Clinical and labarotory characteristics of the patients with and without acute-on chronic liver faliure
The data of 202 patients who fit the inclusion criteria were analyzed. Of these patients, 74 patients (37%) suffered from acute on chronic liver failure, and 128 patients (60%) were not. About 38% of the ACLF group died, while just 8.5% of the non-ACLF group died (Table 1).
Patients with ACLF were older than the non-ACLF group (68.5±5 vs 62.7±11.1, p<0.05) (Table 1). Both groups had the same BMI and same number of type 2 diabetes mellitus patients. The main artery pressure was lower in the ACLF group (81.5±12.6 vs. 88 ±13.2, p<0.05) (Table 1). Cirrhosis was more severe in the ACLF group as seen by the MELD score (18.5±7.9 vs. 11.3±5.1, p<0.001), the Child-Pugh score (8.6±1.9 vs. 7.6±1.8, p<0.001), CLIF-C score (44.8±8.9 vs. 38±7.5, p<0.001), and the VTE predicating score, the PADUA score (3.0±2.3 vs. 1.7±1.5, p<0.001) (Table 1). Liver function was reduced in the ACLF group: INR was higher in the ACLF group (1.6±0.7 vs 1.4±0.4, p<0.05), total bilirubin also was higher in the ACLF group (2.6± 4.2 vs. 1.7± 1.5, p<0.05), and albumin was lower (2.9±0.6 vs 3.1±0.7, p<0.05) (Table 1). Although the liver enzymes were high in the ACLF group, there was no significant differences between liver enzymes (AST, ALT, GGT and fibrinogen) in the ACLF and non-ACLF groups (Table 1). LDH was higher in the ACLF group (831±490 vs 222±110, p<0.05). Kidney function was disturbed more in the ACLF group with the creatinine being higher in the ACLF group (2.0±1.3 vs. 0.8±0.7, p<0.05) (Table 1). Inflammatory markers like CRP, NLR and WBC were also significantly high in the ACLF group (60±62 vs 44±34 (p<0.01), 5.7±4.3 vs. 5.1±3.8 (p<0.05) and 6.9±4.1 vs 6.1±2.9 (p<0.05), respectively). No significant changes in the platelet count and ammonia levels were seen (Table 1).
Table 1. Clinical Characteristics of the ACLF and non-ACLF groups in ambulatory patients with no cirrhotic livers. NS: non-significant (P value<0.05).
VARIABLE
|
ACLF
|
NON-ACLF
|
P-VALUE
|
Total
|
N=74
|
N=128
Compensated (4%)
|
|
Age
|
68.5±5
|
62.7±11.1
|
0.0001
|
Male (% of total)
|
52%
|
60%
|
0.45
|
Etiology of Cirrhosis
|
Alcohol:15%
NASH: 55%
HBV: 10%
HCV:8%
WILSON: 0.5%
Autoimmune: 1.5%
Cryptogenic 4.5%
Cardiac 4%
Biliary 1.5%
|
Alcohol:16%
NASH: 56%
HBV: 8%
HCV:7%
WILSON: 0.7%
Autoimmune: 2%
Cryptogenic: 1%
Cardiac: 5%
Biliary: 4%
|
>0.05
|
MAP (mmHg)
|
81.5±12.6
|
88 ±13.2
|
0.0008
|
BMI
|
29.6±6.9
|
28±4.7
|
0.1
|
Diabetes
|
67%
|
62%
|
0.5
|
Precipitating events:
Alcohol
Infection
GI bleeding
Idiopathic
Decompensation:
Ascites:
GI bleeding:
Encephalopathy:
Sepsis:
|
8%
47%
35%
10%
30%
25%
42%
2%
|
22%
40%
36.5%
2%
8%
35%
56%
1%
|
>0.05
<0.05
|
MELD score
|
18.5±7.9
|
11.3±5.1
|
0.000
|
Child-Pugh score
|
8.6±1.9
|
7.6±1.8
|
0.000
|
CLIF-C score
|
44.8±8.9
|
38±7.5
|
0.000
|
PADUA score
|
3.0±2.3
|
1.7±1.5
|
0.0003
|
Laboratory parameters
|
|
|
|
INR
|
1.6±0.7
|
1.4±0.4
|
0.001
|
Bilirubin (mg/dl)
|
2.6±4.2
|
1.7±1.5
|
0.002
|
Albumin (g/dl)
|
2.9±0.6
|
3.1±0.7
|
0.002
|
Creatinine (mg/dl)
|
2.0±1.3
|
0.8±0.7
|
0.0001
|
Sodium (mmol/l)
|
136±5.5
|
137±12.0
|
0.82
|
Hemoglobin (g/dl)
|
9.8±2
|
9.7±2.6
|
0.04
|
WBC (x103/µl)
|
6.9±4.1
|
6.1±2.9
|
0.03
|
Platelet (x103/µl)
|
141±87
|
136±86
|
0.9
|
C-reactive protein (CRP) (mg/dl)
|
60±62
|
44±34
|
0.0001
|
Fibrinogen (g/dl)
|
402±192
|
399±1.39
|
0.58
|
Ammonia (micromol/L)
|
77±49
|
84±47
|
0.7
|
ALT (U/L)
|
147±55
|
34±30
|
0.1
|
AST (U/L)
|
340±111
|
96±92
|
0.08
|
GGT (U/L)
|
266±205
|
296±196
|
0.4
|
LDH (U/L)
|
831±490
|
222±110
|
0.04
|
NLR
|
5.7±4.3
|
5.1±3.8
|
0.004
|
Mortality (% of total)
|
38
|
8.5
|
0.00001
|
Association between liver failure scores, demographic parameters and inflammatory markers, and ACLF development
In addition to gender and age, liver failure scores and inflammatory markers were studied for their correlation to organ failure and ACLF development in the cirrhotic patient groups. The results show that the MELD (MELD>18) score was the best predictor of organ failure in a univariate regression (Table 2A). Multivariate analysis showed that both MELD score and CLIF-C ACLF were associated with organ failure (Table 2B).
Table 2. Association between liver faiulre scores, inflammatory markers and organ faliure. A: Univariate analysis of the risk factor strength with organ faliure. B: Multivariate analysis of the risk factor strength. SE are the standard errors of the regression coefficients. T is the quotient of the coefficient. P value is the two-sided p values or observed significance levels.
A.
P-value
|
T
|
Std. Error
|
95% Conf. (±)
|
Coefficient
|
|
|
|
|
|
|
Constant
|
0.85
|
0.179
|
0.0042
|
0.008
|
0.0007
|
Age
|
0.66
|
0.43
|
0.071
|
0.14
|
0.03
|
Gender
|
0.0007
|
3.45
|
0.006
|
0.012
|
0.02
|
MELD score
|
0.40
|
0.83
|
0.022
|
0.044
|
0.01
|
CHILD PUGH score
|
0.14
|
1.48
|
0.0062
|
0.012
|
0.009
|
CLIF-C ACLF
|
0.32
|
0.99
|
0.020
|
0.0397
|
0.02
|
PADUA
|
0.47
|
0.70
|
0.0008
|
0.0016
|
0.0005
|
CRP
|
0.327
|
0.94
|
0.00027
|
0.0005
|
0.0002
|
Fibrinogen
|
0.259
|
1.15
|
0.010
|
0.02
|
-0.011
|
NLR
|
B.
P-value
|
T
|
Std. Error
|
95% Conf. (±)
|
Coefficient
|
|
|
|
|
|
|
Constant
|
0.001
|
4.58
|
0.005
|
0.01
|
0.024
|
MELD score
|
0.001
|
3.21
|
0.0042
|
0.009
|
0.013
|
CLIF-C ACLF
|
The correlation between the MELD score, and CLIF-C score or Child-Pugh scores with the development of ACLF
Severity of liver failure scores were used to predict mortality. We wanted to assess the ability of these scores to predict development of acute or chronic liver failure. The MELD score, especially above 18, significantly correlated with development of ACLF (Fig. 1A&B) more than both CLIF-C and Child-Pugh scores (Fig. 1A&B).
Association between liver failure scores, inflammatory markers and mortality in patients with liver disease
In addition to gender and age, liver failure scores and inflammatory markers were studied for their correlation to mortality. The results show that the MELD and PADUA scores significantly predict mortality in the liver patients suffering from cirrhosis (Table 3A). Furthermore, multivariate analysis showed that both MELD and PADUA scores were significantly and independently associated with mortality (Table 3B).
Table 3. Association between liver faiulre scores, inflammatory markers and mortality. A: Univariate analysis of the risk factors’ strength with mortality. B: Multivariate analysis of the risk factors’ strength with mortality. SE are the standard errors of the regression coefficients. T is the quotient of the coefficient. P-values are two-sided p values or observed significance levels.
A.
P-value
|
T
|
Std. Error
|
95% Conf. (±)
|
Coefficient
|
|
|
|
|
|
|
Constant
|
0.3
|
0.97
|
0.003
|
0.006
|
0.003
|
Age
|
0.9
|
0.08
|
0.054
|
0.108
|
0.005
|
Gender
|
0.001
|
4.78
|
0.0049
|
0.009
|
0.028
|
MELD score
|
0.12
|
1.5
|
0.017
|
0.03
|
0.026
|
CHILD PUGH score
|
0.3
|
0.895
|
0.004
|
0.009
|
0.004
|
CLIF-C ACLF
|
0.01
|
2.3
|
0.0167
|
0.030
|
0.036
|
PADUA
|
0.11
|
1.5
|
0.00064
|
0.001
|
0.0009
|
CRP
|
0.44
|
0.77
|
0.0078
|
0.015
|
0.0060
|
NLR
|
B.
P-value
|
T
|
Std. Error
|
95% Conf. (±)
|
Coefficient
|
|
|
|
|
|
|
Constant
|
0.0001
|
6.39
|
0.003
|
0.007
|
0.025
|
MELD score
|
0.0001
|
3.82
|
0.014
|
0.027
|
0.053
|
PADUA
|
Patient severity in the ACLF group
Patients in the ACLF group had different liver faliure severity. The CLIF-C score had a high accuracy in predicting 28 day mortality, particullary when it was calculated at 48 hours after ACLF diagnosis. CLIF-C scores were higher in the non-ACLF group as seen in figure 1. The ACLF-C score up to 45 was more frequent in the non-ACLF group, but above 45, the ACLF group was more freqent (Fig. 2A). The majority of the patients who suffered from ACLF were with low CLIF-C scores (<45), and few of them had higher scores (>60) (Fig 2B). This group showed a high prevlance of mortality. In the surviving patients, the majority had sepsis followed by acute gastrointestinal bleeding (GIB) then idopathic and finally acute kidney injuries.
Infalmmation marker levels in ACLF and in non-ACLF patients
The liver faliure severity mediated changes in infalmmatory marker levels. CRP and NLR were highly elevated in ACLF, but only moderatly elevated in decompsation liver faliure without ACLF (Fig. 3 A,B). The CRP level is correlated with the patient severity. Patients with high CLIF-C scores had high CRP levels, and low CLIF-C scores has low CRP levels (Fig. 3C).
Discriminant analysis with diagnostic accuracy of the correlations between clinical parameters and survival
The PADUA score, creatinine levels and MAP were able to discriminate between deadand alive patients with an 82% diagnostic accuracy. There is a good correlation between the PADUA score, creatinine and MAP, and disease severity and mortality in the cirrhotic patients. The sensitivity, specificity, positive predictive value and negative predictive value are 97%, 55%, 97% and 63%, respectively (Table 4). In addition, PADUA, MELD, CLIF-C scores and BUN, highly predict mortality (87%) in all cirrhotic patients. The sensitivity, specificity, positive predictive value and negative predictive value are 95%, 63%, 88% and 83%, respectively (Table 5).
Table 4: The validity (predictive power) of age, CRP, CRE and PADUA scores on patient mortality. The accuracy of the calculations is 82%. (A): The number of samples: Predicted condition — 129 survival; 35 death; and true condition, 0 for disease and 1 for no disease. (B): The sensitivity, specificity, positive predictive value and negative predictive value are shown.
A
|
Actual count
|
0
|
1
|
132
|
114
|
18
|
35
|
12
|
23
|
B
|
Specificity
|
55%
|
Sensitivity
|
97%
|
Positive predictive value
|
93%
|
Negative predictive value
|
63%
|
Table 5: The validity (predictive power) of BUN, MELD score, CLIF-C and PADUA score on mortality in cirrhotic patients. The accuracy of the calculations is 87%. (A): The number of samples: Predicted condition — 96 survival; 24 death; and true condition, 0 for disease and 1 for no disease. (B): The sensitivity, specificity, positive predictive value and negative predictive value are shown.
A
|
Actual count
|
0
|
1
|
96
|
84
|
12
|
24
|
4
|
20
|
B
|
Specificity
|
63%
|
Sensitivity
|
95%
|
Positive predictive value
|
88%
|
Negative predictive value
|
83%
|