Predicting Antituberculosis Drug-Induced Liver Injury Using an Interpretable Machine Learning Method: Model Development and Validation Study

Risk Prediction of Liver Injury in Pediatric Tuberculosis Treatment: Development of an Automated Machine Learning Model

Purpose

Drug-induced liver injury (DILI) is one of the most common and serious adverse drug reactions related to first-line anti-tuberculosis drugs in pediatric tuberculosis patients. This study aims to develop an automatic machine learning (AutoML) model for predicting the risk of anti-tuberculosis drug-induced liver injury (ATB-DILI) in children.

Methods

A retrospective study was performed on the clinical data and therapeutic drug monitoring (TDM) results of children initially treated for tuberculosis at the affiliated Changsha Central Hospital of University of South China. After the features were screened by univariate risk factor analysis, AutoML technology was used to establish predictive models. The area under the receiver operating characteristic curve (AUC) was used to evaluate model's performance, and then the TreeShap algorithm was employed to interpret the variable contributions.

Results

A total of 184 children were enrolled in this study, of whom 19 (10.33%) developed ATB-DILI. Univariate analysis showed that seven variables were risk factors for ATB-DILI, including the plasma peak concentration (Cmax) of rifampicin, body mass index (BMI), alanine aminotransferase, total bilirubin, total bile acids, aspartate aminotransferase and creatinine. Among the numerous predictive models constructed by the "H2O" AutoML platform, the gradient boost machine (GBM) model exhibited the superior performance with AUCs of 0.838 and 0.784 on the training and testing sets, respectively. The TreeShap algorithm showed that Cmax of rifampicin and BMI were important features that affect the AutoML model's performance.

Conclusion

The GBM model established by AutoML technology shows high predictive accuracy and interpretability for ATB-DILI in children. The prediction model can assist clinicians to implement timely interventions and mitigation strategies, and formulate personalized medication regimens, thereby minimizing potential harm to high-risk children of ATB-DILI.

Diagnosis, prevention and risk-management of drug-induced liver injury due to medications used to treat mycobacterium tuberculosis

INTRODUCTION

Many of the first line medications for the treatment of active and latent M. tuberculosis are hepatoxic and cause a spectrum of anti-tuberculosis drug induced liver injury (ATLI), including acute liver failure (ALF). Despite advances in recognition of and prevention of ATLI, isoniazid remains one of the leading causes of DILI as well as drug-induced ALF.

AREAS COVERED

A literature search of the incidence, risk factors, current societal guidelines, monitoring, and prophylactic medication usage in ATLI was performed using PubMed and institutional websites. Relevant articles from 1972 to 2024 were included in this review.

EXPERT OPINION

Current societal guidelines regarding ATLI monitoring are mixed, but many recommend liver enzyme testing of high-risk populations. We recommend liver test monitoring for all patients on multi-drug therapy as well as those on isoniazid therapy. Precision medicine practices, such as N-acetyltransferase-2 polymorphism genotyping, are thought to be beneficial in reducing the incidence of ATLI in high-risk populations. However, broader implementation is currently cost prohibitive. Hepatoprotective drugs are not currently recommended, although we do recognize their potential. In patients who develop ATLI but require ongoing anti-TB treatment, strategies to restart the same or less hepatotoxic regimens are currently being followed.

Interpretable machine learning in predicting drug-induced liver injury among tuberculosis patients: model development and validation study

BACKGROUND

The objective of this research was to create and validate an interpretable prediction model for drug-induced liver injury (DILI) during tuberculosis (TB) treatment.

METHODS

A dataset of TB patients from Ningbo City was used to develop models employing the eXtreme Gradient Boosting (XGBoost), random forest (RF), and the least absolute shrinkage and selection operator (LASSO) logistic algorithms. The model's performance was evaluated through various metrics, including the area under the receiver operating characteristic curve (AUROC) and the area under the precision recall curve (AUPR) alongside the decision curve. The Shapley Additive exPlanations (SHAP) method was used to interpret the variable contributions of the superior model.

RESULTS

A total of 7,071 TB patients were identified from the regional healthcare dataset. The study cohort consisted of individuals with a median age of 47 years, 68.0% of whom were male, and 16.3% developed DILI. We utilized part of the high dimensional propensity score (HDPS) method to identify relevant variables and obtained a total of 424 variables. From these, 37 variables were selected for inclusion in a logistic model using LASSO. The dataset was then split into training and validation sets according to a 7:3 ratio. In the validation dataset, the XGBoost model displayed improved overall performance, with an AUROC of 0.89, an AUPR of 0.75, an F1 score of 0.57, and a Brier score of 0.07. Both SHAP analysis and XGBoost model highlighted the contribution of baseline liver-related ailments such as DILI, drug-induced hepatitis (DIH), and fatty liver disease (FLD). Age, alanine transaminase (ALT), and total bilirubin (Tbil) were also linked to DILI status.

CONCLUSION

XGBoost demonstrates improved predictive performance compared to RF and LASSO logistic in this study. Moreover, the introduction of the SHAP method enhances the clinical understanding and potential application of the model. For further research, external validation and more detailed feature integration are necessary.

Causal Evaluation of Post-Marketing Drugs for Drug-induced Liver Injury from Electronic Health Records

Drug-induced liver injury (DILI) is one of the most common and serious adverse drug reactions that can lead to acute liver failure and death. Detection of DILI and causal estimation of drug-hepatotoxicity association are of great importance for patient safety. This paper proposes a framework for causal estimation of post-marketing drugs for DILI from real-world electronic health record (EHR) data. Randomized clinical trials were replicated at scale by automatically generating different user and non-user cohorts for each potential drug, and average treatment effects (ATEs) of drugs were estimated using targeted maximum likelihood estimation. Ten years of real-world EHRs were used to validate the framework. Of all 1199 single-ingredient drugs analyzed, 7 novel and 7 known drug-hepatotoxicity associations were found to be causal.

A nomogram model to predict the risk of drug-induced liver injury in patients receiving anti-tuberculosis treatment

Objectives: To establish an individualized nomogram to predict the probability of drug-induced liver injury (DILI) in tuberculosis patients receiving anti-tuberculosis treatment. Methods: The clinical information of patients admitted to a tertiary hospital between January 2010 and December 2022 was retrospectively reviewed from the clinical records. Patients with baseline liver diseases (hepatis B or C infection and fatty liver) or taking liver protective drugs were excluded. The maximum values in liver function test within 180 days after anti-tuberculosis treatment were collected to determine the occurrence of DILI. The candidate variables used for establishing prediction model in this study were the last results within the 30 days before the treatment onset. The final variables were included after univariate and multivariate logistic regression analyses and applied to establish the nomogram model. The discrimination power and prediction accuracy of the prediction model were assessed using the area under the receiver operating characteristic (AUC) curve and a calibration chart. The clinical effectiveness was assessed via decision curve analysis (DCA). The established model was validated in two validation groups. Results: A total of 1979 patients with 25 variables were enrolled in this study, and the incidence of DILI was 4.2% (n = 83). The patients with complete variables were divided into training group (n = 1,121), validation group I (n = 492) and validation group II (n = 264). Five variables were independent factors for DILI and included in the final prediction model presented as nomogram: age (odds ratio [OR] 1.022, p = 0.023), total bilirubin ≥17.1 μmol/L (OR 11.714, p < 0.001), uric acid (OR 0.977, p = 0.047), neutrophil count (OR 2.145, 0.013) and alcohol consumption (OR 3.209, p = 0.002). The AUCs of the prediction model in the training group, validation group I and validation group II were 0.833, 0.668, and 0.753, respectively. The p-values of calibration charts in the three groups were 0.800, 0.996, and 0.853. The DCA curves of the prediction model were above the two extreme curves. Conclusion: The nomogram model in this study could effectively predict the DILI risk among patients under anti-tuberculosis drug treatment.

Highlights of the drug-induced liver injury literature for 2021

INTRODUCTION

In 2021, over 3,000 articles on Drug-Induced Liver Injury (DILI) were published, nearly doubling the annual number compared to 2011. This review selected DILI articles from 2021 we felt held the greatest interest and clinical relevance.

AREAS COVERED

A literature search was conducted using PubMed between 1 March 2021 and 28 February 2022. 86 articles were included. This review discusses new and established cases of hepatotoxins, including new FDA approvals and COVID-19 therapeutics. Developments in biomarkers and causality assessment methods are discussed. Updates from registries are also explored.

EXPERT OPINION

DILI diagnosis and prognostication remain challenging. Roussel Uclaf Causality Assessment Method (RUCAM) is the best option for determining causality and has been increasingly accepted by clinicians. Revised Electronic Causality Assessment Method (RECAM) may be more user-friendly and accurate but requires further validation. Quantitative systems pharmacology methods, such as DILIsym, are increasingly used to predict hepatotoxicity. Oncotherapeutic agents represent many newly approved and described causes of DILI. Such hepatotoxicity is deemed acceptable relative to the benefit these drugs offer. Drugs developed for non-life-threatening disorders may not show a favorable benefit-to-risk ratio and will be more difficult to approve. As the COVID-19 landscape evolves, its effect on DILI deserves further investigation.

Clinical standards for the dosing and management of TB drugs

BACKGROUND: Optimal drug dosing is important to ensure adequate response to treatment, prevent development of drug resistance and reduce drug toxicity. The aim of these clinical standards is to provide guidance on 'best practice´ for dosing and management of TB drugs.METHODS: A panel of 57 global experts in the fields of microbiology, pharmacology and TB care were identified; 51 participated in a Delphi process. A 5-point Likert scale was used to score draft standards. The final document represents the broad consensus and was approved by all participants.RESULTS: Six clinical standards were defined: Standard 1, defining the most appropriate initial dose for TB treatment; Standard 2, identifying patients who may be at risk of sub-optimal drug exposure; Standard 3, identifying patients at risk of developing drug-related toxicity and how best to manage this risk; Standard 4, identifying patients who can benefit from therapeutic drug monitoring (TDM); Standard 5, highlighting education and counselling that should be provided to people initiating TB treatment; and Standard 6, providing essential education for healthcare professionals. In addition, consensus research priorities were identified.CONCLUSION: This is the first consensus-based Clinical Standards for the dosing and management of TB drugs to guide clinicians and programme managers in planning and implementation of locally appropriate measures for optimal person-centred treatment to improve patient care.

An Investigation of the Risk Factors Associated With Anti-Tuberculosis Drug-Induced Liver Injury or Abnormal Liver Functioning in 757 Patients With Pulmonary Tuberculosis

Objectives: To identify the risk factors associated with anti-tuberculosis drug-induced liver injury (AT-DILI) or abnormal living functioning from 757 patients with pulmonary tuberculosis (TB) registered at Nanshan Center for Chronic Disease Control (Nanshan CCDC), Shenzhen, Guangdong Province, China. Design and methods: We identified 757 TB patients who met our inclusion criteria by screening the Hospital Information System (HIS) at Nanshan CCDC. Next, we identified positive cases of AT-DILI or abnormal liver functioning based on results of the first-time liver function tests (LFTs) after taking anti-TB drugs. The χ2 test was used to relate the positive rate with a variety of factors. A logistic regression model was also used to identify statistically significant risk factors. Results: Of the 757 patients, the positive rate of AT-DILI or abnormal liver functioning was 37.9% (287/757). Univariate analysis revealed that the positive rate was 42.91% (212/494) for males and 28.52% (75/263) for females. The positive rate was significantly higher in males (p <0.001). Patients with an annual income of 9,231-13,845 USD had a significantly higher positive rate (67.35%; 33/49) than those with an income of 1,540-4616 USD (37.97%; 30/79) (p = 0.022). The most frequent prescription regime among positive cases was a 2 months supply of fixed dose combination Ethambutol Hydrochloride, Pyrazinamide, Rifampicin and Isoniazid Tablets (Ⅱ) 450 mg) followed by a 4 months supply of fixed dose combination Rifampin and Isoniazid Capsules (2FDC-HRZE half/4FDC-HR) at 56.03% (144/257). The least frequent prescription regime was a 2 months supply of fixed dose combination Rifampin, Isoniazid and Pyrazinamide Capsules with Ethambutol independently followed by a 4 months supply of fixed dose combination Rifampin and Isoniazid Capsules (2FDC-HRZ + EMB/4FDC-HR) at 24.27% (25/103). The difference between these two different regimes was significant (p = 0.022). With an increase in the duration of medication, patients under various prescription regimes all showed a gradual increase in the positive rate of AT-DILI or abnormal liver functioning. Conclusion: We identified several risk factors for the occurrence of AT-DILI or abnormal liver functioning, including gender, annual income, prescription regime, dosage, and treatment time.

Prediction of readmission in patients with acute exacerbation of chronic obstructive pulmonary disease within one year after treatment and discharge

Background

To investigate the risk factors and construct a logistic model and an extreme gradient boosting (XGBoost) model to compare the predictive performances for readmission in acute exacerbation of chronic obstructive pulmonary disease (AECOPD) patients within one year.

Methods

In total, 636 patients with AECOPD were recruited and divided into readmission group (n = 449) and non-readmission group (n = 187). Backward stepwise regression method was used to analyze the risk factors for readmission. Data were divided into training set and testing set at a ratio of 7:3. Variables with statistical significance were included in the logistic model and variables with P < 0.1 were included in the XGBoost model, and receiver operator characteristic (ROC) curves were plotted.

Results

Patients with acute exacerbations within the previous 1 year [odds ratio (OR) = 4.086, 95% confidence interval (CI) 2.723–6.133, P < 0.001), long-acting β agonist (LABA) application (OR = 4.550, 95% CI 1.587–13.042, P = 0.005), inhaled corticosteroids (ICS) application (OR = 0.227, 95% CI 0.076–0.672, P = 0.007), glutamic-pyruvic transaminase (ALT) level (OR = 0.985, 95% CI 0.971–0.999, P = 0.042), and total CAT score (OR = 1.091, 95% CI 1.048–1.136, P < 0.001) were associated with the risk of readmission. The AUC value of the logistic model was 0.743 (95% CI 0.692–0.795) in the training set and 0.699 (95% CI 0.617–0.780) in the testing set. The AUC value of XGBoost model was 0.814 (95% CI 0.812–0.815) in the training set and 0.722 (95% CI 0.720–0.725) in the testing set.

Conclusions

The XGBoost model showed a better predictive value in predicting the risk of readmission within one year in the AECOPD patients than the logistic regression model. The findings of our study might help identify patients with a high risk of readmission within one year and provide timely treatment to prevent the reoccurrence of AECOPD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-021-01692-3.