Editor’s Note:
Drug-induced liver injury (DILI) is liver damage caused by prescription or over-the-counter chemical medicines, biological products, traditional Chinese medicines, herbal and natural medicines, health products, dietary supplements, or their metabolites, additives, contaminants, and impurities. It is one of the most concerning drug-related diseases globally. At the inaugural 2024 Jinling Liver Disease Conference, Professor Dong Ji from the Fifth Medical Center of the Chinese PLA General Hospital delivered a keynote report on the prognosis of DILI, discussing the histological and biochemical characteristics of chronic DILI and their relationship to its prognosis, and introducing a new non-invasive predictive model for forecasting the prognosis of chronic DILI.
Incidence of DILI in China:
A retrospective study in China estimates the annual incidence of DILI at 23.80 per 100,000 individuals, with a noted increase in recent years. Although most DILI cases recover liver function after discontinuation of the offending drug, 8%-25% progress to chronic DILI. Recurrent episodes can lead to cirrhosis or even liver failure, imposing significant psychological and financial burdens on patients and their families. Therefore, identifying and determining factors affecting the prognosis of chronic DILI and implementing individualized management (such as follow-up frequency, treatment intensity, liver transplant decisions) for patients at different risk levels are crucial clinical issues needing resolution.
Clinical Classification of DILI:
Based on the mechanism of injury, DILI can be divided into intrinsic (the inherent toxicity of a drug or its metabolite to the liver, dose-dependent, commonly seen with acetaminophen, amiodarone, statins); idiosyncratic (metabolic or immune idiosyncrasy, usually dose-independent but may require a certain dose threshold, common with antibiotics, some traditional medicines); and indirect (drugs causing liver injury indirectly by altering existing liver disease or immune status, with unclear dose correlation, common with anti-cancer drugs, immune checkpoint inhibitors).
Current Research on DILI:
Domestic studies are mostly retrospective cross-sectional surveys with relatively few cases (less than 500), low proportions of liver biopsies, focusing on incidence, causative drugs, age distribution, and clinical characteristics, with few cohort studies on prognosis; international studies pay more attention to biomarkers, mechanisms, specific drug-induced liver injury research, and the development of animal models. Some also focus on the risk factors for short-term prognosis, but the case numbers are relatively low.
A Spanish prospective study followed up DILI patients for up to three years to analyze the time of liver enzyme resolution to determine the best definition and risk factors for chronic DILI. It found that older age, female gender, and cholestatic or mixed liver injury were associated with poor prognosis or prolonged recovery, with one year as the optimal time point for defining chronic or prolonged recovery. ALP>1.1×ULN and TBIL>2.8×ULN were the best thresholds for predicting chronicity or prolonged recovery.
A prospective study from the USA defined persisters as patients with persistent liver injury (serum AST or ALT>1.5×ULN or ALP>1×ULN) 12 months after a DILI episode, with non-qualifiers classified as resolvers. Follow-ups exceeding two years were conducted, with life quality assessments (SF-36 scores) performed at 6, 12, and 24 months. The study found significant reductions in the physical functioning score (PF) (51.9 vs. 77, P=0.001), overall health score (GH) (56.2 vs. 66.3, P=0.044), and total body score (PCS) (35 vs. 42.4, P=0.003) among persisters, indicating that persistent liver injury severely affects patients’ quality of life. Another prospective study based on the US Drug-Induced Liver Injury Network (DILIN) assessed the frequency and risk factors for persistent biochemical abnormalities in DILI patients followed for more than two years. Including 113 patients, multivariate analysis revealed that older age (OR[95%CI] 1.03[0.99, 1.07], P=0.067) and higher baseline ALP (OR[95%CI] 1.18[1.03, 1.35], P=0.03) were independent risk factors for persisters; patients with cholestatic liver injury had a higher risk of becoming persisters during long-term follow-up (54% vs. 20%, P<0.01).In this study, only 17 patients underwent liver biopsies, with 9 patients (53%) showing pathological features of chronic cholestasis, 3 exhibiting chronic hepatitis or steatohepatitis, and 6 with bile duct loss (all of whom showed progression of fibrosis during follow-up).
In 2019, a study published in Hepatology explored biological markers for the diagnosis and prognosis of drug-induced liver injury (DILI). It found that traditional biomarkers, including International Normalized Ratio (INR), AST, and TBIL, along with potential markers such as osteopontin (OPN), cytokeratin-18 (K18), macrophage colony-stimulating factor receptor (MCSFR), caspase-cleaved cytokeratin-18 (ccK18), fatty acid-binding protein 1 (FABP1), and AFP, were significantly elevated and could predict mortality/liver transplantation. The strongest correlations with mortality/transplantation were seen with INR (AUC 0.920) and OPN (AUC 0.858).
A large retrospective cohort study developed a prediction model based on platelets and bilirubin to assess the risk of acute liver failure (ALF) in patients with DILI. It found that patients who developed ALF had higher levels of ALT, AST, ALP, and TBIL, and lower platelets at the time of DILI diagnosis compared to those who did not develop ALF. The model built with platelets and TBIL demonstrated the highest discrimination, with a high-risk threshold of -1.08141, a sensitivity of 0.91 (95% CI: 0.71-0.99), a specificity of 0.76 (95% CI: 0.75-0.77), a negative predictive value of 0.99 (95% CI: 0.99–1.0), and a positive predictive value of 0.01 (95% CI: 0.005–0.01). These results were validated in a sample of 76 patients from the Penn cohort (University of Pennsylvania).
A recent multicenter retrospective study published in J Hepatol included 294 patients from the iDILIC network who met the criteria for analysis. The study found that ALP, TBIL, the timing of the DILI episode, and the extent of liver metabolism were independent predictors of prolonged recovery from DILI. Using these indicators, an AFT model was constructed to calculate patient scores, dividing them into high-risk (score >1.30), uncertain (1.30 < score ≤ 0.44), and low-risk (score ≤ 0.44) groups. Log-rank tests showed significant differences between the high and low-risk groups (P<0.0001). These results were successfully validated in the Spanish cohort (N=257) and LiverTox cohort (N=191). However, only 5 patients in the study were followed up for at least 12 months, so the model could not predict chronic DILI persisting for ≥1 year, and there was a lack of histological assessment.
To explore the prognostic model of chronic DILI：
Professor Dong Ji shared recent work his team has done on predicting outcomes in chronic DILI. They analyzed 441 elderly DILI patients (>60 years old) and found a significantly higher proportion of females had autoimmune phenomena (IgG >1.1×ULN or anti-nuclear/anti-smooth muscle antibodies >1:80), and those with autoimmune phenomena had notably higher levels of liver inflammation and fibrosis, thus necessitating more aggressive follow-up and treatment.
A second study established a prognosis model for drug-induced acute liver failure (DIALF-5 model), consisting of artificial liver support, N-acetylcysteine, vasopressor drugs, INR, and hepatic encephalopathy. This model was as accurate as the previously established ALFSG-PI but simpler to calculate and easier to implement clinically.
The third study established a liver biochemical non-resolution model (BNR-6) based on a multicenter liver biopsy cohort of 5,326 patients diagnosed with chronic DILI. The model consists of six parameters: age, gender, AST, TBIL, PT, and PLT. Risk stratification is divided based on ROC threshold values: low risk 0-28 points (≤10%), medium risk 28-50 points (10%-70%), and high risk >50 points (>70%). Clinical application has shown that the BNR-6 model highly correlates with liver histology, allowing for the individualized calculation of BNR risk for each patient. When the BNR-6 score is significantly high, it predicts a high risk of delayed recovery in chronic DILI patients, suggesting a need for more aggressive treatment and close follow-up (8% of the high-risk group should consider more aggressive treatments, including steroids and artificial liver); when the BNR-6 score is significantly low, it indicates a lower risk of BNR (50.4% of the low-risk group can avoid liver biopsy using the BNR-6 model), allowing for fewer follow-ups, saving medical expenses, and more economical monitoring of prognosis. With the BNR-6 model, chronic DILI patients can be accurately differentiated, avoiding the potential harm of biopsy to the liver, making it easy to implement, and widely applicable in clinical diagnosis and treatment of liver diseases and safe medication practices.
