Comparison of Prediction Models for Acute Kidney Injury Among Patients with Hepatobiliary Malignancies Based on XGBoost and LASSO-Logistic Algorithms

Figures & data

Table 1 Demographic and Clinical Factors of AKI in Patients with Hepatobiliary Malignancies

Variables	AKI	Non-AKI	Stat. Value	p-value
Age (yr)	60.3±11.7	57.9±11.3	−5.731	<0.001^a
Gender (male)	621 (78.4)	4970 (82.1)	6.354	0.012^b
BMI	22.9±3.4	23.3±3.4	2.895	0.004^a
Hypertension (yes, %)	136 (17.2)	655 (10.8)	27.656	<0.001^b
Diabetes (yes, %)	59 (7.4)	455 (7.5)	0.004	0.947^b
Stroke (yes, %)	11 (1.4)	21 (0.3)	—	0.001^c
CHD (yes, %)	10 (1.3)	35 (0.6)	—	0.030^c
COPD (yes, %)	13 (1.6)	30 (0.5)	—	0.001^c
CKD (yes, %)	97 (12.2)	99 (1.6)	283.620	<0.001^c
Metastatic tumors (yes, %)	83 (12.0)	609 (88.0)	0.136	0.712^b
Emergent condition (yes, %)	122 (21.0)	459 (79.0)	55.179	<0.001^b
Treat (surgery, %)	163 (20.6)	326 (5.4)	262.733	<0.001^b
Treat (chemotherapy, %)	364 (46.0)	2771 (45.8)	—	—
Treat (interventional, %)	257 (32.4)	2820 (46.6)	—	—
Treat (palliative, %)	8 (1.0)	137 (2.3)	—	—
SBP (mmHg)	123.2±15.3	125.8±15.5	4.515	<0.001^a
DBP (mmHg)	76.4±9.4	78.1±9.5	4.716	<0.001^a
Glucose (mmol/L)	6.0 [5.0, 7.2]	5.7 [4.9, 6.9]	—	0.002^d
ALT (U/L)	49.0 [25.0, 85.0]	42.0 [24.0, 68.0]	—	<0.001^d
AST (U/L)	52.0 [35.0, 91.0]	45.0 [30.0, 73.0]	—	<0.001^d
TBiL (μmol/L)	15.8 [9.2, 29.8]	13.0 [8.7, 22.7]	—	<0.001^d
BUN(mmol/L)	5.3 [4.0, 8.1]	4.5 [3.7, 5.5]	—	<0.001^d
SCr (μmol/L)	93.8±46.1	70.6±18.3	−26.315	<0.001^a
eGFR (mL/min/1.73m^2)	85.7±33.1	106.7±26.5	20.326	<0.001^a
SUA (μmol/L)	336.8±132.2	291.2±88.2	−12.802	<0.001^a
Album (g/L)	35.7±6.0	37.2±5.3	7.318	<0.001^a
Hemoglobin (g/L)	115.0±22.3	125.5±20.5	13.362	<0.001^a
Leukocyte (*10^9)	5.7 [4.0, 9.1]	5.1 [3.7, 7.1]	—	<0.001^d
Sodium (mmol/L)	137.3±4.9	139.1±3.9	12.030	<0.001^a
Potassium (mmol/L)	3.9±0.6	3.9±0.5	2.136	0.033^a
Chloride (mmol/L)	99.5±5.4	101.5±4.2	11.726	<0.001^a
Calcium (mmol/L)	2.2±0.1	2.3±0.1	9.700	<0.001^a
Magnesium (mmol/L)	0.9±0.1	0.9±0.1	−5.248	<0.001^a
Phosphorus (mmol/L)	1.0±0.2	1.0±0.2	2.035	0.042^a
CO2 (mmol/L)	23.0±3.6	23.9±3.0	7.687	<0.001^a
Anion gap (mmol/L)	15.0±3.6	13.7±2.9	−10.489	<0.001^a

Notes: ^aStudent’s t-test; ^bPearson chi-square test; ^cFisher’s exact test; ^dWilcoxon test.

Abbreviations: AKI, acute kidney injury; BMI, body mass index; CHD, coronary heart disease; COPD, chronic obstructive pulmonary disease; CKD, chronic kidney injury; SBP, systolic blood pressure; DBP, diastolic blood pressure; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TBiL, total bilirubin; BUN, blood urea nitrogen; SCr, serum creatinine; eGFR, estimated glomerular filtration rate; SUA, serum uric acid.

Figure 1 (A) Importance matrix plot of AKI predictors in XGBoost model among liver cancer patients. (B) Importance matrix plot of AKI predictors in XGBoost model among gallbladder cancer patients.

Table 2 Predictive Variables of AKI Selected by LASSO-Logistic Regression

Variables	Liver Cancer			Gall Bladder Cancer
	OR	95% CI	p-value	OR	95% CI	p-value
Age (yr)	—	—	—	1.03	0.99~1.06	0.127
Gender (male)	0.46	0.35~0.61	<0.001	—	—	—
BMI	—	—	—	0.87	0.79~0.96	0.005
CKD (yes, %)	—	—	—	2.21	0.58~8.37	0.245
Stroke (yes)	2.88	1.19~6.94	0.019	—	—	—
COPD (yes)	3.75	1.75~8.06	0.001	—	—	—
Metastatic Tumors (yes, %)	—	—	—	1.84	0.77~4.43	0.171
Emergent Condition (yes)	1.47	1.11~1.97	0.008	2.64	1.13~6.16	0.025
Treat (surgery)	6.42	2.84~14.51	<0.001	6.22	0.82~47.15	0.077
Treat (chemotherapy)	1.44	0.66~3.15	0.367	0.50	0.07~3.57	0.491
Treat (interventional)	1.61	0.74~3.53	0.233	1.71	0.19~15.13	0.631
SBP (mmHg)	0.99	0.99~1.00	0.003	—	—	—
AST (U/L)	—	—	—	1.00	1.00~1.00	0.450
TBiL (μmol/L)	1.01	1.00~1.01	<0.001	—	—	—
BUN(mmol/L)	1.05	1.00~1.10	0.034	1.04	0.93~1.16	0.465
SCr (μmol/L)	1.03	1.02~1.04	<0.001	1.00	1.00~1.01	0.335
eGFR (mL/min/1.73m2)	0.99	0.99~1.00	0.021	0.99	0.98~1.00	0.047
SUA (μmol/L)	1.00	1.00~1.00	0.007	—	—	—
Album (g/L)	—	—	—	0.93	0.87~1.00	0.040
Hemoglobin (g/L)	0.99	0.99~1.00	<0.001	0.99	0.97~1.00	0.110
Sodium (mmol/L)	0.94	0.91~0.97	<0.001	0.92	0.86~0.98	0.007
Potassium (mmol/L)	0.77	0.64~0.92	0.004	—	—	—
Chloride (mmol/L)	0.97	0.94~1.00	0.058	—	—	—
Calcium (mmol/L)	0.61	0.24~1.54	0.297	—	—	—
Phosphorus (mmol/L)	—	—	—	0.16	0.05~0.50	0.002
Anion gap (mmol/L)	1.04	1.01~1.08	0.012	1.12	1.01~1.24	0.030

Abbreviations: AKI, acute kidney injury; BMI, body mass index; CKD, chronic kidney injury; COPD, chronic obstructive pulmonary disease; SBP, systolic blood pressure; AST, aspartate aminotransferase; TBiL, total bilirubin; BUN, blood urea nitrogen; SCr, serum creatinine; eGFR, estimated glomerular filtration rate; SUA, serum uric acid.

Table 3 External Validation of AKI Prediction Models Based on XGBoost and LASSO-Logistic Algorithms

Model Index	Liver Cancer		Gallbladder Cancer
	XGBoost	LASSO-logistic	XGBoost	LASSO-logistic
Sensitivity	66.3%	62.3%	78.3%	69.6%
Specificity	81.2%	79.3%	81.1%	63.2%
False positive rate	18.8%	20.7%	18.9%	36.8%
False negatively rate	33.7%	37.7%	21.7%	30.4%
Accuracy rate	79.5%	77.5%	80.5%	64.4%
Youden’s index	47.5%	41.6%	59.3%	32.7%
Positive predictive value	30.3%	27.1%	50.0%	31.4%
Negative predictive value	95.1%	94.5%	93.9%	89.6%

Notes: Sensitivity is the positively classified cases divided by the positive cases TP/(TP+FN). Specificity is the negatively classified cases divided by the negative cases TN/(FP+TN). False-positive rate (FPR) is the false positively classified cases divided by the negative cases FP/(FP+TN). False negatively rate (FNR) is the false negatively classified cases divided by the positive cases FN/(TP+FN). Accuracy rate is the sum of correctly classified cases test divided by the data set size (TP+TN)/(TP+TN+FP+FN). Youden’s index is the difference between specificity and false-positive rate TP/(TP+FN)–FP/(FP+TN). Positive predictive value (PPV) is positive cases that are true positives TP/(TP+FP). Negative predictive value (NPV) is the proportion of negative cases that are true negatives TN/(TN+FN).

Figure 2 (A) Receiver operating characteristic curves of AKI prediction models among liver cancer patients; (B) receiver operating characteristic curves of AKI prediction models among gallbladder cancer patients; (C) left chart of AKI prediction models among liver cancer patients; (D) left chart of AKI prediction models among gallbladder cancer patients; (E) learning curve of AKI prediction models among liver cancer patients; (F) learning curve of AKI prediction models among gallbladder cancer patients.