Computed tomography manifestations of drug-induced liver injury according to type and severity of injury

Abstract

Background

Drug-induced liver injury (DILI) has become a major cause of acute liver failure, and its incidence has been increasing steadily in recent years. This study aimed to compare the clinical and computed tomography (CT) imaging features of the variable biochemical damage and severity of DILI to establish a radiological model for predicting high-risk DILI based on CT image features.

Methods

The eligible patients with DILI (January 2016 to March 2024) who underwent serum laboratory examination and contrast abdominal CT within 3 months of onset were retrospectively analyzed at Affiliated Xihu Hospital of Hangzhou Medical College (Institution I) and The Second Affiliated Hospital of Zhejiang Chinese Medical University (Institution II). The severity-associated CT features were determined via binomial logistic regression analysis, and the efficacy of the different models were compared. The odds ratios (ORs) and corresponding 95% confidence intervals (CIs) provided were not adjusted.

Results

The injury types included hepatocellular (n=68, 45.64%), mixed (n=28, 18.79%), and cholestatic (n=53, 35.57%). The proportion of splenomegaly in patients with cholestatic injury (56.60%) was significantly higher than that in those with hepatocellular (35.71%) and mixed injury (22.06%) (P<0.001). Regarding severity, 127 (85.23%) patients had mild-to-moderate injury, and 22 (14.77%) had severe-to-fatal injury or required liver transplantation (LT). Injury severity was independently associated with quantitative liver-spleen contrast (Q-LSC) (OR =0.002; 95% CI: 0.00–0.13), and ascites (OR =70.83; 95% CI: 16.34–306.99). The prediction of the new model employing Q-LSC and ascites for high-risk DILI demonstrated excellent performance [area under the receiver operating characteristic (ROC) curve (AUC) =0.929; sensitivity=0.818; specificity =0.953].

Conclusions

Statistical differences are observed in the serum biomarkers of DILI according to varying biochemical damage and degree of severity. Q-LSC and ascites were associated with the severity of DILI, and a combined model incorporating Q-LSC and ascites can effectively predict high-risk DILI.

Introduction

Drug-induced liver injury (DILI) refers to the hepatic damage caused by prescription or nonprescription drugs, including small chemical molecules, biological agents, traditional Chinese medicines (TCMs), natural medicines, health products, and dietary supplements (1). DILI is a prevalent etiology of liver injury manifestations in China and remains a significant contributor to acute liver failure in Western nations, posing a 10% mortality risk in the United States (2-4). Epidemiological studies from Western countries have reported an estimated annual incidence of DILI ranging from 2.3 to 40.6 per 100,000 person-years and a 6-month mortality rate among patients with DILI approximates 8% (5). Therefore, the early stratification for the identification and management of patients with high-risk and severe DILI is crucial to improving the prognosis of patients following artificial liver transplantation (LT).

Prediction models incorporating various serum biomarkers or histologic patterns for predicting the outcome of DILI have been developed. Among these, the predictive models of Hy’s law and New Hy’s law (6) are widely used due to their high sensitivity and specificity in accurately predicting the occurrence of acute liver failure following the onset of DILI. Furthermore, several studies have indicated that histologic patterns exhibit a stronger correlation with clinical outcomes in DILI than do serum biochemical models (7,8). However, apart from invasive liver biopsies, there is currently no evidence linking the histologic features observed postbiopsy to the long-term clinical effects of DILI.

Radiologic evaluation plays a crucial role in diagnosing, monitoring, and managing patients with DILI. Recent advancements in ultrasound, computed tomography (CT), and magnetic resonance imaging (MRI) have significantly enhanced our understanding of DILI. Masuoka et al. (9) conducted a retrospective study on the CT imaging of 19 cases of immune checkpoint inhibitor-induced liver injury and identified biliary dilatation, biliary wall thickening, and non-edematous thickening of the gallbladder wall as the most common CT features among these patients. Wu et al. (10) investigated the MRI features associated with the severity of DILI and found that hepatic surface irregularities, transient differences in hepatic attenuation, and splenomegaly were significantly correlated with disease severity. Fu et al. (11) retrospectively analyzed 168 DILI cases using contrast-enhanced MRI, employed radiomics to extract meaningful features, and ultimately developed a combined clinical and imaging feature prediction model for chronic DILI. However, MRI imposes significant burdens on patients due to its high cost, time-consuming nature, and limited availability. In contrast, CT offers advantages such as affordability, widespread accessibility in primary care hospitals, detection of diffuse or focal liver diseases, and monitoring for manifestations of progression, including liver fibrosis and chronic liver failure (12-14). Additionally, quantitative CT techniques can provide valuable indicators for assessing hepatic steatosis, steatohepatitis, and hepatic fibrosis (15).

Therefore, to enhance our comprehension of the clinical and CT imaging characteristics of DILI, this study aimed to investigate the correlation of the type of biochemical damage and its severity in DILI with CT imaging features. Furthermore, we developed a radiological model based on characteristic CT risk factors and assessed the efficacy of this model in predicting high-risk DILI. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-24-2110/rc).

Methods

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Institutional Review Boards of the Affiliated Xihu Hospital of Hangzhou Medical College (No. 20241114/32/01/001) and The Second Affiliated Hospital of Zhejiang Chinese Medical University (No. 2025-LW-096-01). The requirement for individual consent was waived due to the retrospective nature of the analysis.

Study population

A retrospective study on hospitalized patients who had a confirmed diagnosis of DILI was conducted from January 2016 to March 2024. By searching the electronic medical record system for records of discharge diagnoses containing the keyword “drug-induced liver injury”, we collected comprehensive data including medication history, laboratory biochemical tests, imaging examination, and histopathologic findings. In order to establish a definitive diagnosis of DILI, we employed the Roussel-Uclaf causality assessment method (RUCAM) (16) for each patient, which is a validated tool used to assess the causal relationship between medications and liver injury events. The RUCAM scores were categorized as highly probable (>8 points), possible (6–8 points), probable (3–5 points), unlikely (1 or 2 points), or excluded (0 points).

The inclusion criteria (Figure 1) were as follows: (I) the RUCAM score was ≥6 points or 3–5 points. (II) Patients had a history of drug use and abnormal liver biochemistry, according to the DILI liver biochemistry diagnostic criteria delineated by the European Association for the Study of the Liver (EASL) (17) and met one or more of the following criteria: (i) aspartate aminotransferase (AST) or alanine aminotransferase (ALT) levels >5 time the upper limit of normal (ULN) or alkaline phosphatase (ALP) levels >2 times ULN; (ii) serum total bilirubin (TBIL) levels >2.5 mg/dL with elevated AST, ALT, or ALP; and (iii) international normalized ratio (INR) >1.5 with elevated AST, ALT, or ALP. (III) The time interval between the onset of DILI and the follow-up CT examination was within 3 months (10,18). The exclusion criteria were as follows: (I) viral hepatitis; (II) autoimmune hepatitis, primary biliary cholangitis, or primary sclerosing cholangitis; (III) nonalcoholic steatohepatitis and alcoholic liver disease; (IV) hepatocellular carcinoma; (V) a history of chronic liver disease and cirrhosis; and (VI) a lack of laboratory biochemical markers or imaging CT.

Figure 1.

Flowchart of enrollment of patients with DILI. Institution I, Affiliated Xihu Hospital of Hangzhou Medical College; Institution II, The Second Affiliated Hospital of Zhejiang Chinese Medical University. ΔT, the interval from DILI onset to CT examination; CT, computed tomography; DILI, drug-induced liver injury; LT, liver transplantation.

Definition of injury type and severity

The clinical, biochemical, serologic, and imaging data of DILI were extracted from the electronic medical record system. The R value was defined as the ratio of serum ALT level (as a multiple of its ULN) to ALP level (as a multiple of its ULN). Hepatocellular DILI was characterized by an R ≥5, cholestatic DILI by an R ≤2, and mixed DILI by an R >2 and <5. Patient severity was categorized as mild, moderate, severe, lethal, or requiring LT as per the consensus outlined in the International Guidelines (19).

CT examinations

All examinations were conducted with a 64- or 128-slice multidetector CT scanner (SOMATOM Definition AS, Siemens Healthineers, Erlangen, Germany). Subsequently, multiple-phase contrast-enhanced CT scans were performed to acquire precontrast, arterial, portal venous, and equilibrium phase images at 0, 25, 70, and 180 seconds after the administration of iobitridol injection (iodine concentration: 350 mg/mL). Iobitridol was administered at a 1.5 mL/kg dose with an injection rate of 3 mL/s. The patient underwent a supine position examination, with the feet positioned first, and the scanning range extended from the superior aspect of the diaphragm to the inferior margin of the liver. Scanning and imaging parameters included a tube voltage of 120 kVp, a tube current of 200 mA, a helical scanning mode with a pitch value of 1.35, a detector collimation of 128 mm × 0.6 mm, an acquisition matrix size of 512×512 pixels, a scanning field of view of 500 mm × 500 mm, a scanning layer thickness and spacing of 5 mm, and reorganization layer thickness and spacing of 1 mm.

CT evaluation

All imaging features were independently reviewed on a picture archiving and communication system (PACS; Zhejiang Greenlander Information Technology Co., Hangzhou, China) in our institution by two abdominal radiologists (with 5 and 10 years of respective work experience) using uniform criteria and region of interest (ROI) selection methods in a double-blind manner. Quantitative parameters were measured and averaged when measured in pairs, with any discrepancies being discussed and resolved by the board-certified radiologists.

In the evaluation based on CT examination, although the imaging principles are different from those of MRI, the imaging manifestations corresponding to some basic pathological changes caused by DILI share similarities, such as hepatocyte injury, biliary tract injury, and signs of ascites (9,20,21). CT features were assessed based on consistent criteria, including liver surface irregularities (surface nodularity) (10), transient hepatic attenuation difference (THAD) (areas of parenchymal the appear with hyperenhancement only or primarily during the hepatic arterial phase) (22), gallbladder wall edema (thickening or swelling of the gallbladder wall with low density) (23), periportal edema (abnormal hypoattenuation around the left and right portal veins and their branches) (9), bile duct dilatation (intrahepatic ducts >3 mm or extrahepatic ducts >8 mm in diameter) (24), portal lymphadenopathy (≥1 cm in the short axis of at least one lymph node in the porta hepatis) (10), splenomegaly (the edge of the spleen located below the left margin of the rib cage or over five ribs) (10), and ascites (an abnormal accumulation of fluid within the peritoneal cavity with low density) (25). The spectrum of CT findings of the liver were classified into three categories: normal, diffuse hepatic injury, and focal hepatic injury. Four circular ROIs measuring 1.0–3.5 cm² were manually positioned by each radiologist to target the left and right lobes of the liver within the focal area demonstrating drug-induced hepatic injury on CT images. Density intensity (DI) measurements of the hepatic parenchyma on nonenhanced images were obtained from these ROIs after imaging artifacts, visible vascular structures, areas of biliary dilatation, or possible non-DILI-induced lesions were excluded. The mean DI value from these ROIs represented the DI of the liver (DI_liver). The DI measurement for the spleen (DI_spleen) was taken at its midpoint. Quantitative liver-spleen contrast (Q-LSC) was calculated as follows: Q-LSC = DI_liver/DI_spleen (26,27).

Liver pathology assessment

The liver tissues were subjected to hematoxylin and eosin (HE) staining and well Masson’s trichrome staining, which was followed by an evaluation from two expert liver histopathologists. The histologic pattern was classified into hepatitis, cholestatic hepatitis, and cholestatic hepatitis based on the pathologic features proposed by Kleiner et al. (28). Furthermore, the severity of histological findings was categorized as mild, moderate, or severe according to the criteria proposed by Ishak et al. (29).

Statistical analysis

All statistical analyses were performed with SPSS software version 23.0 (IBM Corp., Armonk, NY, USA) and R software version 4.2.0 (The R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was set at P<0.05. The normality of distribution of continuous variables was assessed with a combination of the Shapiro-Wilk test, quantile-quantile (QQ) plots, and histograms. The results are expressed as the mean ± standard deviation for normally distributed variables, while skewed variables are presented as the median and interquartile range (IQR). The Student t-test or Mann-Whitney test was employed to compare continuous variables, while categorical variables were compared with the χ² test. The Cohen’s kappa test was employed to compute the kappa value between the independent evaluation from two radiologists. Univariate and multivariate analyses were applied to the identify independent risk factors for DILI and subsequently recorded the odds ratios (ORs) along with their corresponding 95% confidence intervals (CIs). The performance of the model was assessed with various metrics, including the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, and specificity.

Results

Demographic data and clinical characteristics

The demographic characteristics and laboratory parameters of patients with DILI are presented in Table 1. A total of 149 patients, with a mean age of 57.76±14.53 years (range, 11–88 years), were included in the study, of whom 83 (55.70%) were females. The types of liver injury observed were hepatocellular (n=68, 45.64%), mixed (n=28, 18.79%), and cholestatic (n=53, 35.57%). Significant differences in ALT, AST, ALP, TBIL, total cholesterol (TC), triglyceride (TG), low-density lipoprotein (LDL) levels, and INR values were found between the three clinical types of DILI (all P values <0.05). The severity of DILI was mild in 43 (28.86%) cases, moderate in 84 (56.38%), severe in 18 (12.08%), and fatal or required LT in 4 (2.68%). There were no statistically significant differences observed in TG, high-density lipoprotein (HDL), or alpha-fetoprotein (AFP) levels between different severity groups of DILI (Table S1). Figure 2 shows the main types of insulting drugs that induced DILI in different biochemical injury types. TCMs were the leading cause of DILI (n=66, 44.30%), followed by antituberculosis agents (n=28, 18.79%) and antineoplastic and biochemical agents (n=19, 12.75%).

Table 1. Demographic data, clinical, and laboratory characteristics of the enrolled patients with DILI.

Characteristics	Total (n=149)	Biochemical injury type			P value
		Hepatocellular (n=68, 45.64%)	Mixed (n=28, 18.79%)	Cholestatic (n=53, 35.57%)
Age (years)	57.76±14.53	58.04±15.84	54.21±13.13	59.26±13.36	0.325
Sex					0.423
Male	66 (44.30)	29 (42.65)	10 (35.71)	27 (50.94)
Female	83 (55.70)	39 (57.35)	18 (64.29)	26 (49.06)
ALT (U/L)	289 [75–573]	608 [374.75–894.5]	277.5 [152.5–380]	43 [20–91.5]	<0.001
AST (U/L)	182 [75.5–394.5]	343 [216.25–813.75]	234.5 [114.75–370.5]	64 [32.5–96]	<0.001
AST/ALT	0.84 [0.5–1.5]	0.6 [0.47–1.03]	0.81 [0.55–1.06]	1.47 [0.85–1.9]	<0.001
ALP (U/L)	230 [143–326.5]	149.5 [120–194.75]	249 [151.5–371]	319 [248.5–432.5]	<0.001
GGT (U/L)	191 [107–369.5]	201 [119.25–323.5]	213.5 [101.25–714]	166 [76.5–387]	0.349
TBIL (μmol/L)	45.6 [18.8–139.1]	47.75 [23.08–184.45]	110.2 [26.5–171.43]	21.9 [14.5–55.8]	0.001
DBIL (μmol/L)	14.4 [4.6–78.5]	20.7 [8.15–133.05]	62.25 [10.23–124.88]	5.2 [2.8–21.8]	<0.001
IBIL (μmol/L)	23.1 [13.2–49.5]	25.6 [14.7–53.7]	40.6 [14.38–53.03]	16.1 [10.7–40.65]	0.004
TBA (μmol/L)	37 [18.07–125.55]	51.5 [19.13–137.85]	54.5 [25.32–166.95]	28 [14.45–116.4]	0.139
ALB (g/L)	36.9 [32.15–40.35]	36.2 [31.9–38.83]	37 [30.6–41.35]	37.5 [33.65–41.2]	0.364
Cr (mmol/L)	64.2 [53.25–76.3]	65.05 [54.93–77.7]	63.75 [51.55–74.55]	64.1 [52.6–76.15]	0.404
TC (mmol/L)	4.36 [3.52–5.34]	3.83 [3.07–4.41]	5.12 [4.05–6.28]	4.64 [3.93–5.71]	<0.001
TG (mmol/L)	1.37 [0.94–2.14]	1.25 [0.95–1.77]	2.17 [1.20–3.12]	1.2 [0.865–2.385]	0.018
HDL (mmol/L)	0.93 [0.7–1.12]	0.94 [0.69–1.1]	0.82 [0.67–1.02]	1.01 [0.73–1.185]	0.13
LDL (mmol/L)	2.32 [1.86–2.83]	2.15 [1.81–2.64]	2.64 [2.37–3.15]	2.29 [1.51–2.84]	0.004
INR	1.09 [1.02–1.18]	1.14 [1.06–1.31]	1.04 [0.97–1.16]	1.08 [0.98–1.16]	0.002
AFP (μg/L)	3.3 [2.4–5.4]	3.25 [2.23–7.95]	4.2 [2.53–7.55]	3.3 [2.5–3.8]	0.277
Class of agent(s)
TCM	66 (44.30)	33 (48.53)	9 (32.14)	24 (45.28)
Antituberculosis	28 (18.79)	13 (19.12)	7 (25.00)	8 (15.09)
Antineoplastic and biochemical	19 (12.75)	8 (11.76)	4 (14.29)	7 (13.21)
HDS	11 (7.38)	6 (8.82)	2 (7.14)	3 (5.66)
HDS and TCM	7 (4.70)	3 (4.41)	2 (7.14)	2 (3.77)
Biochemical	11 (7.38)	3 (4.41)	3 (10.71)	5 (9.43)
Others	7 (4.70)	2 (2.94)	1 (3.57)	4 (7.55)
Severity					<0.001
Mild	43 (28.86)	5 (7.35)	6 (21.43)	32 (60.38)
Moderate	84 (56.38)	49 (72.06)	19 (67.86)	16 (30.19)
Severe	18 (12.08)	11 (16.18)	3 (10.71)	4 (7.55)
Fatal/LT	4 (2.68)	3 (4.41)	0 (0.00)	1 (1.89)

Data are presented as mean ± standard deviation, median [IQR], and n (%). AFP, alpha-fetoprotein; ALB, albumin; ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate transaminase; Cr, creatinine; DBIL, direct bilirubin; DILI, drug-induced liver injury; GGT, gamma-glutamyl transferase; HDL, high-density lipoprotein; HDS, herbal and dietary supplements; IBIL, indirect bilirubin; INR, international normalized ratio; IQR, interquartile range; LDL, low-density lipoprotein; LT, liver transplantation; TBA, total bile acid; TBIL, total bilirubin; TC, total cholesterol; TCM, traditional Chinese medicine; TG, triglyceride.

Figure 2.

The primary types of insulting drugs that induced DILI. DILI, drug-induced liver injury; HDS, herbal and dietary supplements; TCM, traditional Chinese medicine.

CT features of DILI

The CT image characteristics of the different clinical subtypes are summarized in Table 2. Contrast-enhanced CT revealed abnormal liver injury in 61 patients (40.94%), with focal liver injury observed in 36 patients (24.16%) and diffuse liver injury observed in 25 patients (16.78%). Splenomegaly was present in 55 (36.91%) patients, while gallbladder wall edema and periportal edema were found in 36 (24.16%) and 32 (21.48%) patients, respectively. Additionally, ascites, THAD, bile duct dilatation, and hepatic surface irregularities were identified in 28 (18.79%), 25 (16.78%), 24 (16.11%), and 15 (10.07%) patients, respectively. The mean value of Q-LSC was determined to be 1.16±0.15, which was higher in the cholestatic liver injury group as compared to hepatocellular liver injury and mixed liver injury groups. The interrater agreement between the two radiologists for hepatic surface irregularities, THAD, and portal lymphadenopathy yielded kappa values of high significance (κ=0.915, κ=0.987, and κ=0.893, respectively).

Table 2. Comparison of CT features between the different biochemical injury types in patients with DILI.

Parameters	Total (n=149)	Biochemical injury type			P value
		Hepatocellular (n=68)	Mixed (n=28)	Cholestatic (n=53)
Irregularity of liver surface	15 (10.07)	8 (11.76)	2 (7.14)	5 (9.43)	0.824
THAD	25 (16.78)	9 (13.24)	6 (21.43)	10 (18.87)	0.502
Gallbladder wall edema	36 (24.16)	21 (30.88)	7 (25.00)	8 (15.09)	0.128
Periportal edema	32 (21.48)	15 (22.06)	7 (25.00)	10 (18.87)	0.809
Biliary duct dilatation	24 (16.11)	12 (17.65)	4 (14.29)	8 (15.09)	0.916
Portal lymphadenopathy	16 (10.74)	9 (13.24)	3 (10.71)	4 (7.55)	0.608
Splenomegaly	55 (36.91)	15 (22.06)	10 (35.71)	30 (56.60)	<0.001
Ascites	28 (18.79)	13 (19.12)	6 (21.43)	9 (16.98)	0.856
Q-LSC	1.14±0.19	1.05±0.16	1.16±0.15	1.23±0.20	0.439
Spectrum of CT findings	61 (40.94)	27 (39.71)	14 (50.00)	20 (37.74)	0.79
Diffuse hepatic injury	25 (16.78)	10 (14.71)	5 (17.86)	10 (18.87)
Focal hepatic injury	36 (24.16)	17 (25.00)	9 (32.14)	10 (18.87)

Data are presented as n (%) or mean ± standard deviation. CT, computed tomography; DILI, drug-induced liver injury; Q-LSC, quantitative liver-spleen contrast; THAD, transient hepatic attenuation difference.

Comparison of CT features between different biochemical injury types

The incidence of splenomegaly was notably elevated in patients with cholestatic injury (56.60%) as compared to those with hepatocellular (35.71%) and mixed injury (22.06%) (P<0.001). However, no statistically significant distinctions were found between the three groups in terms of CT imaging features such as liver surface irregularities, THAD, gallbladder wall edema, bile duct dilatation, portal lymphadenopathy, ascites, and Q-LSC. The CT findings are shown in Figure 3.

Figure 3.

The multiphase enhanced CT imaging features in three patients with DILI. (A,B) A 55-year-old female patient with pancreatic cancer developed liver injury after combination therapy with paclitaxel and gemcitabine. (A) Non-contrast CT image demonstrating hepatomegaly with hypoattenuation of the liver parenchyma consistent with hepatic steatosis. (B) Contrast-enhanced CT image obtained 1.5 months after chemotherapy showing mottled hepatic enhancement. (C,D) A 33-year-old female patient with rectal cancer who underwent preoperative oxaliplatin and capecitabine regimen for adjuvant chemotherapy. (C) Pretreatment image showing no liver abnormalities. (D) Contrast-enhanced CT image obtained 2 months after chemotherapy showing focal hepatic injury. (E,F) A 68-year-old female patient with rectal cancer who underwent postoperative immunosuppressive therapy. (E) Pretreatment image showing no liver abnormalities. (F) Contrast-enhanced CT image obtained 1 month after treatment demonstrating heterogeneous enhancement of the liver with periportal oedema, left intrahepatic bile duct dilatation, and effusion in the abdominal cavity. CT, computed tomography; DILI, drug-induced liver injury.

Identification of CT features associated with DILI severity

In the comparison of the imaging findings between the level 3–4 group and level 1–2 group (Table 3), there were no significant differences in THAD, splenomegaly, or the spectrum of CT findings (P>0.05). However, significant differences were found regarding hepatic surface irregularity, gallbladder wall edema, periportal edema, bile duct dilatation, portal lymphadenopathy, ascites, and Q-LSC (P<0.05). To identify factors associated with severity grades 3–4, multivariate factor analysis was conducted by including factors with P<0.05 as independent variables. The results revealed that the independent contributors to increased risk were ascites (OR =70.83; 95% CI: 16.34–306.99; P<0.001) and Q-LSC (OR =0.002; 95% CI: 0.00–0.13; P=0.003).

Table 3. Identification of CT features independently associated with DILI severity.

Parameters	Level 1–2 (n=127, 85.23%)	Level 3–4 (n=22, 14.77%)	Univariate			Multivariate
			OR (95% CI)	P value		OR (95% CI)	P value
Irregularity of liver surface	8 (6.30)	7 (31.82)	6.94 (2.20–21.87)	0.001
THAD	22 (17.32)	3 (13.64)	0.75 (0.21–2.77)	0.67
Gallbladder wall edema	22 (17.32)	14 (63.64)	8.35 (3.13–22.32)	<0.001
Periportal edema	21 (16.54)	11 (50.00)	5.05 (1.94–13.16)	0.001
Biliary duct dilatation	17 (13.39)	7 (31.82)	3.02 (1.08–8.48)	0.036
Portal lymphadenopathy	10 (7.87)	6 (27.27)	4.39 (1.41–13.71)	0.011
Splenomegaly	44 (34.65)	11 (50.00)	1.89 (0.76–4.70)	0.173
Ascites	10 (7.87)	18 (81.82)	52.65 (14.92–185.85)	<0.001		70.83 (16.34–306.99)	<0.001
Q-LSC	1.16±0.19	1.02±0.11	0.001 (0.001–0.19)	0.002		0.002 (0.00–0.13)	0.003
Spectrum of CT findings	46 (36.22)	13 (59.09)		0.342			0.004
Diffuse hepatic injury	21 (16.54)	4 (18.18)	1.89 (0.508–7.025)
Focal hepatic injury	25 (19.69)	9 (40.91)	0.583 (0.164–2.081)

Data are presented as n (%) or mean ± standard deviation. CI, confidence interval; CT, computed tomography; DILI, drug-induced liver injury; OR, odds ratio; Q-LSC, quantitative liver-spleen contrast; THAD, transient hepatic attenuation difference.

The progression of DILI severity grading was assessed with three models developed via binary logistic regression analysis. Model 1 included Q-LSC, Model 2 included ascites, and Model 3 included both Q-LSC and ascites. Figure 4 presents the predictive accuracy of three models. Model 1 demonstrated a moderate AUC (AUC =0.628; 95% CI: 0.487–0.769) in predicting the level 3–4 group and level 1–2 group. Model 2 had a higher AUC, achieving a better prediction of severity (AUC =0.870; 95% CI: 0.772–0.968). By combining Q-LSC and ascites in Model 3, the AUC further increased, indicating enhanced predictive power (AUC =0.929; 95% CI: 0.867–0.991; sensitivity =0.818; specificity =0.953).

Figure 4.

The ROC curves of three different models for predicting the progression of DILI severity. AUC, area under the receiver operating characteristic curve; DILI, drug-induced liver injury; ROC, receiver operating characteristic.

Histopathological features of DILI

A total of seven patients underwent liver biopsy, with histological classification revealing acute hepatitis (n=0), chronic hepatitis (n=4), cholestatic hepatitis (n=2), acute cholestasis (n=0), and chronic cholestasis (n=1). Common findings included hepatocellular steatosis, necrosis, inflammation in the portal area, and bile duct hyperplasia. Focal necrosis was observed in 5 (71.4%) cases, fusional necrosis in 1 (14.3%) case, and focal necrosis with fusional necrosis in 1 (14.3%) case. Interfacial inflammation was present in 5 (71.4%) cases to varying degrees, while high infiltration of lymphocytes with eosinophils occurred in 4 (57.1%) cases. Histological features of chronic hepatitis included mild-to-moderate interfacial hepatitis, inflammation of the confluent area, fibroplasia, and formation of fibrous septa. The typical pathological staging is depicted in Figure 5. Immunohistochemical staining was conducted in seven cases, which resulted in positive staining for CD34, CK7, CK19, and CK8/18 and negative staining for CD5.

Figure 5.

Histopathological characteristics of patients with DILI (HE staining; 100×). (A) HE staining revealing focal hepatocyte vesicular steatosis, some hepatocellular swelling with ballooning and feathery degeneration, and the pseudoglandular arrangement of hepatocytes. (B) HE staining indicating centrilobular pallor in the hepatocytes surrounding the central vein and shrunken nuclei in some cells. (C) HE staining demonstrating well-defined areas of hepatocyte necrosis with lymphocytes, eosinophil, and plasma cell infiltration. Structural disruption of the hepatic lobules with visible bridging necrosis. (D) Masson staining showing liver fibrosis. DILI, drug-induced liver injury; HE, hematoxylin and eosin.

Discussion

In this study, splenomegaly was significantly more frequent in patients with cholestatic injury. The univariate analysis revealed that irregularity of the liver surface, gallbladder wall edema, periportal edema, biliary duct dilatation, portal lymphadenopathy, ascites, and Q-LSC were independently associated with severe injury. We developed a novel logistic binary analysis model incorporating ascites and Q-LSC to effectively identify the independent prognostic factors in DILI, which demonstrated excellent performance, with an AUC of 0.929, a sensitivity of 0.818, and a specificity of 0.953. To the best of our knowledge, this study is the first to visually evaluate severe DILI based on noninvasive CT imaging features, and our findings can be used to assist in decision-making for the timely intervention into adverse events in patients with DILI and the early identification of patients who require transplantation.

The pathogenesis of DILI is complex and involves a combination of concurrent mechanisms that have not yet been fully elucidated. These potential pathogenic processes can be categorized into direct hepatotoxicity and idiosyncratic hepatotoxicity. Direct hepatotoxicity of drugs refers to the direct damage caused by the ingested drug and/or its metabolites, typically exhibiting a predictable dose-dependent pattern (30). Conversely, specific hepatotoxicity is unpredictable and occurs when drugs and their reactive metabolites induce mitochondrial damage, leading to oxidative stress in hepatocytes and subsequent injury or death (31). Additionally, genetic polymorphisms may result in enzyme dysfunction and abnormalities in transporter proteins, thereby increasing susceptibility to DILI. Notably, drugs triggering liver injury also initiate restorative tissue repair, which plays a crucial role in limiting and reversing liver injury (1).

The distinctive pattern of elevated biochemical markers offers valuable insights into the clinical type, etiology, and prognosis of biochemical damage in DILI, thereby enhancing the diagnostic accuracy of DILI. It has been suggested that patients with hepatocellular DILI exhibit higher TBIL levels as compared to those with cholestatic or mixed DILI. This may be attributed to the fact that cholestasis-related biochemical markers such as ALP and gamma-glutamyl transferase are only present during the early stages of cholestasis. However, this study demonstrates that TBIL levels are elevated in patients with mixed DILI relative to those with other types of biochemical impairment (32). This observation can be attributed to concurrent damage to both hepatocytes and bile ducts in individuals with mixed-type DILI resulting in impaired bilirubin esterification and secretion barriers.

Among the various clinically staged CT imaging features observed in patients with DILI, focal and diffuse liver injuries were the most common positive manifestations on CT scans, followed by splenomegaly and gallbladder wall edema. Diffuse injury manifestations included hepatitis, steatosis, steatohepatitis, cholestasis, hepatic graft-versus-host disease, sinusoidal obstruction syndrome (SOS), fibrosis, and cirrhosis. Focal hepatic lesions encompassed treatment-related vascular injuries such as focal nodular hyperplasia-like lesions and hepatic perfusion abnormalities, as well as hepatic infections resulting from immunosuppression. Our study revealed that focal liver lesions were more prevalent than were diffuse liver injuries in patients with DILI. Additionally, this study assessed the extent of hepatic steatosis using Q-LSC measured through CT imaging. The liver-spleen ratio progressively decreased in patients with cholestatic, mixed-type, and hepatocellular DILI due to the expected hypodensity of the liver caused by hepatic steatosis. Furthermore, we found a higher incidence of splenomegaly in cases of cholestatic liver injury, which may be attributed to the functional hemodynamic changes associated with necrosis (33).

The results of univariate analysis from our revealed that severe DILI cases were more likely to exhibit liver surface irregularity, gallbladder wall edema, periportal edema, bile duct dilatation, periportal lymph node enlargement, and ascites. An MRI study indicated that liver surface irregularity could be attributed to multilobar collapse with compliant hepatic necrosis and disproportionate hepatic regeneration (34). Wu et al. (10) also reported a higher prevalence of gallbladder wall edema in severe-to-fatal or long-term DILI cases, which is in line with the results obtained from our experimental findings. Shu et al. (35) found there to be an association between gallbladder wall edema and the degree of hepatic necrosis inflammatory activity due to the extension of hepatic necro-inflammatory activity along Glisson’s capsule to involve the wall of the gallbladder. Masuoka et al. (9) observed that bile duct abnormalities are frequently observed on CT images in patients with immune checkpoint inhibitor-induced liver injury, suggesting their potential utility for the early recognition and prognosis improvement in such cases. Our study provides novel evidence indicating a higher propensity for portal lymphadenopathy development among individuals at high risk for DILI; additionally, enlarged lymph nodes in the hepatic hilar region often indicate conditions such as inflammation of the liver or biliary system. Ascites may result from acute severe liver injury impairing short-term synthetic function or be associated with cirrhotic portal hypertension (36,37).

In this study, binary logistic regression analysis based on CT imaging features revealed that the presence of ascites and liver-spleen ratio were independent predictors for identifying unfavorable outcomes in DILI. A retrospective investigation conducted by Sharma et al. (38) concluded that chemotherapy-induced liver injury with CT image presentation features such as hepatosplenomegaly and ascites was associated with SOS. However, another retrospective study reported that hepatic surface irregularity, THAD, and splenomegaly predicted poor prognosis in patients with DILI, which contradicts our predictor results. This discrepancy may be attributed to variations in the selection of examination types resulting in different interpretations of the imaging features. Our study identified Q-LSC as a significant imaging feature, which potentially reflects the interplay between hepatic and splenic metabolic activities in DILI. The hepatic steatosis score can be quantitatively assessed through Q-LSC value measurements. Excessive hepatic lipid deposition induces a state of chronic low-grade inflammation, while the spleen plays a pivotal role in modulating metabolic processes. Disruption of the liver-spleen axis may influence the progression and severity of DILI. Tsushima et al. (39) demonstrated through CT imaging that splenomegaly is frequently observed in patients with nonalcoholic fatty liver disease. Additionally, a study employing dynamic positron emission tomography to assess the metabolic rate of glucose (MRglu) demonstrated that MRglu is consistently and significantly elevated in patients with hepatic steatosis, exhibiting an inverse correlation with liver CT density (40). Notably, in male patients, the spleen longitudinal diameter was significantly correlated with hepatic CT density of hepatic MRglu and splenic MRglu.

There are certain limitations in this study that should be acknowledged. First, the diagnosis of DILI was primarily based on clinical assessment and laboratory indicators, which might have introduced selection bias, and a comprehensive gold standard of histopathological examination was lacking. Second, as different clinicians treated the patients included in this retrospective study, the potential influence of varying treatment approaches on the prognosis of patients with DILI cannot be disregarded. Third, the 8-year time span of the dataset and the absence of patients with DILI limited the model’s ability to determine the value of stage-specific intervals. Future studies will include multiple centers and increase the diversity and number of DILI cases. A 3-month interval may lead to the regression of notable acute manifestations such as edema, ascites, and inflammation prior to follow-up scans, causing misclassification of severity and potential bias in the risk stratification models. In future research, we will enlarge the sample size and categorize DILI follow-up scans within 4 weeks, 4–8 weeks, and 8 weeks to 3 months to better characterize the dynamic imaging changes from acute to chronic DILI. Finally, our study solely relied on enhanced abdominal CT imaging features to evaluate the biochemical injury type and severity; however, we aim to expand our investigation by incorporating MRI examinations in future studies.

Conclusions

Significant variations in serological indices of DILI were observed across different biochemical injury types and severity grades. Moreover, ascites and Q-LSC emerged as potential risk factors for predicting the severity of DILI. Combining ascites and Q-LSC could aid in distinguishing between mild-to-moderate and severe-to-fatal or LT-requiring cases of DILI. Future studies will incorporate multicenter data to develop a comprehensive model integrating CT and MR imaging features for identifying high-risk groups and predicting poor prognosis in patients with DILI.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by Institutional Review Boards of the Affiliated Xihu Hospital of Hangzhou Medical College (No. 20241114/32/01/001) and The Second Affiliated Hospital of Zhejiang Chinese Medical University (No. 2025-LW-096-01). The requirement for individual consent was waived due to the retrospective nature of the analysis.

Data Sharing Statement

Available at https://qims.amegroups.com/article/view/10.21037/qims-24-2110/dss

qims-15-09-8553-dss.pdf ^{(125.6KB, pdf)}

DOI: 10.21037/qims-24-2110