Data were obtained from ADACS, which was approved by the Ethics Committee of Center for Tuberculosis Control and Prevention of China and carried out from October 2007 to March 2009[¹³,¹⁴]. And written informed consent was obtained from every participant or surrogate before enrolment. ADACS was a prospective longitudinal study of anti-TB drugs induced adverse effects, which consisted of a multi-center cohort of 4488 pulmonary TB patients and receiving standard short-course chemotherapy (primary and re-treatment patients would take HRZE/HRZES for initial intensive phase and HR/HRE for consolidation phase respectively) in four geographically and economically diverse areas of China[¹³,¹⁴]. During the study, patients were asked to take liver function test at baseline and within 2 months after treatment initiation in addition to developing unbearable symptoms [¹³,¹⁴], and for patients with abnormal baseline liver function the anti-TB treatment would not be started until their liver function turned normal through liver protective treatment, so the baseline liver function was normal for all patients. After the 6–9 months follow up period, 273 patients appeared liver dysfunction and 106 patients were diagnosed as ATLI according to the ATS criteria [¹⁴]. Of the 273 patients with liver dysfunction, 111 patients who were diagnosed through scheduled liver function test within two months after initiation of anti-TB treatment formed scheduled monitoring group, others who were diagnosed due to developing symptoms formed passive detection group (n = 162) (Figure 1). In this study, we compared the clinical features, prognosis of liver dysfunction and impact on anti-TB treatment between the two groups to determine the effect of liver function monitoring on identification of asymptomatic liver damage and anti-TB treatment outcomes during anti-TB treatment. Besides, all patients were required to fill in their ADACS calendars and record their everyday feelings as well as drug usages during the whole follow up period with the same pattern of treatment supervision [¹³].

Liver dysfunction was diagnosed as an increase in serum alanine aminotransaminase (ALT) or aspartate transaminase (AST) or total bilirubin (TBil) that was greater than the upper limit of normal (ULN), when other causes were excluded: new viral hepatitis infections, other liver diseases, other potentially hepatotoxic medications that would confound the picture. All assessments were double checked by two experienced physicians. The severity of liver dysfunction was classified according to the WHO Toxicity Classification Standards [¹²]: mild (ALT/AST ≤ 3 times ULN, or TBil ≤ 2 times ULN), moderate (3 times ULN < ALT/AST ≤ 5 times ULN, or 2 times ULN < TBil ≤ 5 times ULN) and severe (ALT/AST/TBil > 5 times ULN).

During the follow-up in ADACS, when liver dysfunction developed, patients would be given intensive monitoring and/or treatment (such as liver protective drugs or even hospitalization) according to the doctor’s clinical adjudication. The anti-TB treatment would not be changed immediately, only when ALT/AST/TBil elevated to greater than 2 times ULN, changes of anti-TB treatment would be considered by the doctor, including drug replacement, interruption, discontinuation, dose decrease and changes of medication administration.

For comparison of the two groups, the following parameters were noted: 1) the time period for the development of liver dysfunction from initiation of anti-TB treatment (liver dysfunction finding time); 2) requirement of hospitalization; 3) severity of liver dysfunction; 4) prognosis of liver dysfunction, which was defined and classified as recovered (ALT/AST/TBil turned normal), continuing with abnormal ALT/AST/TBil and death; 5) impact on anti-TB treatment.

The baseline characteristics and clinical features of participants were described as median (inter-quartile range, IQR) for continuous variables (not subject to normal distribution), and percentages for categorical variables. Non-parameter test was used for continuous variables (not subject to normal distribution) and χ2 test was used for categorical variables.

In order to control the confounding bias in this observational study, propensity score weighting was used to compare the prognosis of liver dysfunction and impact on anti-TB treatment between participants in scheduled monitoring group and passive detection group. A propensity score model was developed with logistic regression model to predict patients’ assignment in this cohort. Covariate variables included in the model were as follows according to the selection principle [¹⁵]: age, gender, weight, marriage status, income per year, type of anti-TB treatment, history of anti-TB adverse reactions, history of liver diseases, HBSAg, pre-existing disease, pattern of taking anti-TB drugs (every day or every other day), and type of preventive liver protective drugs [¹²,¹⁶]. Liver protective drugs were classified as four categories (herbal preparations, finished manufactured herbal products, combinations of vitamins and other non-herbal substances, and pharmaceutical preparations) according to a modified version of WHO standard definitions for the evaluation and research of herbal medicines [¹⁷,¹⁸]. After the calculation of propensity score (PS), standardized mortality ratio weighting (SMRW) was used to adjust the data with formulas as follows:

Then χ2 test was used to compare the prognosis of liver dysfunction and impact on anti-TB treatment between the two groups.

The statistical analyses were performed using SPSS for Windows (version 13.0; SPSS Inc.). A two-sided P value less than 0.05 was set as the significant level.

Of the 273 patients with liver dysfunction, 111 patients were in scheduled monitoring group while 162 patients were in passive detection group. The baseline characteristics of 273 patients were shown in Table 1. The median age of 273 patients was 40.0 years (IQR: 28.0-53.0), and 212 (77.7%) were male. Among the demographic features and weight, there was no significant difference between the two study groups. And no statistically significant difference was found in type of anti-TB treatment, history of anti-TB adverse reactions, history of liver diseases, HBSAg and pattern of taking anti-TB drugs between the two groups. 75 patients (67.6%) in scheduled monitoring group and 94 (58.0%) in passive detection group took preventive liver protective drugs during the anti-TB treatment (P = 0.111). 57 (51.4%), 3 (2.7%), 5 (4.5%) and 10 (9.0%) patients in scheduled monitoring group took finished manufactured herbal preparations, combinations of vitamins and other non-herbal substances, pharmaceutical preparations and combinations of above liver protective drugs, respectively. On the contrary, 59 (36.4%), 5 (3.1%), 7 (4.3%) and 23 (14.2%) patients in passive detection group were prescribed four categories of liver protective drugs above, respectively.

91 (33.3%) of 273 patients did not have any clinical symptoms, including 8 with severe hepatotoxicity. Among 111 patients in scheduled monitoring group, only 20 patients (18.0%) developed symptoms of nausea (11 patients, 55%), vomiting (9 patients, 45%), anorexia (5 patients, 25%) and abdominal symptoms (4 patients, 20%) after detecting liver dysfunction. None of the patients developed icterus and dark urine. All 162 patients (100%) in passive detection group were symptomatic. 116 (71.6%), 110 (67.9%), 31 (19.2%) and 51 (31.5%) patients experienced nausea, vomiting, anorexia and abdominal symptoms, respectively. 5 patients (3.1%) developed icterus and 1 patient (0.6%) developed dark urine (Table 2).

The median interval in days between the initiation of anti-TB treatment and the development of liver dysfunction was 52 (IQR: 30, 61) for scheduled monitoring group and 39 (IQR: 28, 60) for passive detection group. The difference between patients in the two groups was not statistically significant (P = 0.205) (Table 2).

In terms of the severity of liver dysfunction, more patients (28 patients, 17.3%) in passive detection group experienced severe liver dysfunction in comparison with those in scheduled monitoring group (13 patients, 11.7%), although there was no statistically significant difference between the two groups (P = 0.443). Besides, 8 patients with severe hepatotoxicity in scheduled monitoring group did not have any clinical symptoms. Only 2 patients (1.8%) in scheduled monitoring group required hospitalization, whereas in passive detection group a total of 18 patients (11.1%) were hospitalized. The difference between the two groups was statistically significant (P = 0.004) (Table 2). When analyzing patients with liver dysfunction finding time less than 1 month, no one in scheduled monitoring group required hospitalization in contrast to 8 patients (11.9%) in passive detection group (P = 0.025).

Regarding the prognosis of liver dysfunction, 1 patient in passive detection group was excluded from the analysis due to death of lung cancer. 109 patients (98.2%) recovered, 2 patients (1.8%) continued with abnormal transaminase, and none of the patients died in scheduled monitoring group, whereas in passive detection group 158 patients (98.1%) recovered, 1 patient (0.6%) continued with abnormal transaminase, and 2 patients (1.3%) died of hepatotoxicity. A 68 years old woman who experienced ALT/AST elevation to 693.1/510.2 U/L after 35 days of anti-TB treatment was with no improvement after hospitalization and finally died of hepatic failure on the 45th day; another 69 years old man who experienced ALT elevation to 210 U/L with jaundice after 65 days of anti-TB treatment died after 1 week in hospital. The propensity score weighting analysis showed that there was not significantly different between the two groups (P = 0.965).

In scheduled monitoring group, 39 patients (35.1%) changed their anti-TB treatment, whereas 92 patients (56.8%) in passive detection group changed the treatment. The difference was statistically significant by propensity score weighting analysis (P = 0.001). As shown in Table 3, more patients in passive detection group interrupted and discontinued the anti-TB treatment in comparison with patients in scheduled monitoring group (P = 0.007 and P = 0.026, respectively).

When analyzing the impact of liver dysfunction on TB condition, 4 patients in passive detection group were excluded because they could not be followed up for TB treatment outcomes due to discontinuation after the onset of hepatotoxicity (3 patients) and death of lung cancer (1 patient). 89 patients (80.2%) in scheduled monitoring group and 112 patients (70.9%) in passive detection group had TB cured on time. None of the patients had TB aggravated or died in scheduled monitoring group, whereas there was 1 (0.6%) with TB aggravation and 2 (1.3%) deaths in passive detection group (Table 4).

Note: ^*Including hepatitis B, icterohepatitis and cholelithiasis.

Note: ^* 4 patients were excluded due to lost to follow up after discontinuation and death of lung cancer.

^# P value was calculated after propensity score weighting analysis.