108	0.296	0.002
Hydrothorax on chest X ray or CT scans	88	0.178	0.097
Red blood cell	108	0.050	0.609
Hemoglobin	108	0.173	0.073
White blood cell	108	0.268	0.005
Platelet	108	−0.015	0.881
Total bilirubin	108	0.461	<0.001
Albumin	107	−0.324	0.001
Alanine aminotransferase	108	0.298	0.002
Aspartate aminotransferase	108	0.421	<0.001
Alkaline phosphatase	108	0.232	0.016
Gamma-glutamyl transpeptidase	108	0.254	0.008
Blood urea nitrogen	106	−0.179	0.066
Creatinine	106	0.060	0.543
Potassium	108	−0.158	0.102
Sodium	108	−0.238	0.013
Prothrombin time	105	0.153	0.120
International normalized ratio	105	0.199	0.041
Child-Pugh score	104	0.450	<0.001
Model for end stage liver diseases (MELD) score	104	0.343	<0.001

HA level. Ascites, hydrothorax, RBC, PLT, TBIL, ALB, PT, INR, Child-Pugh score, and MELD score significantly correlated with HA level (Table 5).

Table 5

Correlation of HA with clinical and laboratory data by Spearman non-parametric tests.

Variables	No. Pts available	Correlation coefficient	Sig. (2-tailed)
Age	108	0.062	0.525
Sex	108	−0.038	0.694
Hepatic encephalopathy	108	0.164	0.089
Ascites	108	0.300	0.002
Hydrothorax on chest X ray or CT scans	88	0.274	0.010
Red blood cell	108	−0.258	0.007
Hemoglobin	108	−0.104	0.286
White blood cell	108	0.000	1.000
Platelet	108	−0.268	0.005
Total bilirubin	108	0.240	0.012
Albumin	107	−0.248	0.010
Alanine aminotransferase	108	0.046	0.636
Aspartate aminotransferase	108	0.041	0.671
Alkaline phosphatase	108	−0.060	0.536
Gamma-glutamyl transpeptidase	108	0.075	0.442
Blood urea nitrogen	106	0.004	0.965
Creatinine	106	0.144	0.141
Potassium	108	−0.180	0.062
Sodium	108	−0.022	0.824
Prothrombin time	105	0.239	0.014
International normalized ratio	105	0.229	0.019
Child-Pugh score	104	0.325	0.001
Model for end stage liver diseases (MELD) score	104	0.282	0.004

Prognostic factors

The 6-month survival data was available in 97 patients. The 6-month mortality was 14.4% (14/97). The logistic regression univariate analysis of factors associated with the 6-month mortality included the absence of ascites (OR=0.184, 95%CI=0.038–0.899, p=0.037) and increased RBC (OR=0.355, 95%CI=0.159–0.795, p= 0.012), TBIL (OR=1.023, 95%CI=1.008–1.039, p=0.003), HA (OR=1.00003, 95%CI=1.000004–1.000056, p=0.022), Child-Pugh score (OR=1.561, 95%CI=1.113–2.152, p=0.007), and MELD score (OR=1.29, 95%CI=1.122–1.483, p<0.001) (Table 6).

Table 6

Logistics regression analysis of factors associated with 6-month death.

Variables	No. Pts available	6-month death vs. alive	P value	OR	95%CI
Age	97	14 vs. 83	0.237	1.031	0.980–1.085
Sex (Male vs. Female)	97 (57 vs. 40)	14 vs. 83	0.894	0.925	0.294–2.908
Hepatic encephalopathy (No vs. Yes)	97 (91 vs. 6)	14 vs. 83	0.999	2.937	NA
Ascites (No vs. Yes)	97 (39 vs. 58)	14 vs. 83	0.037	0.184	0.038–0.899
Hydrothorax on chest X ray or CT scans (No vs. Yes)	79 (72 vs. 7)	11 vs. 68	0.258	0.357	0.060–2.123
Red blood cell	97	14 vs. 83	0.012	0.355	0.159–0.795
Hemoglobin	97	14 vs. 83	0.062	0.980	0.959–1.001
White blood cell	97	14 vs. 83	0.99	0.998	0.775–1.285
Platelet	97	14 vs. 83	0.93	1.000	0.992–1.009
Total bilirubin	97	14 vs. 83	0.003	1.023	1.008–1.039
Albumin	96	14 vs. 82	0.08	0.916	0.831–1.011
Alanine aminotransferase	97	14 vs. 83	0.57	0.993	0.971–1.016
Aspartate aminotransferase	97	14 vs. 83	0.996	1.000	0.985–1.015
Alkaline phosphatase	97	14 vs. 83	0.488	1.002	0.996–1.008
Gamma-glutamyl transpeptidase	97	14 vs. 83	0.424	0.997	0.989–1.005
Blood urea nitrogen	95	14 vs. 81	0.102	1.121	0.978–1.286
Creatinine	93	13 vs. 80	0.111	1.002	0.999–1.006
Potassium	97	14 vs. 83	0.075	0.338	0.103–1.114
Sodium	97	14 vs. 83	0.069	0.896	0.795–1.009
Prothrombin time	94	14 vs. 80	0.161	1.093	0.965–1.237
International normalized ratio	94	14 vs. 80	0.128	3.084	0.723–13.155
Amino-terminal pro-peptide of type III pro-collagen	97	14 vs. 83	0.265	1.007	0.995–1.020
IV-collagen	97	14 vs. 83	0.235	1.001	0.999–1.004
Laminin	97	14 vs. 83	0.899	1.000	0.999–1.001
Hyaluronic acid	97	14 vs. 83	0.022	1.00003	1.000004–1.000056
Child-Pugh score	93	14 vs. 79	0.007	1.561	1.113–2.152
Model for end stage liver diseases (MELD) score	93	14 vs. 79	<0.001	1.290	1.122–1.483

ROC analysis

The AUC of Child-Pugh score, MELD score, and HA level for predicting the 6-month mortality was 0.692 (95%CI=0.587–0.783, p=0.0276), 0.803 (95%CI=0.707–0.878, p=0.0002), and 0.612 (95%CI=0.508–0.709, p=0.1531), respectively (Figure 1). There was a statistically significant difference between Child-Pugh score and HA level (p=0.0124), but there was not a statistically significant difference between Child-Pugh score and HA level (p=0.3421).

Figure 1

ROC analysis of Child-Pugh score (A), MELD score (B), and HA level (C) for predicting the 6-month mortality rate of cirrhotic patients.

Discussion

The present study had 2 primary objectives. The first objective was to validate our previous retrospective observational study regarding the correlation of the 4 serum liver fibrosis markers with the severity of liver dysfunction 17. The major similarity and discrepancy are summarized as follows. First, our previous study demonstrated that CIV (coefficient r: 0.2361, p=0.0006), LN (coefficient r: 0.2445, p=0.0004), and HA (coefficient r: 0.1612, p=0.0203) levels significantly correlated with Child-Pugh score, but not PIIINP level (coefficient r: 0.02665, p=0.7031). Similarly, the present prospective observational study confirmed that CIV (coefficient r: 0.658, p<0.001), LN (coefficient r: 0.450, p<0.001), and HA (coefficient r: 0.325, p=0.001) levels significantly correlated with Child-Pugh score, but not PIIINP level (coefficient r: 0.081, p=0.414). Second, our previous study also demonstrated that CIV (coefficient r: 0.1795, p=0.0108) and LN (coefficient r: 0.2588, p=0.0002) levels significantly correlated with MELD score, but not PIIINP (coefficient r: 0.04573, p=0.5191) or HA (coefficient r: 0.07926, p=0.2633) level. In contrast, the present study showed that CIV (coefficient r: 0.368, p<0.001), LN (coefficient r: 0.343, p<0.001), and HA (coefficient r: 0.282, p=0.004) levels significantly correlated with MELD score, but not PIIINP (coefficient r: 0.090, p=0.363) level. The possible causes for such a discrepancy could be: 1) the patient characteristics were different between the 2 studies; 2) Pearson chi-square test was used in the previous study, but Spearman non-parametric test was used in the present study; and 3) the correlation of HA level with MELD score might be unstable.

The second objective was to explore the effect of the 4 serum liver fibrosis markers on the survival of liver cirrhosis patients. Logistic regression analysis showed that only HA level, but not PIIINP, CIV, or LN level, was significantly associated with the 6-month mortality in cirrhotic patients. However, this association was very weak. When we used the ROC analysis to evaluate the effect of HA level for the predicting the 6-month mortality, the significance disappeared. Indeed, the prognostic value of HA level may be inferior to those of the traditional prognostic models (i.e., Child-Pugh and MELD scores). Therefore, we do not recommend the prognostic values of the 4 serum liver fibrosis markers in liver cirrhosis.

Several limitations of the present study should be mentioned. First, the in-hospital mortality was very low (0.9%, 1/108), and the logistic regression analysis of factors associated with the in-hospital mortality was not performed. Second, the information on 6-month mortality was missing in 11 patients (10.1%, 11/108). Third, long-term follow-up was lacking. Fourth, the causes of admission were heterogeneous.

Conclusions

CIV, LN, and HA levels significantly correlated with the severity of liver dysfunction, but they could not predict the 6-month mortality rate of cirrhotic patients. Therefore, the current evidence does not recommend the prognostic value of serum liver fibrosis markers in liver cirrhosis patients.