Abstract
Aims and background
Sepsis-associated liver dysfunction (SALD) represents a prevalent and critical complication frequently observed in patients with sepsis. The association between the initial aspartate aminotransferase (AST)-to-platelet (PLT) ratio index (APRI) and SALD is unclear in adult patients diagnosed with sepsis.
Patients and methods
We retrospectively analyzed data from the Medical Information Mart for Intensive Care-IV database. Sepsis-associated liver dysfunction was defined as an elevated serum aminotransaminase (>800 IU/L) or total bilirubin (>2 mg/dL) level. Multivariate and smoothing curve analyses were performed to investigate the relationship between the APRI [APRI = (AST (IU/L)/upper limits of normal)/PLT (k/uL)×100] and SALD. Subgroup analysis was additionally conducted to assess the robustness of the finding. Receiver operating characteristic (ROC) curve was performed to evaluate the discriminatory ability of SALD. External validation was performed using our own dataset.
Results
Overall, 6,334 sepsis patients (SALD, n = 985; no-SALD, n = 5,349) were included. Initial APRI was positively associated with SALD occurrence after controlling for potential confounding variables [odds ratio (OR) = 1.17; 95% confidence interval (CI): 1.15–1.20; p < 0.001]. A nonlinear dose-dependent relationship was found between initial APRI and SALD (p < 0.001). Subgroup analysis revealed no significant interaction between initial APRI and each subgroup divided by age, sex, albumin level, and Sequential Organ Failure Assessment score (p > 0.05). The area under the curve (AUC) for APRI was 0.769 (95% CI: 0.752–0.786), and the optimal cutoff was 0.95. External validation also exhibited good consistency (AUC: 0.761; 95% CI: 0.680–0.842).
Conclusion
A high initial APRI was linked to an elevated risk of developing SALD in adult patients with sepsis, as shown by the non-linear dose-dependent relationship.
Clinical significance
Initial APRI is an easy and accessible tool that can be adopted for timely detection of the risk of SALD and prompt initiation of interventions for adult patients with sepsis.
How to cite this article
Zhang B, Li X, Qin Z, Dong D, Yu W. Initial Aspartate Aminotransferase-to-platelet Ratio Index is Associated with Sepsis-associated Liver Dysfunction in Adult Patients with Sepsis: A Retrospective Cohort Study. Indian J Crit Care Med 2025;29(11):916–924.
Keywords: Aspartate aminotransferase-to-platelet ratio index, Medical Information Mart for Intensive Care IV database, Sepsis-associated liver dysfunction
Introduction
Sepsis, marked by severe organ dysfunction caused by an uncontrolled immune response to infection, frequently presents a life-threatening scenario in clinical environments.1 Sepsis has the potential to advance into multiple organ dysfunction, leading to high mortality rates.2 The liver, which serves as a crucial immune and metabolic organ, is frequently affected during this progression, resulting in significant adverse outcomes.3 Compared to typical sepsis scenarios, sepsis-associated liver dysfunction (SALD) has a poor prognosis, including no improvement in disease progression or survival outcomes, with incidence rates as high as 68.6%.4
The aspartate aminotransferase (AST)-to-platelet (PLT) ratio index (APRI), derived from conventional laboratory parameters, has been a longstanding valuable tool for evaluating liver damage and fibrosis in individuals suffering from hepatitis and nonalcoholic fatty liver disease.5,6 Aspartate aminotransferase is a sensitive biomarker of early liver injury, and PLT plays a crucial role in sepsis-induced multiple organ dysfunction syndrome (MODS) by modulating inflammation, preserving tissue integrity, and providing defense against infection.7 The APRI has been found to serve as a significant predictor of SALD in individuals diagnosed with sepsis.8,9 Additionally, the APRI has been utilized to forecast the prognosis of septic patients without a history of chronic liver disease.10
Nevertheless, the available evidence on the relationship between APRI and SALD is inadequate. The primary objective of this study was to examine whether the association between APRI and SALD follows a non-linear pattern. Additionally, SALD can be categorized into two distinct types: Hypoxic hepatitis and sepsis-induced cholestasis (SIC), which arise from varying underlying pathophysiological processes.2 The secondary objective was to assess the significance of APRI in these two subcategories. We conducted a retrospective analysis using data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) version 2.2 database and our hospital to investigate the potential correlation between the initial APRI and SALD, including hypoxic hepatitis and SIC.
Patients and Methods
Data Sources
The MIMIC-IV version 2.2 contains unidentified adult health-related data for over 73,000 adult admissions to the intensive care unit (ICU) at the Beth Israel Deaconess Medical Center, spanning the years from 2008 to 2019.11,12 The Computational Physiology Laboratory of the Massachusetts Institute of Technology (Cambridge, Massachusetts, USA) is responsible for maintaining this public database. We obtained access to the database by successfully completing an online training course provided by the National Institutes of Health and passing online examinations (certification number: 10246997). Since all data were anonymized, the requirement for informed consent was waived.
Data for the external validation of our research findings were obtained from the Nanjing Drum Tower Hospital (Nanjing, Jiangsu, China) after receiving approval from the institution's Medical Ethics Committee.
Patients
We enrolled adult patients with sepsis (≥18 years); however, individuals with multiple entries in the database were only included if information regarding their initial hospitalization and first admission to ICU was available. Exclusion criteria were as follows: (1) all types of chronic liver disease; (2) viral hepatitis; (3) primary acute cholangiopathies; (4) cholecystitis; (5) hepatic infarction; (6) liver necrosis; (7) liver trauma; (8) acute myocardial injury; (9) hemolysis; (10) malignant hemopathy; and (11) length of stay (LOS) in the ICU <24 hours. Patients who did not have concurrent data for AST and PLT within the first 24 hours following their admission to ICU were excluded from the study. The participant flowchart is depicted in Figure 1.
Fig. 1.
Patient flowchart
ICU, intensive care unit; MIMIC-IV, Medical Information Mart for Intensive Care IV; SALD, sepsis-associated liver dysfunction
Data Extraction
Navicat Premium version 15.0.12 (PremiumSoft CyberTech Ltd., Hong Kong, China), was used to establish a connection with the MIMIC-IV database, facilitating the extraction of required data through the utilization of structured query language (SQL).13 All codes used for extracting data from the database are accessible on GitHub.
The data of subsequent indicators were either obtained directly or extracted from the database: Age, weight, sex, comorbidities, infection sites, disease severity scores on the initial day of admission to ICU, mean value of vital signs recorded within the first 24 hours of admission, therapies administered within the first day of admission, initial laboratory values, and clinical outcomes.
Disease severity scores were calculated using the SQL code, proposed by Johnson et al.14 The APRI was calculated using the formula: [AST (IU/L)/upper limits of normal]/PLT (k/uL) × 100. Initial APRI refers to the APRI value obtained for the first time within the first 24 hours following admission to the ICU. According to standard guidelines, sepsis was identified in patients with confirmed or suspected infectious sources and a simultaneous Sequential Organ Failure Assessment (SOFA) score ≥2.1 Sepsis-associated liver dysfunction was diagnosed according to the following criteria: (1) alanine aminotransferase (ALT) or AST levels >800 IU/L (20 times the upper limit of normal), or (2) serum total bilirubin (TBIL) levels >2.0 mg/dL3; it was categorized as hypoxic hepatitis and SIC. Patients were diagnosed with hypoxic hepatitis when they met criteria (1) alone, and were diagnosed with SIC when they met criteria (2). After the exclusion of ineligible participants, the dataset was subjected to an extensive evaluation, leading to the removal of outlier or incorrect values that failed to meet logical validation criteria. Furthermore, any variables exhibiting missing data that constituted 30% or more of the total sample size were discarded (Supplementary Table 1). The mean imputation method was applied to the remaining variables with missing values.
Table 1.
General characteristics of septic patients stratified by quartiles of initial APRI level
| Variables | Total (n = 6,334) |
Quartile 1
(<0.278) (n = 1,583) |
Quartile 2
(0.278–0.550) (n = 1,584) |
Quartile 3
(0.551–1.395) (n = 1,583) |
Quartile 4
(> 1.395) (n = 1,584) |
p |
|---|---|---|---|---|---|---|
| Demography | ||||||
| Age (years), mean ± SD | 66.55 ± 17.34 | 67.16 ± 16.54 | 68.26 ± 16.96 | 66.98 ± 17.26 | 63.81 ± 18.23 | <0.001 |
| Weight (kg), mean ± SD | 81.98 ± 23.26 | 81.09 ± 23.66 | 82.11 ± 24.05 | 82.21 ± 23.04 | 82.52 ± 22.26 | 0.340 |
| Male, n (%) | 3,555 (56.13) | 776 (49.02) | 921 (58.14) | 915 (57.80) | 943 (59.53) | <0.001 |
| Comorbidity | ||||||
| Hypertension, n (%) | 2,481 (39.17) | 651 (41.12) | 642 (40.53) | 616 (38.91) | 572 (36.11) | 0.018 |
| Old myocardial infarct, n (%) | 1,363 (21.52) | 243 (15.35) | 301 (19.00) | 359 (22.68) | 460 (29.04) | <0.001 |
| Chronic heart failure, n (%) | 2,182 (34.45) | 490 (30.95) | 538 (33.96) | 591 (37.33) | 563 (35.54) | 0.002 |
| Cerebrovascular disease, n (%) | 1,028 (16.23) | 300 (18.95) | 275 (17.36) | 241 (15.22) | 212 (13.38) | <0.001 |
| Chronic pulmonary disease, n (%) | 1,708 (26.97) | 482 (30.45) | 438 (27.65) | 411 (25.96) | 377 (23.80) | <0.001 |
| Chronic renal disease, n (%) | 1,504 (23.74) | 410 (25.90) | 385 (24.31) | 379 (23.94) | 330 (20.83) | 0.008 |
| Diabetes, n (%) | 1,897 (29.95) | 543 (34.30) | 485 (30.62) | 472 (29.82) | 397 (25.06) | <0.001 |
| Infection site | ||||||
| Blood, n (%) | 4,120 (65.05) | 1,019 (64.37) | 1,001 (63.19) | 991 (62.60) | 1,109 (70.01) | <0.001 |
| Lung, n (%) | 1,308 (20.65) | 284 (17.94) | 349 (22.03) | 334 (21.10) | 341 (21.53) | 0.020 |
| Skin and soft tissue, n (%) | 81 (1.28) | 24 (1.52) | 12 (0.76) | 27 (1.71) | 18 (1.14) | 0.084 |
| Abdominal cavity, n (%) | 474 (7.48) | 115 (7.26) | 106 (6.69) | 123 (7.77) | 130 (8.21) | 0.403 |
| Urinary tract, n (%) | 2,020 (31.89) | 486 (30.70) | 514 (32.45) | 486 (30.70) | 534 (33.71) | 0.194 |
| Scoring systems | ||||||
| APS III, mean ± SD | 49 (37, 65) | 47 (36, 59) | 46 (36, 60) | 50 (37, 65) | 56 (42, 74) | <0.001 |
| SOFA, mean ± SD | 6 (4, 9) | 4 (3, 7) | 5 (3, 7) | 6 (4, 9) | 8 (5, 11) | <0.001 |
| LODS, mean ± SD | 5 (3, 8) | 5 (3, 7) | 5 (3, 7) | 5 (4, 8) | 7 (5, 9) | <0.001 |
| OASIS, mean ± SD | 34 (29, 41) | 34 (28, 40) | 34 (29, 40) | 34 (29, 40) | 36 (30, 43) | <0.001 |
| SAPS II, mean ± SD | 40 (31, 50) | 38 (30, 47) | 39 (31, 47) | 40 (31, 50) | 43 (33, 55) | <0.001 |
| Vital signs in first 24 hours | ||||||
| Mean HR (bpm), mean ± SD | 88.03 ± 17.30 | 87.94 ± 16.40 | 86.49 ± 17.18 | 87.59 ± 17.25 | 90.10 ± 18.13 | <0.001 |
| Mean SBP (mm Hg), mean ± SD | 115.84 ± 16.03 | 118.61 ± 17.19 | 117.07 ± 15.89 | 115.28 ± 15.66 | 112.41 ± 14.64 | <0.001 |
| Mean DBP (mm Hg), mean ± SD | 62.50 ± 10.76 | 62.78 ± 10.59 | 62.10 ± 10.58 | 62.01 ± 10.94 | 63.12 ± 10.90 | 0.008 |
| Mean MAP (mm Hg), mean ± SD | 77.40 ± 10.60 | 77.90 ± 10.81 | 77.31 ± 10.58 | 76.95 ± 10.53 | 77.46 ± 10.45 | 0.084 |
| Mean RR (bpm), mean ± SD | 20.49 ± 4.17 | 20.09 ± 3.94 | 20.17 ± 4.11 | 20.41 ± 4.10 | 21.27 ± 4.41 | <0.001 |
| Mean T (°C), mean ± SD | 36.93 ± 0.70 | 36.95 ± 0.52 | 36.98 ± 0.59 | 36.94 ± 0.73 | 36.85 ± 0.90 | <0.001 |
| Mean SpO2 (%), mean ± SD | 96.87 ± 2.40 | 96.90 ± 2.03 | 96.89 ± 2.13 | 96.95 ± 2.29 | 96.73 ± 3.04 | 0.061 |
| Treatment received in first 24 hours | ||||||
| Invasive ventilation, n (%) | 3,349 (52.87) | 715 (45.17) | 774 (48.86) | 886 (55.97) | 974 (61.49) | <0.001 |
| Vasopressor therapy, n (%) | 3,022 (47.71) | 602 (38.03) | 670 (42.30) | 789 (49.84) | 961 (60.67) | <0.001 |
| Renal replacement therapy, n (%) | 366 (5.78) | 74 (4.67) | 70 (4.42) | 81 (5.12) | 141 (8.90) | <0.001 |
| Initial laboratory parameters | ||||||
| WBC count (k/uL), mean ± SD | 13.87 ± 9.03 | 14.54 ± 8.82 | 13.23 ± 8.02 | 13.38 ± 8.79 | 14.33 ± 10.26 | <0.001 |
| Hemoglobin (gm/dL), mean ± SD | 10.87 ± 2.35 | 10.57 ± 2.10 | 11.01 ± 2.28 | 10.94 ± 2.39 | 10.97 ± 2.59 | <0.001 |
| Hematocrit (%), mean ± SD | 33.19 ± 7.04 | 32.55 ± 6.23 | 33.61 ± 6.86 | 33.24 ± 7.09 | 33.36 ± 7.83 | <0.001 |
| MCH (pg), mean ± SD | 29.94 ± 2.60 | 29.33 ± 2.67 | 29.88 ± 2.44 | 30.24 ± 2.60 | 30.32 ± 2.56 | <0.001 |
| MCHC (gm/dL), mean ± SD | 32.72 ± 1.71 | 32.39 ± 1.70 | 32.71 ± 1.66 | 32.87 ± 1.68 | 32.90 ± 1.75 | <0.001 |
| MCV (fL), mean ± SD | 91.62± 7.13 | 90.63 ± 7.22 | 91.45 ± 6.70 | 92.08 ± 7.18 | 92.31 ± 7.30 | <0.001 |
| RBC (m/uL), mean ± SD | 3.65 ± 0.80 | 3.62 ± 0.74 | 3.70 ± 0.78 | 3.64 ± 0.81 | 3.64 ± 0.87 | 0.028 |
| RDW (%), mean ± SD | 15.02 ± 2.15 | 15.12 ± 2.08 | 14.91 ± 2.07 | 15.00 ± 2.24 | 15.05 ± 2.21 | 0.046 |
| Platelet (k/uL), median (IQR) | 196 (139, 266) | 280 (220, 365) | 198 (156, 252) | 167 (123, 223) | 143 (87, 202) | <0.001 |
| ALT (IU/L), median (IQR) | 26 (16, 53) | 15 (11, 23) | 21 (15, 33) | 31 (19, 50) | 78 (39, 189) | <0.001 |
| AST (IU/L), median (IQR) | 36 (23, 72) | 20 (15, 26) | 29 (23, 39) | 47 (33, 70) | 118 (60, 262) | <0.001 |
| TBIL (mg/dL), median (IQR) | 0.60 (0.40, 1.00) | 0.40 (0.30, 0.70) | 0.60 (0.40, 0.90) | 0.60 (0.40, 1.00) | 0.80 (0.50, 1.40) | <0.001 |
| ALP (IU/L), median (IQR) | 74 (55, 104) | 75 (58, 101) | 72 (55, 98) | 73 (52, 103) | 77 (55, 116) | <0.001 |
| Albumin (gm/dL), mean ± SD | 3.02 ± 0.55 | 3.04 ± 0.57 | 3.09 ± 0.55 | 3.01 ± 0.52 | 2.94 ± 0.55 | <0.001 |
| Bun (mg/dL), median (IQR) | 23 (15, 38) | 21 (14, 37) | 22 (14, 36) | 23 (15, 38) | 25 (16, 41) | <0.001 |
| Creatinine (mg/dL), median (IQR) | 1.10 (0.80, 1.80) | 1.00 (0.70, 1.70) | 1.10 (0.80, 1.60) | 1.10 (0.80, 1.70) | 1.30 (0.90, 2.00) | <0.001 |
| Bicarbonate (mEq/L), mean ± SD | 22.01 ± 5.14 | 22.98 ± 5.42 | 22.80 ± 4.91 | 22.07 ± 4.76 | 20.19 ± 4.98 | <0.001 |
| Anionic gap (mEq/L), mean ± SD | 15.69 ± 4.63 | 15.23 ± 4.25 | 14.93 ± 4.00 | 15.46 ± 4.56 | 17.16 ± 5.28 | <0.001 |
| Sodium (mEq/L), mean ± SD | 138.43 ± 6.04 | 138.46 ± 6.35 | 138.68 ± 6.15 | 138.30 ± 5.86 | 138.29 ± 5.79 | 0.223 |
| Potassium (mEq/L), mean ± SD | 4.23 ± 0.85 | 4.17 ± 0.76 | 4.17 ± 0.75 | 4.24 ± 0.86 | 4.34 ± 0.98 | <0.001 |
| Chloride (mEq/L), mean ± SD | 104.13 ± 7.15 | 103.49 ± 7.31 | 104.14 ± 7.08 | 104.35 ± 7.04 | 104.56 ± 7.11 | <0.001 |
| Total calcium (mg/dL), mean ± SD | 8.16 ± 0.93 | 8.32 ± 0.88 | 8.23 ± 0.87 | 8.13 ± 0.89 | 7.97 ± 1.03 | <0.001 |
| Glucose (mg/dL), median (IQR) | 136 (113, 172) | 134 (11, 171) | 133 (111, 164) | 135 (113, 168) | 141 (118, 185) | <0.001 |
| PT(s), median (IQR) | 14.20 (12.60, 16.20) | 13.90 (12.40, 16.00) | 13.90 (12.47, 16.00) | 14.30 (12.70, 16.40) | 14.90 (12.90, 17.90) | <0.001 |
| PTT (s), median (IQR) | 30.80 (27.30, 37.40) | 30.50 (26.90, 34.35) | 30.70 (26.80, 35.90) | 30.80 (27.40, 37.20) | 32.20 (28.20, 42.90) | <0.001 |
| INR, median (IQR) | 1.30 (1.10, 1.50) | 1.30 (1.10, 1.40) | 1.30 (1.10, 1.40) | 1.30 (1.10, 1.50) | 1.30 (1.20, 1.60) | <0.001 |
| SALD, n (%) | 985 (15.55) | 66 (4.17) | 140 (8.84) | 216 (13.64) | 563 (35.54) | <0.001 |
| Hypoxic hepatitis, n (%) | 209 (3.30) | 10 (0.63) | 11 (0.69) | 19 (1.20) | 169 (10.67) | <0.001 |
| SIC, n (%) | 776 (12.25) | 56 (3.55) | 129 (8.15) | 197 (12.44) | 394 (24.87) | <0.001 |
| Length of stay | ||||||
| Hospital (days), median (IQR) | 9.74 (5.72, 16.82) | 9.70 (5.66, 16.08) | 9.48 (5.75, 15.95) | 9.65 (5.83, 16.90) | 10.10 (5.41, 18.31) | 0.402 |
| ICU (days), median (IQR) | 3.92 (2.15, 7.76) | 3.62 (2.01, 6.91) | 3.73 (2.11, 7.26) | 4.16 (2.20, 8.26) | 4.25 (2.46, 8.54) | <0.001 |
| Mortality | ||||||
| ICU, n (%) | 1,006 (15.88) | 181 (11.43) | 189 (11.93) | 250 (15.79) | 386 (24.37) | <0.001 |
| Hospital, n (%) | 1,300 (20.52) | 246 (15.54) | 278 (17.55) | 320 (20.21) | 456 (28.79) | <0.001 |
| 30-day, n (%) | 1,545 (24.39) | 337 (21.29) | 352 (22.22) | 357 (22.55) | 499 (31.50) | <0.001 |
ALP, alkaline phosphatase; ALT, alanine aminotransferase; APRI, aspartate aminotransferase-to-platelet ratio index; APS III, Acute Physiology Score III; AST, aspartate aminotransferase; Bun, blood urea nitrogen; DBP, diastolic blood pressure; HR, heart rate; ICU, intensive care unit; INR, international normalized ratio; LODS, Logistic Organ Dysfunction System Score; MAP, mean arterial pressure; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MCV, mean corpuscular volume; OASIS, Oxford Acute Severity of Illness Score; PT, prothrombin time; PTT, partial thromboplastin time; RBC, red blood cell; RDW, red blood cell distribution width; RR, breathing rate; SALD, sepsis-associated liver dysfunction; SAPS II, Simplified Acute Physiology Score II; SBP, systolic blood pressure; SIC, sepsis-induced cholestasis; SOFA, sequential organ failure assessment; SpO2, pulse oxygen saturation; T, temperature; TBIL, total bilirubin; WBC, white blood cell
Data for external validation was retrospectively collected between January 2023 and March 2024, based on the previously mentioned inclusion and exclusion criteria.
Statistical Analysis
Continuous variables that exhibited a normal or nearly normal distribution were represented as mean ± standard deviation (SD), whereas those characterized by a skewed distribution were denoted as median (interquartile range). Group comparisons were conducted using the one-way analysis of variance and Kruskal–Wallis tests, as deemed suitable. Categorical variables were represented as absolute values (in percentages) and analyzed through the Chi-square test. Both univariate and multivariate logistic regression analyses were conducted to assess the relationship between the initial APRI and SALD. Smooth curve fitting was performed to explore the non-linear relationship between APRI and SALD. A two-piecewise linear regression model was employed based on the smoothed curve to determine the threshold effect. A likelihood-ratio test was conducted to compare the two linear regression models. The 30-day mortality rates were compared using the log-rank tests, and the results were presented using the Kaplan–Meier curves. To evaluate the accuracy of the initial APRI in discriminating SALD, receiver operating characteristic (ROC) curves were generated, and the area under the curve (AUC) was calculated. All analyses were performed using the statistical software package R and Free Statistics software version 1.8. p < 0.05 was considered statistically significant.
Results
General Characteristics of the Study Population
A total of 6,334 patients were included in the final analysis; 985 patients with SALD and 5,349 without SALD. The incidence of SALD was 15.55%. Among patients with SALD, 209 and 776 were classified as having hypoxic hepatitis and SIC, respectively. The general characteristics of the study population based on the quartiles of the initial APRI are presented in Table 1.
Patients in the fourth quartile of the APRI had the highest scores on the Acute Physiology Score III (APS III), SOFA, Logistic Organ Dysfunction Score, Oxford Acute Severity of Illness Score (OASIS), and Simplified Acute Physiology Score II (SAPS II); highest proportions of invasive ventilation, vasopressor therapy, and renal replacement therapy; highest initial levels of mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration (MCHC), mean corpuscular volume, ALT, AST, TBIL, alkaline phosphatase (ALP), blood urea nitrogen (Bun), creatinine, anionic gap, potassium, chloride, and glucose; highest prothrombin time (PT), and partial thromboplastin time (PTT); highest rates of SALD, hypoxic hepatitis, and SIC; highest hospital mortality; longest ICU LOS; highest ICU mortality; and highest 30-day mortality (all p < 0.05). Furthermore, the Kaplan–Meier survival curves showed that the 30-day mortality in the fourth quartile of APRI was significantly higher than the mortality rates in the remaining quartiles (p < 0.05) (Supplementary Figure 1). Moreover, the median and quartiles of the initial APRI in patients without SALD, with hypoxic hepatitis, with SIC, and with SALD were 0.48 (0.26, 1.05), 9.61 (2.33, 25.89), 1.45 (0.57, 5.12), and 2.00 (0.66, 9.14), respectively. The initial APRI values of patients with hypoxic hepatitis, SIC, and SALD were higher than those of patients without SALD (all p < 0.05) (Supplementary Figure 2).
Fig. 2.
Association between initial APRI level and SALD in adult patients with sepsis
APRI, aspartate aminotransferase-to-platelet ratio index; SALD, sepsis-associated liver dysfunction; the red solid curve represents the OR, and the coral area indicates its 95% CI; the blue histograms show the percentage of non-SALD participants belonging to each level of initial APRI value; the red histograms show the percentage of SALD participants belonging to each level of initial APRI value
Association between Initial APRI and SALD
The findings of the univariate analysis are displayed in Supplementary Table 2. Age, sex, weight, chronic heart failure, cerebrovascular disease, chronic renal disease, diabetes, infection site, scoring system, mean heart rate, mean systolic blood pressure, mean arterial pressure, mean respiratory rate, mean pulse oxygen saturation, vasopressor therapy, invasive ventilation, renal replacement therapy, white blood cell count, levels of mean corpuscular hemoglobin, MCHC, red blood cell distribution width (RDW), PLT, ALP, albumin, Bun, creatinine, bicarbonate, anionic gap, sodium, potassium, chloride, total calcium, and glucose, PT, PTT, international normalized ratio, and APRI were significantly associated with SALD (all p < 0.05).
Table 2.
Multivariate analyses for SALD
| Variables | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p | |
| APRI | 1.25 (1.22–1.28) | <0.001 | 1.19 (1.16–1.21) | <0.001 | 1.17 (1.15–1.20) | <0.001 |
| APRI | ||||||
| Quartile 1 | 1 (Ref) | 1 (Ref) | 1 (Ref) | |||
| Quartile 2 | 2.24 (1.65–3.03) | <0.001 | 2.06 (1.51–2.80) | <0.001 | 2.21 (1.59–3.07) | <0.001 |
| Quartile 3 | 3.63 (2.73–4.83) | <0.001 | 2.60 (1.94–3.49) | <0.001 | 2.57 (1.85–3.57) | <0.001 |
| Quartile 4 | 12.48 (9.55–16.32) | <0.001 | 6.67 (5.05–8.83) | <0.001 | 6.01 (4.30–8.39) | <0.001 |
| Trend test | <0.001 | <0.001 | <0.001 | |||
Model 1: adjusted for age, gender, weight; model 2: adjusted for model 1 and APS III score, SOFA score, LODS score, OASIS score, SAPS II score; model 3: adjusted for model 2 and chronic heart failure, cerebrovascular disease, chronic renal disease, diabetes, infection site, mean HR, mean SBP, mean MAP, mean RR, mean SpO2, vasopressor therapy, invasive ventilation, renal replacement therapy, WBC count, MCH, MCHC, RDW, platelet, ALP, albumin, bun, creatinine, bicarbonate, anionic gap, sodium, potassium, chloride, total calcium, glucose, PT, PTT, and INR. APRI, aspartate aminotransferase-to-platelet ratio index; CI, confidence interval; OR, odds ratio; SALD, sepsis-associated liver dysfunction
The independent effect of the initial APRI on the occurrence of SALD was determined by running three models (Table 2). When the APRI was set as a continuous variable, higher initial APRI was associated with a higher SALD occurrence in all three models [model 1: odds ratio (OR) = 1.25, 95% confidence interval (CI): 1.22–1.28, p < 0.001; model 2: OR = 1.19, 95% CI: 1.16–1.21, p < 0.001; and model 3: OR = 1.17, 95% CI: 1.15–1.20, p < 0.001]. In the sensitivity analysis, the APRI was converted from a continuous variable to categorical variable quartiles. Compared with the first quartile of the initial APRI, the multivariate ORs for SALD occurrence were 2.21 (1.59–3.07) for the second quartile, 2.57 (1.85–3.57) for the third quartile, and 3.75 (2.58–5.44) for the fourth quartile. The p-value for the trend of the initial APRI with categorical variables in the fully adjusted model was consistent with the results reported when the APRI was a continuous variable (p < 0.05). Other factors, including age, weight, cerebrovascular disease, pulmonary infection, APS III, SOFA, OASIS, SAPS II, MCHC, and RDW, were independently related to SALD (p < 0.05).
Multivariate analysis also revealed that the initial APRI was significantly associated with SIC and hypoxic hepatitis occurrence in all three models (all p < 0.05) (Supplementary Tables 3 to 4).
Curve Fitting and Subgroup Analysis
Smoothing curve analysis revealed a non-linear association between the initial APRI and SALD (p for non-linearity <0.001; Fig. 2) following model 3 adjustment. A two-piecewise linear regression model revealed that the critical threshold for the initial APRI was 0.516 (Supplementary Table 5). Before the inflection point, the OR was 35.25 (95% CI: 7.49–165.86, p < 0.001), whereas after the inflection point, the OR was 1.29 (95% CI: 1.03–1.61, p = 0.03), suggesting that for every 0.1 unit increase in APRI, the risk of SALD increased by 342.5% until the APRI reaches 0.516. Thereafter, the risk of SALD started to increase by 2.9% per 0.1 unit change in APRI after 0.516.
Smooth curve analysis also revealed a non-linear association between the initial APRI and SIC occurrence (p for non-linearity <0.001; Supplementary Figure 3). The critical threshold for the initial APRI was 0.566. Before the inflection point, the OR was 26.50 (95% CI: 6.49–108.13, p < 0.001). After the inflection point, the OR was 1.19 (95% CI: 1.01–1.41, p = 0.04, Supplementary Table 6). However, no non-linear relationship was found between initial APRI and occurrence of hypoxic hepatitis (p for non-linearity = 0.148, Supplementary Figure 4).
Figs 3A and B.
ROC curve of initial APRI for SALD. (A) Derived from MIMIC-IV v2.2 data; (B) External validation using our own data
APRI, aspartate aminotransferase-to-platelet ratio index; APS III, Acute Physiology Score III; AUC, the area under the curve; MCHC, mean corpuscular hemoglobin concentration; MIMIC-IV, v2.2, Medical Information Mart for Intensive Care IV version 2.2; OASIS, Oxford Acute Severity of Illness Score; RDW, red blood cell distribution width; ROC, receiver operating characteristic; SALD, sepsis-associated liver dysfunction; SAPS II, Simplified Acute Physiology Score II; SOFA, Sequential Organ Failure Assessment; SOFA + APRI, SOFA combined with APRI
Subgroup analyses were performed to stratify the association between initial APRI and SALD according to age, sex, albumin level, and SOFA score, as shown in Supplementary Figure 5. A positive correlation between the initial APRI and risk of SALD occurrence in the adult sepsis population remained consistent across all subgroup variables, with no interaction (p > 0.05).
ROC Analysis of the Initial APRI for SALD and External Validation
As shown in Figure 3A and Table 3, ROC analysis showed that the AUC of initial APRI for SALD occurrence was 0.769 (95% CI: 0.752–0.786), indicating the promising discriminative ability of APRI. The AUCs of the APS III, SOFA, OASIS, and SAPS II scores for SALD discrimination were 0.690 (95% CI: 0.672–0.707), 0.752 (95% CI: 0.736–0.769), 0.601 (95% CI: 0.581–0.621), and 0.627 (95% CI: 0.607–0.647), respectively. The AUCs of the MCHC, RDW, and SOFA combined with the APRI were 0.526 (95% CI: 0.506–0.546), 0.554 (95% CI: 0.534–0.573), and 0.807 (95% CI: 0.791–0.823), respectively. Using the Youden index, the cutoff values of the APS III, SOFA, OASIS, SAPS II, MCHC, RDW, APRI, and SOFA combined with APRI were 50.50, 6.50, 39.50, 43.50, 34.45, 15.85, 0.95, and 0.15, respectively. Furthermore, new data (n = 134) collected from our hospital for external validation also showed similar results, with the AUC of the initial APRI for SALD being 0.761 (95% CI: 0.680–0.842) (Fig. 3B). The AUC of the initial APRI for SIC occurrence was 0.739 (95% CI: 0.720–0.758) (Supplementary Figure 6), and the AUC of that for hypoxic hepatitis occurrence was 0.879 (95% CI: 0.849–0.910) (Supplementary Figure 7).
Table 3.
ROC analysis of the initial APRI for SALD
| Sensitivity (%) | Specificity (%) | Threshold | AUC | 95% CI | |
|---|---|---|---|---|---|
| APS III | 72.08 | 56.70 | 50.50 | 0.690 | 0.672–0.707 |
| SOFA | 73.50 | 62.63 | 6.50 | 0.752 | 0.736–0.769 |
| OASIS | 42.74 | 73.88 | 39.50 | 0.601 | 0.581–0.621 |
| SAPS II | 57.06 | 63.62 | 43.50 | 0.627 | 0.607–0.647 |
| MCHC | 19.29 | 86.15 | 34.45 | 0.526 | 0.506–0.546 |
| RDW | 32.69 | 74.82 | 15.85 | 0.554 | 0.534–0.573 |
| APRI | 66.09 | 72.72 | 0.95 | 0.769 | 0.752–0.786 |
| SOFA + APRI | 68.63 | 78.03 | 0.15 | 0.807 | 0.791–0.823 |
APRI, aspartate aminotransferase-to-platelet ratio index; APS III, Acute Physiology Score III; AUC, area under the curve; CI, confidence interval; MCHC, mean corpuscular hemoglobin concentration; OASIS, Oxford Acute Severity of Illness Score; RDW, red blood cell distribution width; ROC, receiver operating characteristic; SALD, sepsis-associated liver dysfunction; SAPS II, Simplified Acute Physiology Score II; SOFA, Sequential Organ Failure Assessment; SOFA + APRI, SOFA combined with APRI
Discussion
Evaluating and stratifying the risks linked to SALD can enable early detection of SALD and timely adjustment of management strategies for sepsis patients, preventing progression to MODS. Accordingly, clinical or laboratory indices need to be examined to identify sensitive, accurate, and convenient predictors of SALD. To our knowledge, our study is the first to assess the relationship between the APRI and SALD in adult patients with sepsis. The results revealed that a high initial APRI was significantly associated with a high risk of developing SALD, showing a non-linear, positive relationship. In particular, as the APRI increased to 0.516, a substantial increase in the rate of SALD was observed.
Owing to differences in diagnostic criteria and populations enrolled in studies, varying incidence rates of SALD have been reported, ranging from 1.3 to 45.9%.15–17 In our study, the included population comprised adult patients with sepsis, and SALD diagnosis was based on transaminase and bilirubin levels. The incidence of SALD was 15.55%, which was consistent with the aforementioned results. Sepsis-associated liver dysfunction can be categorized into hypoxic hepatitis and SIC based on the presence of hyperbilirubinemia.3 Hypoxic hepatitis is mainly caused by a decrease in oxygen supply to the liver, including inadequate oxygen concentration in the blood, reduced blood flow, and anemia. These factors induce hepatocellular damage. Sepsis-induced cholestasis primarily results from impaired bile formation at the hepatocellular level or disrupted bile flow at the small or large bile duct level. In our study, up to 78.78% of patients with SALD were classified as SIC, and cholestasis was the most common gastrointestinal sign in patients with SALD. This result is consistent with the findings of another study, where the incidence of cholestasis (serum TBIL >3 mg/dL) in patients with sepsis was as high as 83.73%.18
Aspartate aminotransferase is predominantly found in hepatocytes. When liver cell damage occurs, intracellular AST is released into the serum. Plasma AST levels are widely used to assess liver tissue damage caused by a diverse range of factors, including hepatitis and fatty liver disease. Sepsis can lead to thrombocytopenia through mechanisms involving diminished PLT production, shortened PLT lifespan, and increased consumption. Importantly, PLT is involved in the process of immune-mediated thrombus development and generation of neutrophil extracellular traps, which facilitate the trapping and elimination of pathogens, including bacteria and viruses.7 Sepsis-associated thrombotic microangiopathy has gained significant attention in recent years, primarily due to its hallmark features, including thrombocytopenia, microangiopathic hemolytic anemia, and the consequent organ damage.19 Thrombocytopenia is strongly linked to disease severity and prognosis in individuals with sepsis. For patients with SALD, PLT is implicated in the early response of the liver to sepsis, playing a crucial role in hepatocyte injury by interacting with neutrophils, Kupffer cells, and endothelial cells.20 Inpatients with sepsis, PLTs are recruited to the sinusoidal and Disse spaces, resulting in hepatic injury.21 These factors ultimately reduce the level of PLT in the peripheral circulation. Dou et al.22 retrospectively analyzed 1,141 children with sepsis and found that the initial APRI level within 24 hours after pediatric ICU admission was an independent risk factor for the development of SALD and had excellent predictive ability for SALD, with a sensitivity of 84.6% and specificity of 84.3%. A retrospective cohort study conducted by Zhu et al.10 comprising 1,562 patients with sepsis and APRI data showed no significant association between basal APRI at ICU admission and mortality. Xie et al.8 constructed three different prediction models for SALD, based on procalcitonin, APRI, and lactate levels, with all models showing high predictive value. He et al.9 performed a retrospective analysis of clinical data of 1,573 patients with sepsis, and found that the AUC of APRI for predicting SALD was 0.797, with a sensitivity of 83.19 and specificity of 65.27. Our model yielded a similar result, with an AUC of 0.769, and demonstrated good consistency in predicting SIC and hypoxic hepatitis.
The relationship and optimal range of the initial APRI with SALD have not been clarified in previous studies. To elucidate this relationship, we utilized the MIMIC-IV database to systematically analyze the dose–response relationship and threshold effect between initial APRI and SALD. After controlling for potential confounding variables, a non-linear dose-dependent relationship was observed between initial APRI and occurrence of SALD during ICU stay in adult patients with sepsis. The critical threshold of the initial APRI was 0.516; therefore, an increase toward 0.516 may lead to a rapidly increasing risk of SALD. Similar results have been found in patients with SIC, but not patients with hypoxic hepatitis. Nevertheless, the small sample size of these patients (n = 209) may be the main cause of non-linearity (p > 0.05).
Our study has some limitations. First, the main source of data in our study was the MIMIC-IV database, which was derived exclusively from a single medical center and covers the specific time period. Consequently, selection bias is an inevitable concern that should be considered when interpreting our findings, despite the external validation performed in our study. Second, our study only focused on the clinical and laboratory information obtained within 24 hours of ICU admission, without considering the subsequent changes in these variables. Third, this was a retrospective observational study; further multicenter prospective studies are required to validate our findings.
Conclusion
In summary, an elevated initial APRI within 24 hours of ICU admission was significantly and positively correlated with SALD in adult sepsis patients. Thus, initial APRI might be a useful tool for detecting adult patients with sepsis who are at a high risk of developing SALD. Nevertheless, our findings must be validated in future prospective studies.
Author Contributions
Beiyuan Zhang: Study concept and design, methodology, investigation, writing and revising manuscript; Xiaoyao Li: Investigation, validation, writing original manuscript; Zimeng Qin: Collection and assembly of data, writing original manuscript; Danjiang Dong: Data analysis and interpretation; Wenkui Yu: Study concept and study design, article revision, and article approval.
Clinical Significance
Aspartate aminotransferase-to-platelet ratio index, functioning as a composite metric derived from standard laboratory parameters, has exhibited significant clinical utility in evaluating SALD in individuals suffering from sepsis. Its fundamental importance is evident in three key dimensions: (1) clinical convenience; (2) reliable risk assessment; and (3) general applicability.
Supplementary Materials
The Supplementary Figures 1 to 7 and Tables 1 to 6 are available on the journal website www.ijccm.org
Orcid
Beiyuan Zhang https://orcid.org/0000-0001-8760-6520
Xiaoyao Li https://orcid.org/0000-0002-2516-8710
Zimeng Qin https://orcid.org/0009-0000-6195-5677
Danjiang Dong https://orcid.org/0009-0007-7738-1576
Wenkui Yu https://orcid.org/0000-0003-1305-2005
Footnotes
Source of support: Nil
Conflict of interest: None
References
- 1.Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis and septic shock (sepsis-3) JAMA. 2016;315(8):801–810. doi: 10.1001/jama.2016.0287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Knox DB, Lanspa MJ, Kuttler KG, Brewer SC, Brown SM. Phenotypic clusters within sepsis-associated multiple organ dysfunction syndrome. Intensive Care Med. 2015;41(5):814–822. doi: 10.1007/s00134-015-3764-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Strnad P, Tacke F, Koch A, Trautwein C. Liver - guardian, modifier and target of sepsis. Nat Rev Gastroenterol Hepatol. 2017;14(1):55–66. doi: 10.1038/nrgastro.2016.168. [DOI] [PubMed] [Google Scholar]
- 4.Kobashi H, Toshimori J, Yamamoto K. Sepsis-associated liver injury: Incidence, classification and the clinical significance. Hepatol Res. 2013;43(3):255–266. doi: 10.1111/j.1872-034X.2012.01069.x. [DOI] [PubMed] [Google Scholar]
- 5.Dong B, Chen Y, Lyu G, Yang X. Aspartate aminotransferase to platelet ratio index and fibrosis-4 index for detecting liver fibrosis in patients with autoimmune hepatitis: A meta-analysis. Front Immunol. 2022;13:892454. doi: 10.3389/fimmu.2022.892454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Xiao G, Zhu S, Xiao X, Yan L, Yang J, Wu G. Comparison of laboratory tests, ultrasound, or magnetic resonance elastography to detect fibrosis in patients with nonalcoholic fatty liver disease: A meta-analysis. Hepatology. 2017;66(5):1486–1501. doi: 10.1002/hep.29302. [DOI] [PubMed] [Google Scholar]
- 7.Graham SM, Liles WC. Platelets in sepsis: Beyond hemostasis. Blood. 2016;127(24):2947–2949. doi: 10.1182/blood-2016-03-706168. [DOI] [PubMed] [Google Scholar]
- 8.Xie T, Xin Q, Cao X, Chen R, Ren H, Liu C, et al. Clinical characteristics and construction of a predictive model for patients with sepsis related liver injury. Clin Chim Acta. 2022;537:80–86. doi: 10.1016/j.cca.2022.10.004. [DOI] [PubMed] [Google Scholar]
- 9.He Y, Wang C, He W, Zhang H, Ding F, Liu Y, et al. Analysis of risk factors for sepsis-related liver injury and construction of a prediction model. Front Public Health. 2024;12:1475292. doi: 10.3389/fpubh.2024.1475292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Zhu X, Hu X, Qin X, Pan J, Zhou W. An elevated fibrosis-4 score is associated with poor clinical outcomes in patients with sepsis: An observational cohort study. Pol Arch Intern Med. 2020;130(12):1064–1073. doi: 10.20452/pamw.15699. [DOI] [PubMed] [Google Scholar]
- 11.Goldberger AL, Amaral LA, Glass L, Hausdorff JM, Ivanov PC, Mark RG, et al. PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation. 2000;101(23):E215–E220. doi: 10.1161/01.cir.101.23.e215. [DOI] [PubMed] [Google Scholar]
- 12.Johnson AE, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10(1):1. doi: 10.1038/s41597-022-01899-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Yang J, Li Y, Liu Q, Li L, Feng A, Wang T, et al. Brief introduction of medical database and data mining technology in big data era. J Evid Based Med. 2020;13(1):57–69. doi: 10.1111/jebm.12373. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Johnson AE, Stone DJ, Celi LA, Pollard TJ. The MIMIC code repository: Enabling reproducibility in critical care research. J Am Med Inform Assoc. 2018;25(1):32–39. doi: 10.1093/jamia/ocx084. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Palmowski L, Westhus B, Witowski A, Nowak H, Traut I, Canbay A, et al. Subphenotypes and the De Ritis ratio for mortality risk stratification in sepsis-associated acute liver injury: A retrospective cohort study. EClinical Medicine. 2025;82:103173. doi: 10.1016/j.eclinm.2025.103173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Yan J, Li S, Li S. The role of the liver in sepsis. Int Rev Immunol. 2014;33(6):498–510. doi: 10.3109/08830185.2014.889129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Song Y, Maged Abdulsalam Mohammed Ali AM, Yang W, Sun L. Clinical characteristics and prognosis of patients with early sepsis-related liver injury in Northeast China. J Intensive Care Med. 2025;40(3):253–262. doi: 10.1177/08850666241277512. [DOI] [PubMed] [Google Scholar]
- 18.Lin J, Gu C, Zhang S, Tian L, Ren K, Cao Z, et al. Sites and causes of infection in patients with sepsis-associated liver dysfunction: A population study from the medical information mart for intensive care III. Med Sci Monit. 2021;27:e928928. doi: 10.12659/MSM.928928. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Satoh K, Wada T, Tampo A, Takahashi G, Hoshino K, Matsumoto H, et al. Practical approach to thrombocytopenia in patients with sepsis: A narrative review. Thromb J. 2024;22(1):67. doi: 10.1186/s12959-024-00637-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sakurai K, Miyashita T, Okazaki M, Yamaguchi T, Ohbatake Y, Nakanuma S, et al. Role for neutrophil extracellular traps (NETs) and platelet aggregation in early sepsis-induced hepatic dysfunction. In Vivo. 2017;31(6):1051–1058. doi: 10.21873/invivo.11169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Meyer J, Lejmi E, Fontana P, Morel P, Gonelle-Gispert C, Bühler L. A focus on the role of platelets in liver regeneration: Do platelet-endothelial cell interactions initiate the regenerative process? J Hepatol. 2015;63(5):1263–1271. doi: 10.1016/j.jhep.2015.07.002. [DOI] [PubMed] [Google Scholar]
- 22.Dou J, Zhou Y, Cui Y, Chen M, Wang C, Zhang Y. AST-to-platelet ratio index as potential early-warning biomarker for sepsis-associated liver injury in children: A database study. Front Pediatr. 2019;7:331. doi: 10.3389/fped.2019.00331. [DOI] [PMC free article] [PubMed] [Google Scholar]