Risk factors and predictive modeling of early postoperative liver function abnormalities

Abstract

BACKGROUND

Research has shown that several factors can influence postoperative abnormal liver function; however, most studies on this issue have focused specifically on hepatic and cardiac surgeries, leaving limited research on contributing factors in other types of surgeries.

AIM

To identify the risk factors for early postoperative abnormal liver function in multiple surgery types and construct a risk prediction model.

METHODS

This retrospective cohort study involved 3720 surgical patients from 5 surgical departments at Guangdong Provincial Hospital of Traditional Chinese Medicine. Patients were divided into abnormal (n = 108) and normal (n = 3612) groups based on liver function post-surgery. Univariate analysis and LASSO regression screened variables, followed by logistic regression to identify risk factors. A prediction model was constructed based on the variables selected via logistic regression. The goodness-of-fit of the model was evaluated using the Hosmer–Lemeshow test, while discriminatory ability was measured by the area under the receiver operating characteristic curve. Calibration curves were plotted to visualize the consistency between predicted probabilities and observed outcomes.

RESULTS

The key factors contributing to abnormal liver function after surgery include elevated aspartate aminotransferase and alanine aminotransferase levels and reduced platelet counts pre-surgery, as well as the sevoflurane use during the procedure, among others.

CONCLUSION

The above factors collectively represent notable risk factors for postoperative liver function injury, and the prediction model developed based on these factors demonstrates strong predictive efficacy.

Key Words: Perioperative period; Abnormal liver function; Risk factor; Univariate analysis; Risk prediction model

Core Tip: Elevated preoperative liver enzymes, thrombocytopenia, prolonged anesthesia and sevoflurane use are important risk factors for postoperative liver function abnormalities, and the prediction model constructed on the basis of these variables predicted well.

INTRODUCTION

The liver is a vital metabolic organ responsible for synthesizing, degrading, transforming, and storing various substances, playing a crucial role in detoxification and waste excretion. Postoperative hepatic dysfunction refers to abnormal changes in liver function after surgery, including elevated liver enzyme levels, jaundice, and coagulation abnormalities, triggering liver insufficiency or failure in severe cases. Elevated liver enzyme levels are associated with mortality and morbidity; indeed, patients with aspartate aminotransferase (AST) levels > 18 U/L had a threefold increased risk of all-cause mortality[1]. Furthermore, compared to men with AST or alanine aminotransferase (ALT) levels < 20 IU/L, those in the 30–39 IU/L group have relative risks of liver-related death that are 8- and 9.5-fold higher for AST and ALT, respectively[2]. Various preoperative, intraoperative, and postoperative factors influence postoperative liver function. For example, carbon dioxide insufflation during laparoscopic surgery, thermal injury from diathermy, and liver compression—which causes injury to the hepatic arterial branches—can lead to a transient increase in liver transaminases[3]. Perioperative blood transfusion can also result in the transmission of viral hepatitis. However, in most countries, blood donors are not routinely screened for the hepatitis E virus. A 2018 study reported a prevalence of 0.12% and 0.28% for hepatitis E among donors in Germany and China, respectively[4,5], and identified transfusion-transmitted viral hepatitis as a cause of postoperative hepatic dysfunction. Inhalational halogenated agents, commonly used as anesthetics, can cause drug-induced liver injury[6]. Volatile anesthetic-induced liver injury typically causes an increase in ALT levels 2–14 days after surgery. It manifests as asymptomatic hepatic dysfunction[7,8], most commonly characterized by increased levels of liver enzymes, particularly AST, ALT, alkaline phosphatase (ALP), and gamma-glutamyl transferase (GGT)[9,10].

In conclusion, while numerous factors influence postoperative liver function abnormalities, most studies have concentrated on liver and cardiac surgeries, with limited research on the incidence and causes of liver function abnormalities in other types of surgery. Therefore, this study aimed to identify the factors contributing to postoperative liver function abnormalities and to develop a predictive model to prevent them.

MATERIALS AND METHODS

Patient enrollment

This retrospective study included patients who underwent elective tracheal intubation under general anesthesia for surgery in the Departments of Gynecology, Gastrointestinal Surgery, Thyroid Surgery, Urology, and Orthopedics at Guangdong Provincial Hospital of Traditional Chinese Medicine between January 2018 and December 2023. The inclusion criteria were as follows: (1) Age > 18 years; (2) Elective tracheal intubation under general anesthesia for surgery (including gynecological, gastrointestinal, thyroid, urological, and orthopedic surgeries); (3) Operative time > 3 hours; and (4) American Society of Anesthesiologists (ASA) classification I–IV. Patients were excluded if they (1) Underwent hepatic, gallbladder, cardiac, or secondary surgery; (2) Had emergent conditions; or (3) Had > 30% missing data. This study enrolled 3720 patients, see Figure 1 for details. The detailed protocol was approved by the Ethics Committee of Guangdong Provincial Hospital of Traditional Chinese Medicine (approval No. YE2024-041-01), which waived the requirement for informed consent because of the retrospective nature of this study.

Figure 1 Flow chart. VIF: Variance inflation factor.

Data collection

Patients’ demographic data, comorbidities, preoperative laboratory test results, preoperative vital signs, intraoperative anesthetic medications, fluid administration, and urine output were collected from electronic medical records and the anesthesia management systems (Do-Care version 5.0). Demographic data and comorbidity information included sex, age, body mass index, ASA classification, and medical history. The primary outcome was postoperative hepatic dysfunction, as defined by the most recent postoperative liver function test results. Univariate analysis and multivariate logistic regression analysis were applied to investigate the independent risk factors of early postoperative liver dysfunction. Predictive models were developed based on independent risk factors, and their predictive value was assessed using receiver operating characteristic (ROC) curves and calibration curves.

Outcomes

Abnormal liver function was assessed in patients based on the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0: (1) If preoperative liver function at baseline was normal (7 < ALT levels < 40 and 13 < AST levels < 35), then any of the following indicators were considered abnormal postoperatively: ALT ≥ 120 (three times the upper limit of normal value) and AST ≥ 105 (three times the upper limit of normal value); and (2) If the preoperative liver function at baseline was abnormal, then any of the following indicators were considered abnormal postoperatively: ALT ≥ 1.5 × baseline value and AST ≥ 1.5 × baseline value.

Statistical analyses

Normal data were expressed as ± SD, non-normally distributed measurements as M (Q₁, Q₃), and categorical data as the number of cases (percentage). The t-test was employed for normal data, while the Mann–Whitney U rank sum test and χ² test were used for non-normally distributed measurements and count data, respectively. Variables with P < 0.05 were screened using variance inflation factor (VIF) analysis and LASSO regression. Finally, binary logistic regression was conducted for correlation analysis. All analyses were performed using SPSS 27.0 software. For hypothesis testing, we used a two-sided test with a significance level of α = 0.05. To address multicollinearity, variables with a VIF greater than 10 were deleted. In LASSO regression, Lambda was set to 0.00105, and% Deviance was 7.525.

RESULTS

Single-factor analysis

A total of 3720 patients were studied. Among them, 3553 (95.51%) had no postoperative liver function abnormalities, while 167 (4.49%) experienced such abnormalities. The differences between the two groups were statistically significant (P < 0.05) in terms of the length of surgery, anesthesia time, sevoflurane usage, and various laboratory values including AST, ALT, ALP, GGT, total bilirubin (TBIL), TP, ALB, PLT, HDLC, and Glu. Additionally, significant differences were observed regarding gender, ASA II, ASA III, ASA IV classifications, the use of remifentanil, types of surgeries (head and neck, spine and limbs, esophagus, gastrointestinal tract, pancreas, spleen, appendix, and surgeries of the uterus, ovary, and fallopian tube) (P < 0.05). In contrast, no statistically significant differences (P > 0.05) were found for rocuronium bromide, etomidate, intravenous push propofol dosage, dexmedetomidine, age, height, weight, Sufentanil, GLB, HGB, WBC, NEUT, RBC, TC, TG, LDLC, UA, UREA, CREAT, K, Na, Ca, Cl, eGFR, BMI, ASA I, the administration of propofol, surgeries on the kidney, bladder, and ureter, and the presence of diabetes mellitus, hypertension, chronic kidney disease, COPD, hyperlipidemia, chronic hepatitis, cirrhosis, viral hepatitis, cerebral infarction, and coronary artery disease (Tables 1, 2, and 3).

Table 1 General information of the normal and abnormal groups, n (%).

Parameter	Normal group (n = 3553)	Abnormal group (n = 167)	Statistic	P value
Gender, male	1037 (29.19)	88 (52.69)	χ² = 41.78	< 0.01
Age, year	56.00 (47.00, 66.00)	59.00 (50.00, 67.00)	Z = -1.51	0.13
Weight, kg	58.00 (52.00, 65.00)	59.00 (52.35, 66.50)	Z = -0.93	0.35
BMI, kg/m2	22.89 (20.62, 25.33)	22.66 (20.37, 25.56)	Z = -0.31	0.76
Past medical history
High blood pressure	1126 (31.69)	51 (30.54)	χ² = 0.10	0.75
Diabetes	2487 (70.00)	116 (69.46)	χ² = 0.02	0.88
Coronary heart disease	1480 (41.65)	72 (43.11)	χ² = 0.14	0.71
Cerebral infarction	90 (2.53)	2 (1.20)	χ² = 0.69	0.41
Viral hepatitis	211 (5.94)	13 (7.78)	χ² = 0.96	0.33
Chronic hepatitis	17 (0.48)	1 (0.60)	-	0.56
Liver cirrhosis	12 (0.34)	0 (0.00)	-	1.00
Hyperlipidemia	140 (3.94)	10 (5.99)	χ² = 1.73	0.19
COPD	11 (0.31)	1 (0.60)	-	0.42
Chronic kidney disease	91 (2.56)	6 (3.59)	χ² = 0.32	0.57

BMI: Body mass index; COPD: Chronic obstructive pulmonary disease.

Table 2 Results of laboratory testing analysis in the normal and abnormal groups.

Parameter, unit	Normal group (n = 3553)	Abnormal group (n = 167)	Statistic	P value
AST, U/L	18.00 (15.00, 22.00)	30.00 (20.00, 52.00)	Z = -12.42	< 0.01
ALT, U/L	15.00 (11.00, 21.00)	29.00 (17.00, 69.00)	Z = -10.83	< 0.01
ALP, g/L	70.00 (57.00, 87.00)	78.00 (63.64, 111.00)	Z = -4.78	< 0.01
GGT, U/L	19.00 (14.00, 30.00)	38.00 (21.00, 82.50)	Z = -9.77	< 0.01
Tbil, μmol/L	8.60 (6.00, 12.10)	10.30 (7.10, 15.60)	Z = -4.23	< 0.01
Total protein, g/L	71.30 (68.80, 74.60)	70.90 (67.05, 72.70)	Z = -3.35	< 0.01
Albumin, g/L	42.90 (40.20, 45.50)	41.50 (38.20, 44.45)	Z = -4.23	< 0.01
Globulin, g/L	27.20 (25.70, 30.40)	27.20 (25.50, 30.60)	Z = -0.45	0.65
Glu, mmol/L	5.31 (4.78, 6.36)	5.77 (5.00, 6.88)	Z = -3.11	< 0.01
HB, g/L	125.00 (109.00, 136.00)	125.00 (106.50, 140.50)	Z = -0.90	0.37
WBC, × 109/L	6.35 (5.27, 7.72)	6.37 (5.30, 7.94)	Z = -0.53	0.60
N, × 109/L	3.94 (3.07, 5.08)	3.92 (2.90, 5.13)	Z = -0.28	0.78
RBC, × 1012/L	4.33 (3.95, 4.68)	4.26 (3.74, 4.68)	Z = -1.85	0.06
PLT, × 1012/L	259.00 (215.00, 314.00)	232.00 (178.00, 278.00)	Z = -4.72	< 0.01
K+ , mmol/L	4.00 (3.80, 4.20)	4.01 (3.77, 4.21)	Z = -0.14	0.89
Na+ , mmol/L	140.00 (139.00, 142.00)	140.00 (139.00, 142.00)	Z = -1.20	0.23
Ca2+ , mmol/L	2.30 (2.30, 2.30)	2.30 (2.30, 2.30)	Z = -0.05	0.96
CL- , mmol/L	103.80 (102.10, 105.50)	103.70 (101.60, 105.40)	Z = -0.86	0.39
TC, mmol/L	4.71 (4.49, 4.91)	4.71 (4.46, 4.75)	Z = -0.42	0.67
Tg, mmol/L	1.35 (1.19, 1.45)	1.35 (1.29, 1.60)	Z = -1.29	0.20
LDL, mmol/L	3.06 (2.87, 3.20)	3.06 (2.79, 3.06)	Z = -1.14	0.25
HDL, mmol/L	1.18 (1.14, 1.26)	1.18 (0.99, 1.18)	Z = -2.96	< 0.01
Uric, mmol/L	350.00 (350.00, 350.00)	350.00 (350.00, 350.00)	Z = -0.74	0.46
EGFR, mmol/L	94.75 (81.14, 106.41)	93.72 (83.34, 104.34)	Z = -0.20	0.84
Urea, mmol/L	4.68 (3.72, 5.75)	4.81 (3.91, 5.81)	Z = -0.97	0.33
Cr, mmol/L	65.00 (56.00, 78.00)	69.00 (58.00, 83.00)	Z = -1.98	0.05

ALP: Alkaline phosphatase; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; EGFR: Epidermal growth factor receptor; GGT: Gamma-glutamyl transpeptidase; HB: Hemoglobin; HDL: High-density lipoprotein; LDL: Low-density lipoprotein; PLT: Platelets; RBC: Red blood cells; TBIL: Total bilirubin; WBC: White blood cells; Glu: Glucose; Cr: Creatinine.

Table 3 Intraoperative data analysis for the normal and abnormal groups.

Parameter	Normal group (n = 3553)	Abnormal group (n = 167)	Statistic	P value
Head and neck surgery	14 (0.39)	3 (1.80)	-	0.04
Operation on the spine and limbs	643 (18.10)	42 (25.15)	χ² = 5.28	0.02
Operation on the esophagus, gastrointestinal tract, pancreas, spleen, and appendix	907 (25.53)	88 (52.69)	χ² = 60.08	< .01
Operation on the kidneys, bladder, and ureter	218 (6.14)	8 (4.79)	χ² = 0.51	0.48
Operation on the uterus, ovaries, and fallopian tube	1771 (49.85)	26 (15.57)	χ² = 75.04	< 0.01
Time of operation	239.00 (205.00, 288.00)	290.00 (232.00, 359.00)	Z = -7.16	< 0.01
Anesthesia time, minute	295.00 (257.00, 346.00)	352.00 (290.00, 417.00)	Z = -7.50	< 0.01
Length of hospital stay, day	13.00 (9.00, 18.00)	16.00 (11.00, 22.00)	Z = -5.24	< 0.01
ASA rating
I	169 (4.76)	5 (2.99)	χ² = 1.11	0.29
II	2709 (76.25)	115 (68.86)	χ² = 4.76	0.03
III	671 (18.89)	45 (26.95)	χ² = 6.67	< 0.01
IV	4 (0.11)	2 (1.20)	-	0.03
Sufentanil, μg	10.00 (5.00, 15.00)	10.00 (5.00, 15.00)	Z = -1.31	0.19
Whether to pump remifentanil	3131 (88.12)	164 (98.20)	χ² = 16.02	< 0.01
Rocuronium, mg	2.49 ± 10.76	2.94 ± 10.66	t = -0.53	0.59
Etomidate, mg	6.88 ± 16.13	8.70 ± 19.38	t = -1.41	0.16
Whether to pump propofol	2681 (75.46)	119 (71.26)	χ² = 1.51	0.22
Dosage of intravenous propofol, mg	35.40 ± 153.54	30.48 ± 107.01	t = 0.41	0.68
Dexmedetomidine, μg	4.09 ± 12.96	3.92 ± 13.37	t = 0.17	0.87
Sevoflurane, mL	60.00 (40.00, 100.00)	80.00 (50.00, 120.00)	Z = -4.29	< 0.01

ASA: American Society of Anesthesiologists.

Logistic regression analysis

As shown in Figure 2, 18 variables were screened out in LASSO regression. They are age, sex, anesthesia time, ASA II and whether to pump respectively remifentanil, Cr, Sufentanil, AST, ALT, ALP, GGT, TBIL, ALB, PLT, Glu, length of hospital stay, operation on the esophagus, gastrointestinal tract pancreas, spleen, and appendix, operation on the uterus, ovaries, and fallopian tube were included in the binary Logistic regression analysis. The logistic regression analysis identified several variables with statistically significant associations (P < 0.05) with postoperative liver dysfunction. Anesthesia duration had a P value of 0, an odds ratio (OR) of 1.005, and a 95% confidence interval of (1.003, 1.007), indicating that longer anesthesia duration significantly increased the risk of postoperative liver dysfunction. The use of remifentanil revealed a P value of 0.014, an OR of 4.61, and a 95% confidence interval of (1.360, 15.624), suggesting a significant increase in risk associated with its use. Sevoflurane had a P value of 0, an OR of 1.008, and a 95% confidence interval of (1.004, 1.012), also indicating an increased risk. AST had a P value of 0, an OR of 1.043, and a 95% confidence interval of (1.025, 1.062), showing a positive relationship with the risk. ALT had a P value of 0.002, an OR of 1.019, and a 95% confidence interval of (1.007, 1.031), further increasing the risk. PLT had a P value of 0.001, an OR of 0.996, and a 95% confidence interval of (0.994, 0.999), indicating that a higher platelet count was associated with a reduced risk. Operations on the esophagus, gastrointestinal tract, pancreas, spleen, and appendix had a P value of 0, an OR of 2.279, and a 95% confidence interval of (1.456, 3.567), significantly increasing the risk, while surgeries of the uterus, ovary, and fallopian tube had a P value of 0.013, an OR of 0.461, and a 95% confidence interval of (0.251, 0.847), significantly reducing the risk (Table 4).

Figure 2 LASSO regression analysis. A: Lasso coefficient paths; B: Number of selected features.

Table 4 Results of the logistic regression analysis.

Parameter	β	Se	Wald χ2	P value	Or	Lower limit of 95%CI	95%CI upper bound
Age, year	-0.011	0.007	2.018	0.155	0.989	0.975	1.004
Sex, male	-0.043	0.227	0.035	0.851	0.958	0.614	1.495
Anesthesia time, minute	0.005	0.001	25.564	0	1.005	1.003	1.007
ASA II	-0.027	0.215	0.016	0.899	0.973	0.638	1.483
Whether to pump remifentanil	1.528	0.623	6.023	0.014	4.610	1.360	15.624
Cr, mmol/L	-0.002	0.003	0.508	0.476	0.998	0.992	1.004
Sufentanil	0.008	0.002	16.397	0.00	1.008	1.004	1.012
AST, U/L	0.042	0.009	21.633	0.00	1.043	1.025	1.062
ALT, U/L	0.019	0.006	9.411	0.002	1.019	1.007	1.031
ALP, g/L	-0.002	0.002	0.371	0.543	0.998	0.994	1.003
GGT, U/L	0.003	0.002	3.397	0.065	1.003	1.00	1.007
TBIL, μmol/L	0.015	0.011	1.907	0.167	1.015	0.994	1.037
Albumin, g/L	-0.013	0.021	0.385	0.535	0.987	0.946	1.029
PLT, × 1012/L	-0.004	0.001	10.624	0.001	0.996	0.994	0.999
Glu, mmol/L	0.011	0.041	0.078	0.78	1.012	0.934	1.096
Length of hospital stay, day	0.014	0.008	3.37	1	0.066	1.014	0.999
Operation on the esophagus, gastrointestinal tract, pancreas, spleen, and appendix	0.824	0.229	12.988	0.00	2.279	1.456	3.567
Operation on the uterus, ovaries, and fallopian tube	-0.773	0.310	6.231	0.013	0.461	0.251	0.847
Constant	-6.847	1.46	22.009	0.00	0.001

ASA: American Society of Anesthesiologists; ALP: Alkaline phosphatase; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; GGT: Gamma-glutamyl transpeptidase; PLT: Platelets; Glu: Glucose; TBIL: Total bilirubin; Cr: Creatinine.

Predictive model construction

Variables with P < 0.05 in the binary logistic regression analyses in Table 4 were included in the predictive models. The subject job characteristics (ROC) curves (Figure 3) indicate that the area under the curve (AUC) for the training set model (blue) was 0.90 (95%CI: 0.87–0.92), while for the validation set (orange), the AUC was 0.85 (95%CI: 0.78–0.92), both significantly higher than random guesses (diagonal, AUC = 0.5, P < 0.05). The training set curve is closer to the upper left corner, reflecting a high ability to capture true-positive samples; the validation set curve shows a slightly decreasing but stable shape, indicating that the model maintains good discriminative performance on independent data with strong generalization ability. The AUCs of both models were > 0.8, demonstrating that the models have excellent discriminative ability for outcome events and can effectively assist in clinical risk assessment and decision-making.

Figure 3  Receiver operating characteristic curves for training and validation sets.

Predictive model nomogram construction and model evaluation

The nomograms of postoperative liver function abnormality using R language are shown in Figure 4. The calibration curves of the training and validation sets of the prediction model are closely distributed around the 45-degree line, indicating good agreement between the model's predicted probabilities and the actual probabilities. The P values in the Hosmer–lemeshow test are 0.15 and 0.48, respectively, as shown in Figure 5, indicates that the model is well-fitted.

Figure 4  Predictive modeling nomogram.

Figure 5 Calibration curves. A: Calibration curve of the training set; B: Calibration curve of the verification set.

Decision curve analysis (DCA) (Figure 6) shows that the training set model (blue) yields positive net gains (above the green baseline of “None”) at high-risk thresholds of 0-0.8, suggesting clinical benefit from intervention within this range. Negative gains at thresholds above 0.8 imply that no intervention is preferable. The validation set model demonstrates a net benefit of ≥ 0 across the full threshold range (0-1.0) and positive gains at low thresholds (0-0.8), reflecting the model's robustness. The red curves (All) show negative gains, confirming the irrationality of the full intervention strategy. The DCA results for the training and validation sets are consistent, supporting the clinical value of the model in risk stratification, particularly at low-risk thresholds, where the net gain can be enhanced through targeted intervention to optimize clinical decision-making.

Figure 6 Decision curves. A: Decision curves of the training set; B: Decision curve of the validation set.

DISCUSSION

Post-operative abnormal liver function indicators are defined as deviations in liver enzyme activity and metabolic function from the normal physiological range. Such pathological changes are commonly accompanied by multiple complications, such as an imbalance of the coagulation-fibrinolysis system, nutritional metabolism disorders, and abnormal substance exchange. Complication severity is significantly correlated with the clinical outcomes of patients. Current evidence-based medical evidence has revealed an obvious disciplinary tendency in relevant research fields. Most of the existing literature focuses on the analysis of surgical procedures in cardiothoracic surgery and hepatobiliary specialties, while only a small part is in the field of general surgery. This difference in disciplinary distribution urgently needs to be improved by further research. This study identified an incidence of abnormal liver function after surgery of 4.4%. The more common independent risk factors in the total cohort and the three sub-cohorts included: Higher preoperative AST, ALT, decreased PLT, and the use of sevoflurane during the operation. Among the independent risk factors, higher preoperative AST and ALT levels increase the risk of abnormal liver function after surgery. As sensitive indicators of liver damage[11], high levels of ALT and AST may indicate preoperative liver damage or the existence of primary liver diseases. The most common diseases include drug-induced liver injury and viral hepatitis[12]. When the liver loses functionality due to surgery, liver function injury is extremely likely to be aggravated, inducing a series of pathological manifestations such as malnutrition and metabolic disorders, which in turn aggravate the primary liver disease[13]. Platelet count is of great significance for liver function, playing an important role in the processes of blood coagulation[14], liver regeneration[15,16], and liver ischemia/reperfusion[17], as well as in liver function regulation and liver damage response[18,19]. Studies have shown that mice with a PLT count < 100 × 10⁹/L show delayed recovery of liver function after surgery compared to those with a preoperative PLT count < 100 × 10⁹/L[20]. The liver damage caused by sevoflurane can be traced back to an early time. One clinical study showed that among approximately 111 patients (92%) who received sevoflurane anesthesia, the probability of clinically significant volatile anesthetic-induced liver injury after surgery was 4.1%, which is roughly equivalent to the rates for previously reported old drugs (3%)[21]. Furthermore, several individual case reports have been published, as follows: A 20-year-old male developed jaundice 9 days after undergoing general anesthesia surgery with sevoflurane. After the liver function test deteriorated, a liver biopsy showed centrilobular necrosis. The patient developed encephalopathy on the 30th day after surgery, ultimately requiring an emergency liver transplant[22]. A 66-year-old woman with breast cancer received two general anesthetics with sevoflurane within 25 days. She developed jaundice and a sharp increase in transaminase levels soon after the second surgery and died 66 days later. The autopsy results ruled out cardiovascular abnormalities, but histological examination confirmed extensive hepatocyte necrosis[23]. An in vitro hepatocyte assay demonstrated that while opioids such as remifentanil did not cause significant cytotoxicity (defined as a reduction in cell count of at least 50%), there was a significant reduction in three or more of the six tests[24]: Cell count, viability, lactate dehydrogenase, mitochondrial dehydrogenase activity, cytochrome P450 1A2, and albumin synthesis. However, some studies suggest that remifentanil may protect against hepatic ischemia/reperfusion injury by activating the Fmol/Parkin signaling pathway or up-regulating hypoxia-inducible factor 1-alpha expression via the ZEB1/LIF axis, a finding that requires further clinical validation[25,26].

This study has some limitations. Firstly, due to its retrospective design, while a substantial amount of clinical data can be obtained, our analyses relied on existing medical records, and the completeness and accuracy of the data may be influenced by the quality of those records. Additionally, the samples were primarily drawn from the Guangdong population, and the geographical and hospital constraints may limit the generalizability of the findings.

CONCLUSION

In conclusion, our findings suggest that abnormal preoperative elevations of AST and ALT, along with a decrease in PLT, as well as prolonged anesthesia and the metabolic effects of sevoflurane, are considerable risk factors for postoperative hepatic impairment. Additionally, the predictive model developed from the variables identified through logistic regression demonstrates good efficacy.

ACKNOWLEDGEMENTS

I would like to thank Dr. Zhao Gaofeng, Dr. Liao Min and the Department of Anesthesiology of Guangdong Provincial Hospital of Traditional Chinese Medicine for their help.