Risk prediction of biliary infection after endoscopic drainage for malignant perihilar biliary obstruction: A 10-year multicenter retrospective study

Table 1 Baseline demographic and clinicopathological characteristics, n (%)

Characteristics	Modeling group (n = 358)	Validation group (n = 45)	χ2/Z value	P value
Sex
Male	244 (68.2)	29 (64.4)	0.197	0.657
Female	114 (31.8)	16 (35.6)
Age ≥ 65	241 (67.3)	26 (57.8)	1.860	0.173
Chemotherapy	28 (7.8)	6 (13.4)	2.400	0.200
Prior endoscopic intervention	144 (40.2)	17 (37.8)	0.055	0.815
Diabetes	63 (17.6)	7 (15.6)	0.176	0.675
Hypertension	113 (31.6)	17 (37.8)	2.831	0.150
Antibiotic prophylaxis	165 (16.1)	20 (44.4)	3.759	0.053
Plastic stent placement	254 (70.9)	32 (71.1)	0.001	0.977
Metal stent placement	11 (3.1)	2 (4.4)	1.208	0.272
ENBD	179 (50.0)	20 (44.4)	1.033	0.309
Number of stents
0	88 (24.6)	10 (22.2)	0.856	0.836
Single	167 (46.6)	22 (48.9)
Multiple	103 (28.8)	13 (28.9)
EST	105 (29.3)	24 (53.3)	0.440	0.507
RFA	14 (3.9)	6 (13.3)	0.156	0.693
Bismuth-Corlett classification
I	153 (42.7)	23 (51.1)	86.004	0.289
II	135 (37.7)	14 (31.1)
IIIa	21 (5.9)	4 (8.9)
IIIb	14 (3.9)	1 (2.2)
IV	35 (9.8)	3 (6.7)
Transfusion history	18 (5.0)	5 (11.1)	4.803	0.058
Hypokalemia	57 (15.9)	9 (20.0)	27.802	0.351
BMI (kg/m2), median (IQR)	20.12 (18.18-21.96)	21.10 (18.18-22.72)	-2.167	0.301
Total procedure time (minute), median (IQR)	40.00 (25.00-50.00)	35.00 (23.00-45.00)	1.232	0.218
ANC (× 109/L), median (IQR)	5.66 (4.30-7.80)	5.10 (3.76-7.00)	2.213	0.207
NLR, median (IQR)	6.14 (3.80-10.88)	5.00 (3.25-8.71)	2.501	0.102
TBIL (μmol/L), median (IQR)	205.40 (116.00-339.50)	166.60 (62.80-283.00)	2.528	0.124
DBIL (μmol/L), median (IQR)	155.40 (74.42-253.40)	126.10 (47.00-217.20)	2.516	0.052
AST (U/L), median (IQR)	106.38 (60.49-189.85)	80.00 (52.20-124.21)	2.746	0.056

ANC: Absolute neutrophil count; AST: Aspartate aminotransferase; BMI: Body mass index; DBIL: Direct bilirubin; ENBD: Endoscopic nasobiliary drainage; EST: Endoscopic sphincterotomy; IQR: Interquartile range; NLR: Neutrophil-lymphocyte ratio; RFA: Radiofrequency ablation; TBIL: Total bilirubin.

Table 2 Univariate and multivariate logistic regression analyses of the modeling group

Variables	Univariable analysis		Multivariable analysis
	OR (95%CI)	P value	OR (95%CI)	P value
Chemotherapy	0.327 (0.180-6.115)	0.037	0.415 (0.120-1.440)	0.166
Antibiotic prophylaxis	1.255 (0.457-3.448)	0.029	1.041 (0.596-1.820)	0.887
Bismuth-Corlett classification	1.873 (1.258-2.789)	< 0.001	1.412 (1.144-1.743)	0.001
Hypokalemia	2.415 (0.655-8.899)	< 0.001	4.080 (1.958-8.505)	< 0.001
BMI	0.855 (0.151-4.846)	0.025	0.930 (0.863-1.003)	0.059
ANC	2.450 (0.537-11.184)	0.018	1.066 (0.970-1.171)	0.182
NLR	0.997 (0.916-1.086)	0.008	1.007 (0.967-1.048)	0.745
TBIL	1.001 (0.989-1.013)	0.008	1.001 (0.994-1.009)	0.759
DBIL	1.004 (0.986-1.022)	0.006	0.999 (0.998-1.010)	0.888
AST	1.001 (0.994-1.009)	0.005	1.003 (1.000-1.006)	0.021
Serum lipase	0.998 (0.994-1.002)	< 0.001	1.000 (0.999-1.000)	0.226

ANC: Absolute neutrophil count; AST: Aspartate aminotransferase; BMI: Body mass index; CI: Confidence interval; DBIL: Direct bilirubin; NLR: Neutrophil-lymphocyte ratio; OR: Odds ratio; TBIL: Total bilirubin.

Table 3 Predictive performance metrics of the prediction models

	Modeling group		Validation group
	LR	ANN	LR	ANN
AUC	0.734	0.791	0.867	0.940
95%CI	0.655-0.813	0.710-0.872	0.812-0.922	0.928-0.952
Accuracy (%)	71.4	80.2	79.7	84.1
Specificity (%)	79.6	89.8	91.8	93.1
Sensitivity (%)	28.1	46.9	33.3	55.6

ANN: Artificial neural network; AUC: Area under the curve; CI: Confidence interval; LR: Logistic regression.