Transjugular intrahepatic portosystemic shunt vs conservative treatment for recurrent ascites: A propensity score matched comparison

Table 1 Baseline characteristics at hospital admission (all patients)

Characteristics	No TIPS	TIPS	P value
Patients	398	214
Male	269 (68%)	153 (71%)	0.320
Age (yr)	59.5 (26.4-93.4)	59.1 (29.9-80.7)	0.190
Cause of cirrhosis			0.130
-Alcohol	305 (77%)	179 (84%)
-Viral hepatitis	11 (3%)	4 (2%)
-Other	82 (21%)	31 (14%)
Child points (min-max)	10 (7-15)	9 (7-14)	< 0.001
Hepatic encephalopathy (West-Haven)			0.040
-None	276 (69%)	165 (77%)
-Grade 1-2	73 (18%)	30 (14%)
-Grade 3-4	49 (12%)	19 (9%)
ACLF grade at hospital admission			0.020
-No ACLF	294 (74%)	173 (81%)
-ACLF grade 1	64 (16%)	39 (18%)
-ACLF grade 2	32 (8%)	1 (0.5%)
-ACLF grade 3	8 (2%)	1 (0.5%)
Laboratory findings
-Meld	18 (7-40)	14 (7-40)	< 0.001
-Total bilirubin (μmol/L)	49.7 (5.9-668.0)	28.1 (6.1-688.5)	< 0.001
-Creatinine (μmol/L)	100.5 (23.1-781.0)	107.0 (42.3-783.5)	0.180
-INR	1.45 (0.92-9.2)	1.28 (0.97-2.5)	< 0.001
-Sodium (μmol/L)	134 (106-149)	133 (115-146)	0.760
-Albumin (g/L)	22.6 (7.9-48.7); NA: 55	26.1 (11.0-39.6); NA: 23	< 0.001
-CRP (mg/L)	25.6 (1.0-283.0)	18.0 (2.0-181.0)	< 0.001
-Hemoglobin (mmol/L)	6.8 (2.0-9.8)	6.7 (3.4-10.0)	0.310
-Platelets (Gpt/L)	134.5 (22.0-715.0)	153.5 (13.0-668.0)	0.002
-Leucocytes (Gpt/L)	8.76 (1.34-44.90)	7.34 (2.72-33.20)	< 0.001

Continuous variables are given as median and range, categorical variables as total number and percentage. Continuous variables were compared using the Mann-Whitney-U-test and categorical variables using the chi-square test. NA: Not available; ACLF: Acute on chronic liver failure; TIPS: Transjugular intrahepatic portosystemic shunt.

Table 2 Baseline characteristics at hospital admission (matched groups)

Characteristics	No TIPS	TIPS	P value
Patients	214	214
Male	142 (66%)	153 (71%)	0.30
Age (yr)	59.4 (26.4-93.4)	59.1 (29.9-80.7)	0.14
Cause of cirrhosis			0.26
-Alcohol	163 (76%)	179 (84%)
-Viral hepatitis	8 (4%)	4 (2%)
-Other	43 (20%)	31 (14%)
Child points (min-max)	9 (7-14)	9 (7-14)	0.76
Hepatic encephalopathy (west-haven)			0.65
-None	173 (81%)	165 (77%)
-Grade 1-2	22 (10%)	30 (14%)
-Grade 3-4	19 (9%)	19 (9%)
ACLF grade at hospital admission			0.37
-No ACLF	176 (82%)	173 (81%)
-AACLF 1	36 (17%)	39 (18%)
-ACLF 2	2 (1%)	1 (0.5%)
-ACLF 3	0 (0%)	1 (0.5%)
Laboratory findings
-Meld	14 (7-36)	14 (7-40)	0.97
-Total bilirubin (μmol/L)	29.7 (5.9-261.0)	28.1 (6.1-688.5)	0.10
-Creatinine (μmol/L)	90.9 (23.1-781.0)	107.0 (42.3-783.5)	0.08
-INR	1.32 (0.92-2.41)	1.28 (0.97-2.50)	0.28
-Sodium (mmol/L)	135 (106-144)	133 (115-146)	0.10
-Albumin (g/L)	24.3 (7.9-48.7); NA: 27	26.1 (11.0-39.6); NA: 23	0.61
-CRP (mg/L)	18.7 (1.0-283.0)	18.0 (2.0-181.0)	0.10
-Hemoglobin (mmol/L)	6.8 (2.1-9.8)	6.7 (3.4-10.0)	0.43
-Platelets (Gpt/L)	141.0 (23.0-715.0)	154.0 (13.0-668.0)	0.08
-Leucocytes (Gpt/L)	7.58 (1.34-44.90)	7.34 (2.72-33.20)	0.98

Continuous variables are given as median and range, categorical variables as total number and percentage. Continuous variables were compared using Wilcoxon signed-rank test and categorical variables using McNemar-Test. NA: Not available; ACLF: Acute on chronic liver failure; TIPS: Transjugular intrahepatic portosystemic shunt.

Table 3 Changes of acute on chronic liver failure grade during hospital stay and in-hospital mortality (matched groups)

Event	No TIPS	TIPS	P value/OR (95%CI)
Hospital stay (d)	10 (1-78)	14 (3-64)	P < 0.001
Highest ACLF grade			P = 0.041
-No ACLF	157 (73%)	144 (67%)
-ACLF 1	47 (22%)	50 (23%)
-ACLF 2	8 (4%)	15 (7%)
-ACLF 3	2 (1%)	5 (2%)
-Any ACLF	57 (27%)	70 (33%)
Mortality by ACLF
-Over all	13/214 (6.1%)	11/214 (5.1%)	OR: 0.84 (0.33 -2.08)
-No ACLF	3/157 (1.9%)	0/144 (0%)	OR: 0 (0.00 -2.63)
-ACLF 1	3/47 (6.4%)	4/50 (8%)	OR: 1.27 (0.20 -9.18)
-ACLF 2	6/8 (75%)	3/15 (20%)	OR: 0.09 (0.01 -0.87)
-ACLF 3	1/2 (50%)	4/5 (80%)	OR: 3.16 (0.03 -389.17)
-Any ACLF	10/57 (17.5%)	11/70 (15.7%)	OR: 0.88 (0.31-2.52)
Increase in ACLF grade			P = 0.03
-No increase	191 (89.3%)	176 (82.2%)
-1 grade	18 (8.4%)	28 (13.1%)
-2 grades	4 (1.9%)	7 (3.3%)
-3 grades	1 (0.5%)	3 (1.4%)
Mortality by ACLF increase
-No increase	5/191 (2.6%)	2/176 (1.1%)	OR: 0.43 (0.04-2.66)
-1 grade	4/18 (22.2%)	5/28 (17.9%)	OR: 0.77 (0.14-4.55)
-2 grades	3/4 (75.0%)	2/7 (20.0%)	OR: 0.16 (0.003-3.50)
-3 grades	1/1 (100%)	2/3 (66.7%)	OR: 0 (0.00-116.8)
-Any increase	8/23 (34.8%)	9/38 (23.7%)	OR: 0.58 (0.16-2.14)

OR: Odds ratio; ACLF: Acute on chronic liver failure; TIPS: Transjugular intrahepatic portosystemic shunt.

Table 4 Sensitivity analysis: Multivariate regressions (main effects only)

Variable	Estimate	SE	P value	Estimate	SE	P value
	Complete model			Best model
Mortality
Intercept	3.63	3.27	0.268	4.39	3.22	0.173
Creatinine	1.59 × 10-3	1.37 × 10-3	0.246	1.99 × 10-3	1.32 × 10-3	0.132
Bilirubin	7.26 × 10-4	1.10 × 10-3	0.505	-	-	-
INR	3.49 × 10-1	2.15 × 10-1	0.104	3.47 × 10-1	2.12 × 10-1	0.101
CRP	3.27 × 10-3	3.10 × 10-3	0.292	-	-	-
Leucocytes	7.01 × 10-2	2.64 × 10-2	0.008	7.84 × 10-2	2.53 × 10-2	0.002
HE 1-2	-2.88 × 10-1	4.28 × 10-1	0.500	-2.67 × 10-1	4.1 × 10-1	0.523
HE 3-4	2.24	3.56 × 10-1	< 0.001	2.26	3.50 × 10-1	< 0.001
Albumin	-9.61 × 10-2	2.76 × 10-2	< 0.001	-1.02 × 10-1	2.72 × 10-2	< 0.001
Sodium	-6.21 × 10-2	2.45 × 10-2	0.011	-6.51 × 10-2	2.45 × 10-2	0.004
Age	1.22 × 10-4	4.24 × 10-5	0.004	1.16 × 10-4	4.04 × 10-5	0.004
TIPS	-7.29 × 10-1	4.12 × 10-1	0.077	-8.22 × 10-1	4.01 × 10-1	0.040
ACLF
Intercept	-1.822	2.713	0.502	-2.70	8.51 × 10-1	0.002
Creatinine	-1.06 × 10-3	1.21 × 10-3	0.384	-	-	-
Bilirubin	2.81 × 10-3	9.50 × 10-4	0.003	2.99 × 10-3	8.69 × 10-4	0.001
INR	2.16 × 10-1	2.01 × 10-1	0.281	-	-	-
CRP	5.21 × 10-3	2.58 × 10-3	0.043	4.67 × 10-3	2.39 × 10-3	0.050
Leucocytes	6.28 × 10-3	2.43 × 10-2	0.780	-	-	-
HE 1-2	1.22 × 10-1	3.06 × 10-1	0.690	1.47 × 10-1	3.02 × 10-1	0.627
HE 3-4	1.62	3.01 × 10-1	< 0.001	1.63	2.94 × 10-1	< 0.001
Albumin	-3.80 × 10-2	2.00 × 10-2	0.058	-3.98 × 10-2	1.99 × 10-2	0.046
Sodium	-1.03 × 10-2	1.98 × 10-2	0.603	-	-	-
Age	6.39 × 10-5	3.11 × 10-5	0.039	5.63 × 10-5	2.96 × 10-5	0.057
TIPS	5.17 × 10-1	2.64 × 10-1	0.050	4.39 × 10-1	2.55 × 10-1	0.085

Dependent variables were in-hospital mortality (upper panel) and any increase in acute on chronic liver failure grade (lower panel). The full models (left side) included all parameters used for propensity score matching as covariates. After stepwise backward elimination by Akaike information criterion, a model (best model, right side) was selected for each dependent variable. ACLF: Acute on chronic liver failure; TIPS: Transjugular intrahepatic portosystemic shunt.

Table 5 Multivariate logistic regressions with interaction terms

Model	Dependent variable	Parameters	Estimate	SE	z value	P value
A		Intercept	-3.8600	0.4340	-8.887	< 2 × 10-16
	In-hospital	MELD-Score	0.0990	0.0180	5.628	1.82 × 10-8
	Mortality (y/n)	TIPS	-1.3570	1.0590	-1.281	0.200
		MELD: TIPS	0.0330	0.0500	0.668	0.504
B		Intercept	-7.1320	0.9810	-7.271	3.56 × 10-13
	In-hospital	Child (points)	0.4880	0.0840	5.814	6.09 × 10-9
	Mortality (y/n)	TIPS	-1.6760	2.2350	-0.750	0.453
		Child: TIPS	0.0934	0.2020	0.463	0.643
C		Intercept	-3.1860	0.3610	-8.824	< 2 × 10-16
	Increase in	MELD	0.1020	0.0160	6.461	1.04 × 10-10
	ACLF grade	TIPS	-0.7780	0.7120	-1.092	0.275
	(y/n)	MELD: TIPS	0.0480	0.0368	1.318	0.187
D		Intercept	-5.2640	0.7480	-7.040	1.93 × 10-12
	Increase in	Child (points)	0.3880	0.0670	5.807	6.37 × 10-9
	ACLF grade	TIPS	-3.1980	1.5300	-2.090	0.0366
	(y/n)	Child: TIPS	0.3190	0.1145	2.191	0.0285

For models C and D death was treated as an increase in acute on chronic liver failure. Models A and B show an effect of only the MELD/Child scores on mortality. Transjugular intrahepatic portosystemic shunt (TIPS) and the interaction of TIPS and MELD/Child scores (MELD: TIPS, Child: TIPS) have no significant influence on mortality (A and B). In model D a significant interaction term Child:TIPS exists. In model C the interaction term MELD: TIPS is not significant, indicating a weaker interaction than in model D. ACLF: Acute on chronic liver failure; TIPS: Transjugular intrahepatic portosystemic shunt.