Incidence and risk factors of anti- tuberculosis drug induced liver injury (DILI): Large cohort study involving 4652 Chinese adult tuberculosis patients

Background and Aims: Anti- tuberculosis drugs remain as an important cause of drug-induced liver injury (DILI) worldwide. Adverse drug reactions reduce the effective ness of treatment. We aimed to determine the incidence and risk factors associated with anti- tuberculosis DILI (ATDILI). Methods: Using established criteria and causality assessment methods, risk factors for ATDILI were identified in a contemporary cohort and validated in another cohort prospectively. Independent determinants of ATDILI were identified using Cox regres sion analysis. Results: In the derivation cohort (n = 3155), 170 (5.4%) developed ATDILI of which 27 (15.9%) developed jaundice; 9(5.3%) developed acute liver failure (ALF) and 3 died. Among HBsAg positive patients, 11/27 (40.7%) of ATDILI developed after 3 months of starting treatment. In addition, of 218 (6.9%) who developed raised alanine trans ferase (ALT) levels ≥3 times upper limit normal, 193 (88.5%) resolved and 25 (11.4%) progressed to DILI. Age (HR = 1.014, 95% CI: 1.005- 1.023), baseline ALT (HR = 1.014, 95% CI: 1.003- 1.024), haemoglobin (HR = 1.011,


| INTRODUC TI ON
Anti-tuberculosis drug-induced liver injury (ATDILI) is still one of the most important adverse effects with a potential to lead to liver failure and death. 6,7 DILI is the most common adverse reactions resulting in change of treatment or treatment interruption 7 days or more 4 and therefore contributes to reduced effectiveness of treatment as well as non-compliance. 8 Anti-TB drugs are also one of the common drugs in combination to cause DILI and liver failure in low, middle, and high income countries. [9][10][11][12] The incidence of ATDILI varies widely between 2% and 28% dependent upon the characteristics of the particular cohort, drug regimens involved and in particular combination of liver enzymes and bilirubin thresholds used to define DILI. [9][10][11][12] Accurate and timely recognition of DILI is important to prevent serious consequences, however, drug-interruption or withdrawal would reduce its effectiveness. 4 In addition, although several risk factors have been reported to be associated with the development of ATDILI, [13][14][15][16]  In this context, the present study aims to investigate incidence and risk factors associated with ATDILI in a large, contemporary cohort of consecutive TB patients receiving first line intensive therapy based on international consensus case definitions of DILI.

| Patients
All consecutive patients diagnosed with active pulmonary or extrapulmonary TB and treated with first line anti-TB drug including (2018ZD039), Zhejiang Provincial Natural Science Foundation (LGF20H030006). The cooperation is to be established and a plaque will be installed -'Ningbo No.

| Data collection
The following parameters were collected for all the enrolled pa-

| Causality assessment
Before the diagnosis of DILI is confirmed, causality assessment was completed according to the Roussel Uclaf Causality Assessment Method (RUCAM). 18,19 In two groups, cases with scores equal or greater than 9 (categorised as "highly probable") and cases with scores equal or >6 (categorised as "probable") diagnosis of DILI were accepted. Cases with RUCAM scores between 3 and 6 were F I G U R E 1 Flow chart of patient selection. DILI, drug induced liver injury further reviewed by three hepatologists (Dr Yaoren Hu, Ting Hu and Huadong Yan) with expertise in DILI. Cases judged by at least two of the three hepatologists as "probable" were also enrolled in the study.

| Diagnostic criteria, clinical presentation, severity and management of ATDILI
In general, during the first two months of chemotherapy, patients underwent routine blood examination, and the liver function test was evaluated every 2 weeks. Subsequently, patients received routine monthly examination. Decision regarding treatment interruptions, withdrawals and reintroductions were individualized by the clinicians in charge TB care and although no protocol was established or guidelines were adopted to determine these, in general, anti-TB therapies were adjusted based on concomitant clinical symptoms.

| Statistical analysis
Continuous variables were expressed as mean standard deviation (SD) or median with interquartile range. Continuous data between the two groups were compared via Student's t test or Mann-Whitney U-test. Nominal variables were expressed as number/percentage and compared using Chi-square test. The risk factors of ATDILI were determined by univariate and multivariate cox regression analysis.
Candidate variables (P < .10) after a bivariate analysis were entered into a multivariate Cox regression analysis by a backward-forward approach. For multivariate analysis, the entry and removal probability for stepwise was set as .05 and .10, respectively, and variables with P < .05 were kept as risk factors of ATDILI. The cumulative incidence of ATDILI was calculated using Kaplan-Meier method to deal with the censored data. Statistical analysis were performed using SPSS16.0 (SPSS Inc.), GraphPad Prism 5 (GraphPad Software). A two-sided p value <.05 was set as the significant level.

| Patients
As shown in Figure 1, a total of 3155 patients with pulmonary or extra-pulmonary TB were enrolled into the derivation cohort from 5187 initially screened patients. The baseline characteristics of the study cohort are shown in Table 1. There were 2019 (64.0%) males and mean age of the patients was 37 ± 13 years.  (Table S1), 222 patients (7.0%) had previous history for TB treatment.

| Incidence and clinical characteristics of ATDILI patients at DILI onset
Of 3155 patients identified and followed for one year, 170 patients (5.4%) developed DILI after a median 42.9 days from start of treatment. Overall 64 (37.6%) of patients in the derivation cohort presented with one or more of the symptoms (Table S2)

| Risk factors for ATDILI
Results of univariate cox regression analysis of risk factors (at baseline) associated with DILI in study cohort are shown in Table 3.  Table 4.
Abbreviations: ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate Aminotransferase; BMI, body mass index; DILI, drug induced liver injury; DM, diabetes mellitus; GGT, glutamyl transpeptidase; RBC, red blood cell count; WBC, white blood cell.   Table 3. Age, baseline ALT, haemoglobin and HBsAg positivity were still strongly predictive of ATDILI in the validation cohort. Cumulative incidence of ATDILI in validation cohort patients with age ≥34 years, baseline ALT ≥16 IU/L, haemoglobin ≥134 g/L and HBsAg positivity were still significantly high than those with age <34 years, baseline ALT <16 IU/L, haemoglobin <134 g/L and

| Investigations of alternative causes
Two patients each (four in total) were excluded from the derivation and validation cohorts as the events were classified as HBV flare/reactivation by causality assessment with RUCAM score <3 (Figure 1).
Of 325 HBsAg positive patients in the derivation cohort 55

| Long-term outcome
Management of patients with raised liver enzymes were individualized by supervising physicians. In general, investigation to exclude explanations including HBV flare (as appropriate) were performed and liver biochemistry tests were monitored closely. Prompt withdrawal of anti-TB medication was observed in all with DILI, but, drugs and reintroduction regimen of re-introduction was determined by treating physicians. Key outcomes during a 2-year followup in patients with 3 × ULN of ALT as well as those who developed DILI have been presented in Table S5.  Note: Cox regression analysis was performed to identify factors associated with DILI.
liver test abnormalities during anti-TB therapy has long been recognized, however, it should be noted that the original study which included patients treated with INH in which such resolution or nonprogression of DILI was described reported fatality among 13/114 (12.3%) patients. 21 Using clear case definitions, we have been able to report the incidence and natural history of this phenomenon often referred to as 'adaptation', mechanisms underlying which are yet to be elucidated. 22 In addition, although we have adhered to ALT ≥5 × ULN as the threshold to identify DILI, ALF (0.2%, 10/4652) and mortality 0.09%, 4/4652) is still low compared to the large cohort of DILI because of all drugs from a USA-based multicentre study. 23 On the other hand, in a subset of patients with anti-TB DILI related ALF, spontaneous survival of 7/25 (28% [95% CI 12%-49%]) has been reported. 24 In our group, this was 6/9 (67% [95% CI 30%-93%] with a wide and overlapping confidence intervals mean that the difference in outcomes of two studies is not statistically significant and both groups are too small to compare. Moreover, only 18.8% in the former study 24 discontinued the medication before the onset of symptoms and another 14.3% stopped because of symptoms before presenting with jaundice. In our cohort, 84% stopped medication before developing jaundice which may explain the better outcome. Considering the wide use of this particular affordable regimen globally, anti-TB therapy remains an important contributor to the burden of DILI across the world. 11,24,25 Although there are number of studies that have investigated a variety of potential host factors that increase the susceptibility to anti-TB DILI, 22 the information is derived from heterogeneously defined and inconsistently characterised cohort; none has used large, longitudinal and independent cohorts to validate these. We have identified and validated age, ALT at the start of Anti-TB therapy, HBsAg, and high HB as independent risk factors for DILI occurrence. Age and baseline AST were also risk factors for severe DILI with jaundice and ALF, further strengthening the validity of our findings. Age cut-off of 34 years that defines the risk category in our study is in agreement with some studies, 26,27 although age cutoffs of 55 years or over have been proposed as risk factors in other studies previously. 18,19 Overall, even when age cut-offs derived from different cohort studies vary, relationship of age with increased risk of DILI appears to be consistent. In addition, although age specific prevalence rate ratio of TB globally (calculated using prevalence in [15][16][17][18][19][20][21][22][23][24] year age group as a baseline) increases with the age, demography of both testing and validating cohorts in this study with regards to age distribution ( Figure S2) is not dissimilar than expected from that of the population of china 2010-2020.
Baseline ALT and AST were independent risk factors for ATDILI and its severe manifestations respectively. In addition, the cut- cancers, cardiovascular and digestive system diseases. 28 Another longitudinal population based cohort study demonstrated that rise in ALT within the normal range (attributed to accumulation of liver fat) can predict incident diabetes in the long-term. 29 Although fatty liver has been associated with increased risk of DILI, in the context of tuberculosis patients, we do not think that is the likely explanation for our findings. 30 As baseline ALT and AST were identified as risk factors independent of HBV, alcohol intake and unrelated to BMI (which is not associated with DILI), we don't think these reflect undiagnosed underlying liver disorder. Active drinking was a risk factor for the development of jaundice, but, was not associated with the susceptibility to ATDILI itself. As Kupffer cells, resident macrophages of the liver are involved in the uptake and clearance of enzymes such as ALT and AST, levels of latter enzymes may be a reflection of Kupfer cell numbers. 31 We speculate that our findings of association of baseline ALT within the normal range being asso- Interestingly, high baseline haemoglobin level was one of the risk factors for DILI, replicated in an independent validation cohort.
In a population based prospective study based in Ningbo involving >7400 NAFLD free subjects, those with higher baseline haemoglobin had higher incidence of NAFLD over 3 years follow-up. 33 Association of Hb was in the normal range and independent of body mass index, Type 2 diabetes and other metabolic liver diseases. It has been hypothesized that free haemoglobin induced oxidative damage contributed to the development of NAFLD. 34 Haptoglobin, one of the acute phase protein binds to free plasma haemoglobin to reduce oxidative stress induced by free haemoglobin. Therefore, we speculate that the positive association of DILI with baseline Hb may reflect lower degree of physiological defence against oxidative stress related to the drugs. We did not estimate haptoglobin levels in either of the cohorts, hence, are not able to assess this any further.
The current study has a few limitations. Firstly, the study comes from a specialist centre for infectious diseases which while ensuring homogeneity and consistency in clinical practice, might reflect specific case-mix and severity of TB that may not be reflective of patients in the general population. Secondly, we chose a large retrospective cohort to develop the multivariate model; this inevitably introduces some inconsistencies in data acquisition. Even though we identified age, baseline AST, previous TB treatment and active drinking as risk factors for developing jaundice, because of a small number

F I G U R E 2
The cumulative incidence of DILI in subgroups. It was compared using the log-rank test. DILI, drug induced liver injury of life threatening complications, we are not able to develop multivariate models to predict ALF or death related to DILI. In addition, 15/170 (9.2%) of patients with DILI were lost to follow-up (Table S5).
Observational nature of the study precludes investigations of mechanisms underlying the how risk factors influence the incidence of DILI, and its serious outcomes. We have not evaluated the genetic factors that may have particular influence on the risk of DILI in this ethnic group; therefore, these findings may not be applicable to another ethnic group. Finally, although people co-infected with human immunodeficiency virus (HIV) were not excluded, overall there were only five patients with HIV in two cohorts limiting the evaluation of the role of co-infection as a risk factor for DILI. Our evaluation of concomitant drugs was limited to information provided by the patient as only prescription drugs were automatically recorded in the database. In addition, therapeutic regimen involving combination of anti-TB drugs does not permit assessment of individual agent's causality in ATDILI.
In summary, first-line anti-tuberculosis DILI and related death are lower than that has been reported in these large cohorts of patients.
We have demonstrated that 88.5% of those who develop raised ALT while on ATB therapy resolve without ≥7 day drug interruption and therefore are able to complete the crucial treatment. Age, baseline ALT, haemoglobin and HBsAg positivity are risk factors for the development of DILI. Patients with positive HBsAg develop DILI later, which means that monitoring should be adjusted to suit this particular population. Considering the fact that baseline ALT and haemoglobin level cut-offs in the normal range determines the risk, it is unlikely that any change in monitoring is necessary for this group.
Focus on genetic factors has the potential to improve the efficacy and safety of anti-TB therapy further. 4

CO N FLI C T O F I NTE R E S T
The authors who have taken part in this study declared that they do not have anything to disclose regarding funding or conflict of interest related to this manuscript.

AUTH O R CO NTR I B UTI O N S
HY and GPA conceived the study; TC, HY and FJ designed the study;