Outcomes and Predictors of Mortality in Patients With Drug-Induced Liver Injury at a Tertiary Hospital in South India: A Single-Centre Experience

Abstract

Introduction

Drug-induced liver injury (DILI) is an important cause of acute liver failure with significant morbidity and mortality. The outcome of DILI varies widely according to the drug implicated and the type of liver injury. Owing to the heterogeneous nature of liver injury, knowledge on clinical course and prognosis of DILI is limited. We had undertaken this study to determine the clinical characteristics, outcomes, and predictors of mortality in patients with DILI.

Materials and methods

This prospective study was conducted from January 2015 through December 2018. We analyzed the drugs implicated, clinical course, and the outcome. Causality assessment was performed by using Roussel Uclaf Causality Assessment Method scoring. Patients were followed for 6 months until recovery/death or liver transplantation.

Results

There were 133 cases with DILI. The mean age was 47.6 years, and 51.9% of them were men. Drugs causing DILI were antitubercular drugs (37.5%) followed by neuropsychiatric drugs (16.5%), antibiotics/antifungals (12%), complementary and alternative medicine (10.5%), immunomodulatory/chemotherapeutic drugs (10.5%), and nonsteroidal antiinflammatory drugs (7.5%). Eighty-two (61.6%) patients were classified as hepatocellular, 30 (22.5%) as mixed and 21 (15.7%) as cholestatic type of injury. There was no significant difference in the mortality and morbidity between the three types of liver injury. There were 18 deaths (13.5%), of which antitubercular drugs constituted the majority (55.5%) followed by neuropsychiatric drugs (27.7%) and complementary and alternative medicine (16.6%). Based on receiver operating characteristic curve analysis, model for end-stage liver disease (MELD) score >28, mean international normalized ratio (INR) >1.97, mean bilirubin level >15.6 mg/dl, and creatinine level >1.35 mg/dl were associated with mortality.

Conclusion

Although DILI is uncommon, it has significant morbidity and mortality. Antitubercular drugs were the most common cause for DILI and DILI-related mortality in our study. Variables such as MELD, INR, bilirubin, albumin, and creatinine help in predicting severity of liver injury and may help in triaging the patient for referral for liver transplantation.

Diagnosis of drug-induced liver injury (DILI) is uncommon and often challenging. DILI has significant morbidity and mortality.^1,2 The recognition of DILI is often delayed due to the need to exclude more common causes of liver injury. In most instances, it is unrecognized or underreported.^{3, 4, 5} The annual incidence of DILI ranges from 14 to 19 per 100,000 populations, and the mortality ranges between 10% and 17.3%.^{2,5, 6, 7} The spectrum of liver injury in DILI ranges from asymptomatic elevation of liver enzymes to acute liver failure (ALF). Major drugs causing DILI include antibiotics, antiepileptics, nonsteroidal antiinflammatory drugs (NSAIDs), and complementary and alternative medicines (CAMs), and they vary geographically. In India, antitubercular therapy (ATT) is the most common cause of DILI and drug-induced ALF leading to death.^{6, 7, 8, 9} In a large series which studied ALF, ATT contributed to 5.7% of patients with ALF, with 67% mortality.⁸ In India, 11 of 12 deaths due to DILI were associated with ATT.^{6,8, 9, 10} Over-the-counter drugs and CAM-related liver injury are also not uncommon in this part of India, and they mask identification of individual drug leading to DILI.¹¹

Each drug has a signature pattern of liver injury and prognosis varies accordingly. Apart from the type of drugs, the age, bilirubin, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and prothrombin time (PT) are evaluated as predictors of mortality.^6,12,13 However, the critical level of bilirubin, PT, or other factors that predict severity or mortality in patients with DILI are still lacking. This study was conducted to determine causes, clinical characteristics, and outcomes of patients with DILI admitted in our hospital. The clinical and laboratory parameters associated with mortality in these patients were also studied.

Materials and methods

This prospective observational cohort study was conducted in the Department of Gastroenterology, Govt. Medical College, Kozhikode, India, from January 2015 through December 2018. This study was approved by the Institutional Ethics Committee. All patients with suspected DILI were enrolled in the study after exclusion of other causes of acute liver injury. The criteria for DILI were as per the International DILI Expert Working Group.¹⁴ Drug causality assessment was carried out by applying the Roussel Uclaf Causality Assessment Method (RUCAM) model.^15,16 The diagnosis of DILI was confirmed on the basis of documented exposure to drug ingestion leading to hepatotoxicity. The following criteria were used for the diagnosis of DILI. An increase in serum ALT or AST levels >5 times ULN (Upper limit of normal) in patients without symptoms or an elevation of alkaline phosphatase > 2 times ULN or elevation of serum bilirubin >2 mg/dl, or increase in AST or ALT >3 times ULN in patients with symptoms.¹⁶ Other common causes of acute liver injury were excluded by laboratory investigations.Drug causality relationship was classified as highly probable, probable, possible, unlikely, or excluded.¹⁶ Asymptomatic elevation in liver enzymes (<5 times ULN) was not included in the study population, and they were followed up to assess the progression. These patients were included in the study only if they fulfilled criteria for inclusion subsequently. The ULN in our laboratory for ALT and AST was 40 IU/l.

Clinical and laboratory variables studied included the age, gender, body mass index (BMI), symptoms, duration of drug exposure before development of DILI, intake of concomitant drugs, associated comorbidities, liver function tests, complete hemogram, international normalized ratio (INR), creatinine, random blood sugar at the time of diagnosis and then periodically until recovery or for a period 6 months from the time of diagnosis. In selected cases, where the diagnosis of DILI was clinically questionable, additional laboratory tests were performed on a case-to-case basis to exclude other infectious causes of hepatitis and Wilson disease. Ultrasound abdomen was performed in all patients, and MRCP performed in selected cases with cholestasis to rule out bile duct obstruction. Men consuming >40 g/day and women consuming >20 g/day of alcohol on a regular basis were excluded from the study. Patients with evidence of prior chronic liver disease were also excluded. The model for end-stage liver disease (MELD) score was calculated as per standard formula from the highest values of bilirubin, INR, and creatinine.¹⁷ Although the MELD score is traditionally used to determine risk of mortality in patients with cirrhosis, the utility and even superiority has been described in patients with ALF.^{18, 19, 20} The pattern of liver injury is defined by calculating the R value.^15,16 The R ratio applied was calculated based on the initial liver tests at presentation.

Severity of liver injury was classified into mild, moderately severe, and severe. Patients with elevated transaminase (>5ULN) or alkaline phosphatase (>2ULN) level without hyperbilirubinemia were included as mild cases, presence of bilirubin value of >2 mg/dl were included as moderately severe cases, and presence of prolonged INR (>1.5), encephalopathy, or ascites with or without hospitalization accompanied by hyperbilirubinemia were included as severe DILI cases.¹⁸ Hy's law which predicts mortality (bilirubin >3 mg/dl and AST or ALT >3 times ULN) was also calculated.²¹ Persistent DILI is defined as evidence of continued liver injury beyond 3 months for hepatocellular and mixed type of DILI and beyond 6 months for cholestatic DILI after stopping the drug.¹⁴ All patients received N-acetyl cysteine in addition to cessation of the implicated drug. As patients were treated by their physicians with various drugs and various doses of medicines before inclusion in our study, the effects of different forms of treatment were not analyzed for the outcome. Patients with ALF received N-acetyl cysteine, antiencephalopathy measures, and other supportive care. Patients with drug reaction with eosinophilia and systemic symptoms (DRESS) received corticosteroids.

Statistical analysis

The statistical results are presented as mean and standard deviation (SD) for continuous variables and as frequencies and percentages for categorical variables. Continuous variables were compared parametrically using the Student's t test or nonparametrically using the Mann-Whitney test. Categorical variables were compared using the chi-square test. Receiver operating characteristic (ROC) curve analysis was performed for significant variables, and optimum cutoff was calculated. Area under curve (AUC) and its 95% confidence interval were used to present the results. Based on the ROC analysis, sensitivity and specificity for variables were calculated. The level of significance was set at 0.05. The SPSS software, version 22, package for Windows was used for statistical analysis.

Results

We studied 133 cases of DILI over a period of 4 years. Table 1 shows the demographic and clinical characteristics. Sixty-nine patients were men and 64 were women. Based on R factor, 82 (61.6%) patients were classified as hepatocellular, 30 (22.5%) as mixed and 21 (15.7%) as cholestatic type of DILI. The mean age of patients was 44.5 ± 15.3, 45.6 ± 16.2, and 51.1 ± 18.9 years in patients with hepatocellular, mixed, and cholestatic injury, respectively. Men were predominant in cholestatic injury (n = 14, 66.6%). The most common symptoms were anorexia (35.1%), jaundice (30.6%), vomiting (20.9%), pruritus (17.2%), and abdominal pain (11.9%). Except for pruritus which was common in the cholestatic group, all symptoms were almost similar in all the three groups. DRESS was seen in 13 (9.7%) patients. The mean BMI was 21.12 and was nearly similar in all groups. The mean duration of drug exposure was slightly higher in the mixed injury group (74.4 ± 84 days) than in the hepatocellular or cholestatic injury groups (50.3 ± 54 and 50.8 ± 63 days, respectively). On RUCAM scoring, 21 (15.7%) patients were classified as possible, 67 (50.4%) as probable, and 45 (33.8%) as highly probable.

Table 1.

Characteristics of Patients With DILI Based on the Type of Liver Damage.

Patient characteristics	Total n = 133	Hepatocellular n = 82	Mixed n = 30	Cholestatic n = 21	P value
Age (in years)	45.86 ± 16.2	44.5 ± 15.3	45.6 ± 16.2	51.1 ± 18.9	0.24
Gender (male/female)	69/64	41/41	14/16	14/7	0.32
BMI (kg/m2)	21.21 ± 3.86	21.3 ± 3.7	20.7 ± 3.9	21.5 ± 4.3	0.69
Mean duration of drug exposure (in days)	52.56 ± 63.5	50.3 ± 54	74.4 ± 84	50.8 ± 63	0.21
Anorexia	47 (35.1%)	32 (39%)	9 (30%)	6 (28.6%)	0.41
Vomiting	28 (20.9%)	19 (23.2%)	6 (20%)	3 (14.3%)	0.36
Jaundice	41 (30.6%)	22 (26.8%)	11 (36.7%)	8 (38.1%)	0.33
Abdominal pain	16 (11.9%)	11 (13.4%)	2 (6.7%)	3 (14.3%)	0.48
Pruritus	23 (17.2%)	14 (17.1%)	4 (13.3%)	5 (23.8%)	0.80
Hb (gm/dl)	12.2 ± 1.75	12.3 ± 1.6	12.2 ± 2.1	11.9 ± 1.6	0.64
WBC (103/dl)	9228.9 ± 3703	9044.2 ± 3111	9155.0 ± 3343	10056.0 ± 5857	0.53
Platelet (105/dl)	2.53 ± 0.96	2.4 ± 0.8	2.8 ± 1.1	2.6 ± 0.9	0.15
RBS (mg/dl)	121.1 ± 36.8	117.9 ± 34.9	123 ± 44.2	130 ± 32.4	0.38
Creatinine (mg/dl)	0.9 ± 0.44	0.89 ± 0.44	0.94 ± 0.47	0.85 ± 0.42	0.76
Mean INR	1.54 ± 1.27	1.65 ± 1.4	1.36 ± 0.9	1.35 ± 1.1	0.42
Peak INR	1.74 ± 1.79	1.82 ± 1.5	1.58 ± 1.9	1.63 ± 2.3	0.79
Mean bilirubin (mg/dl)	6.49 ± 6.5	6.2 ± 6.3	7 ± 7.3	6.8 ± 6.6	0.81
Peak bilirubin (mg/dl)	8.54 ± 9.0	8.8 ± 9.4	7.4 ± 8.5	8.9 ± 8.1	0.76
Serum protein (gm/dl)	6.69 ± 0.68	6.7 ± 0.6	6.5 ± 0.7	6.7 ± 0.7	0.54
Serum albumin (gm/dl)	3.5 ± 0.57	3.53 ± 0.5	3.55 ± 0.5	3.33 ± 0.6	0.33
Mean ALT (IU/L)	371.8 ± 448	491.3 ± 525	234.2 ± 149	101.5 ± 105	0.00
Mean ALP (IU/L)	185.7 ± 122	137.5 ± 80	243.1 ± 157	291 ± 118	0.00
50% reduction in ALT (in days)	15.63 ± 11.9	14.7 ± 11	17.4 ± 15	16.7 ± 12	0.64
Normalization of LFT (in days)	24.3 ± 23.8	32.4 ± 24	31.8 ± 19	26.2 ± 18	0.64
MELD	14.42 ± 8.6	14.3 ± 8.9	14.6 ± 7.9	14.2 ± 8.9	0.98
DRESS	13 (9.78%)	8.5%	10%	14.3%	0.73
RUCAM (probable/possible/highly probable)	67/21/45	36/14/32	17/4/9	14/3/4	0.35
Severity (mild/moderately severe/severe)	55/36/42	30/21/31	16/9/5	9/6/6	0.30
Mortality	18(13.5%)	12(9.02%)	3(2.2%)	3(2.2%)	0.77

BMI: body mass index; MELD: model for end-stage liver disease; DRESS: drug reaction with eosinophilia and systemic symptoms; Hb: hemoglobin; WBC: white blood cell; RBS: random blood sugar; INR: international normalized ratio; ALT: alanine transaminases; ALP: alkaline phosphatase; RUCAM: Roussel Uclaf Causality Assessment Method; DILI: drug-induced liver injury; LFT, Liver function test.

Drugs causing DILI and type of damage

The different drugs associated with DILI are shown in table 2. ATT drugs constituted the majority (n = 50, 37.5%), followed by neuropsychiatric drugs (n = 22, 16.5%), antibiotics or antifungals (n = 16, 12%), CAM (n = 14, 10.5%), immunomodulatory or chemotherapeutic drugs (n = 14, 10.5%), NSAIDs (n = 7, 5.2%), and miscellaneous drugs (n = 10, 7.5%). The mean duration for reduction in liver enzymes by 50% (ALT) was 14.7, 17.4, and 16.7 days in the hepatocellular, mixed, and cholestatic group, respectively. Persistent DILI was seen in 13 patients, and the drugs implicated were ATT drugs (n = 6), CAMs (n = 2), and dapsone, carbimazole, methotrexate, azathioprine, and telmisartan constituting one each. Among this group, 6 patients had cholestatic injury, 2 mixed, and 5 had hepatocellular type of injury.

Table 2.

Drugs Associated With DILI and Mortality.

Drug	Number	RUCAM (probable/possible/highly probable)	Mortality
ATT drugs	50 (37.5%)	23/3/24	10 (20%)
CAM related	14 (10.5%)	4/4/6	3 (21.4%)
Neuropsychiatric drugs phenytoin (5), carbamazepine (4), valproate (4), lamotrigine (2), chlorpromazine (2), olanzapine (2), clozapine (1), mirtazapine (1), amitriptyline (1)	22 (16.5%)	12/5/5	4 (18.1%)
NSAIDs diclofenac (3), flupirtine (1), ketorolac (1), paracetamol (1), mefenamic acid (1)	7 (5.2%)	4/1/2	0
Antibiotic/antifungals dapsone (4), fluconazole (3), amoxicillin-clavulanate (3), trimethoprim-sulphamethoxozole (2), azithromycin (2), itraconazole (2)	16 (12%)	10/2/4	0
Immunomodulator and chemotherapeutic drugs methotrexate (4), azathioprine (3), cyclophosphamide (3), rituximab (2), gemcitabine (2)	14 (10.5%)	7/5/2	0
Miscellaneous sulphasalazine (4), allopurinol (1), colchicine (1), neomercazole (1), ramipril (1), rosuvastatin (1), telmisartan (1)	10 (7.5%)	7/2/1	1 (10%)

ATT: antitubercular therapy; CAMs: complementary and alternative medicines; NSAIDs: nonsteroidal antiinflammatory drugs; RUCAM: Roussel Uclaf Causality Assessment Method; DILI: drug-induced liver injury.

Among the patients with hepatocellular pattern of injury, the ATT drug (n = 34, 41.5%) was the most common drug implicated, followed by CAM (n = 13, 15.9%), neuropsychiatry drugs (n = 10, 12.2%), and immunomodulatory drugs (n = 8, 9.8%). In the mixed pattern group also, the ATT drug (n = 9, 30%) was the most common drug, followed by neuropsychiatry drugs (n = 7, 23.3%) and antibiotics (n = 5, 16.7%). In patients with cholestatic pattern also, the ATT drug (n = 7, 33.3%) constituted the majority followed by neuropsychiatry drugs (n = 5, 23.8%) and antibiotics (n = 5, 23.8%). Liver biopsy was performed in two cases because of diagnostic dilemma, and both patients showed hepatocellular pattern of injury with acute hepatitis picture and confluent necrosis. Owing to nonavailability of liver transplantation facility in our institution and due to financial constraints, only one patient underwent liver transplantation from outside facility which was successful. This patient had DILI secondary to CAM.

Factors affecting mortality

Of 133 patients, 42 patients (32%) had severe liver injury of which 18 patients died, constituting a mortality of 13.5%. The drugs implicated in mortality were ATT drugs (10; 55.5%), CAMs (3; 16.6%), carbamazepine (2; 11%), valproate, chlorpromazine, and amitriptyline in one each (5.5% each). The mortality rate was 15.4%, 10.7%, and 17.6% among the hepatocellular, mixed, and cholestatic group, respectively. Comparison among survivors and nonsurvivors in patients with DILI is shown in Table 3. Age was slightly higher (53.1 years) among nonsurvivors than among survivors (44.2 years) but was statistically not significant. BMI and R factor had no significant difference between groups. Higher MELD score, INR, higher serum creatinine, serum bilirubin, and low serum albumin levels were significantly associated with higher mortality. The ROC curve was plotted for bilirubin, creatinine, INR, and MELD score, and it showed an optimum AUC as shown in Figure 1. Based on the optimal cutoff obtained in ROC analysis, MELD score of >28, serum creatinine of >1.35 mg/dl, INR of >1.97, and serum bilirubin of >15.6 mg/dl had a specificity of more than 95% for mortality. A MELD score of 16.5 had a sensitivity of 94.4% with specificity of 73.3%. Although creatinine of 1.35 mg/dl had 100% specificity, its sensitivity was less than 50%. Creatinine was included in the ROC model (P = 0.075) as it addresses multiorgan involvement. In patients with severe DILI, symptoms such as jaundice, vomiting, and anorexia did not have any predictive value for mortality. Presence of ascites (64.7% vs 24%), longer duration of drug exposure (96.7 vs 46.6 days), higher MELD score (27.9 vs 19.6), higher creatinine level (1.3 ± 0.9 vs 0.8 ± 0.2 mg/dl), prolonged INR (3.6 vs 1.7), and elevated bilirubin (15.7 vs 11.8 mg/dl) were independent predictors of mortality (Table 4).

Table 3.

Comparison Between Survivors and Nonsurvivors With DILI.

Patient characteristics	Survivors n = 102	Nonsurvivors n = 18	P value
Age (in years)	44.2 ± 16.2	53.1 ± 14.4	0.91
Gender (male/female)	54/48	9/9	0.86
BMI (kg/m2)	21.2 ± 3.8	20.7 ± 3.5	0.82
Ascites	5 (4.9%)	7 (38%)	0.001
Hepatomegaly	15/90	5/13	0.15
Jaundice	26 (25.5%)	9 (50%)	0.035
Vomiting	20 (19.6%)	4 (22.2%)	0.798
Anorexia	35 (34.3%)	8 (44.4%)	0.409
Abdominal pain	9 (8.8%)	4 (22.2%)	0.092
Pruritus	17 (16.67%)	2 (11.1%)	0.552
DRESS	13 (12.7%)	0	0.109
Hb (gm/dl)	12.5 ± 1.5	11.1 ± 1.7	0.05
WBC (103/dl)	8927.2 ± 3244	10679.0 ± 5804	0.02
Platelet (105/dl)	2.6 ± 0.9	1.8 ± 0.6	0.01
RBS (gm/dl)	122 ± 39	124 ± 27	0.82
Creatinine (gm/dl)	0.81 ± 0.1	1.4 ± 0.9	0.001
Mean INR	1.2 ± 0.4	3.5 ± 2.2	0.001
Serum protein (gm/dl)	6.7 ± 0.6	6.4 ± 0.5	0.01
Serum albumin (gm/dl)	3.6 ± 0.5	3.0 ± 0.4	0.001
Mean bilirubin (mg/dl)	4.6 ± 3.2	16.1 ± 5	0.001
Mean ALT (IU/L)	310.6 ± 375	686.9 ± 626	0.003
Mean ALP (IU/L)	185.9 ± 128	197.0 ± 92	0.48
MELD	12 ± 6.1	27.4 ± 10	0.001
Mean duration of drug exposure(in days)	43.2 ± 59.8	94.6 ± 60.9	0.001
R factor (hepatocellular/mixed/cholestatic)	63/25/14	12/3/3	0.67
RUCAM (probable/possible/highly probable)	52/14/39	10/4/4	0.38
Severity (mild/moderate/severe)	51/31/20	0/1/17	0.0001

BMI: body mass index; MELD: model for end-stage liver disease; Hb: hemoglobin; WBC: white blood cell; RBS: random blood sugar; INR: international normalized ratio; ALT: alanine transaminase; ALP: alkaline phosphatase; RUCAM: Roussel Uclaf Causality Assessment Method; DILI: drug-induced liver injury.

Figure 1.

ROC curve for predicting mortality in patients with DILI (MELD > 28, creatinine > 1.35 mg/dl, mean INR > 1.975, mean bilirubin > 15.65 mg/dl). ROC: receiver operating characteristic; DILI: drug-induced liver injury; MELD: model for end-stage liver disease; INR: international normalized ratio.

Table 4.

Significant Parameters Associated With Mortality in Patients With DILI on Regression Model of Univariate Analysis.

Test result variables	Positive: greater than or equal to	Sensitivity of prediction model	Specificity of prediction model
MELD	16.5	94.4%	73.3%
	28	55.6%	99%
Creatinine (mg/dl)	1.35	44.4%	100%
Mean INR	1.375	94.4%	80%
	1.975	83.3%	96.2%
Mean bilirubin (mg/dl)	7.65	100%	75.2%
	15.65	66.7%	95.2%
Albumin (gm/dl)	3.3	74.3%	88.9%

MELD: model for end-stage liver disease; DILI: drug-induced liver injury; INR: international normalized ratio.

Comparison between ATT and non-ATT drugs

ATT was the most common cause of DILI (n = 50, 37.5%). The baseline characteristics and laboratory features between ATT and non-ATT induced patients with DILI are described in Supplementary Table 1. Among them, 34 patients (68%) had hepatocellular injury, 9 (18%) had mixed injury, and 7 (14%) had cholestatic type of liver injury. The clinical features between the two groups did not differ much except vomiting (32% vs 14.5%, P = 0.01) and lack of hypersensitivity features/DRESS in the ATT group. Forty three patients (86%) were on daily ATT regimen, and 7 patients were on alternate day DOTS regime. Six patients had only presumptive diagnosis of tuberculosis. Thirty-four (68%) patients among the ATT group showed recovery, as against 68 patients (81.9%) in the non-ATT group. The mean duration of drug exposure was significantly higher in the ATT group (75.1 days vs 43.3 days, P = 0.006). The ATT group had significantly lower Hb, higher ALT, prolonged INR, and lower albumin as compared with the non-ATT group.

DILI related to CAM

Fourteen patients had CAM-related liver injury and women predominated in this group (71.4%). Hepatocellular type of liver injury was found in 85.7% of them. The mean duration of drug exposure was 57.4 days among the CAM group. Six patients had severe injury, and 2 patients had moderately severe injury. CAM was predominantly used for inducing weight loss, body aches, fever, and joint pains. The patients taking CAM recovered slowly when compared with the non-CAM group (51.1 vs 29 days, P = 0.002). Three patients developed ALF, of which 1 patient underwent liver transplantation successfully. Three patients in this group died which showed a mortality rate of 21.4%.

DILI with and without DRESS

DRESS was seen in 13 patients, and the drugs implicated were anticonvulsants (phenytoin in 3 patients and sodium valproate, carbamazepine, and lamotrigine in one patient each), sulphasalazine and amoxycillin-clauvulanic acid in two patients each, and olanzapine, allopurinol, and CAM in one patient each. All patients received steroids and antibiotics along with supportive care. On analysis of patients with DILI with and without DRESS, only the MELD score was significantly higher in the group without DRESS (9.8 vs 14.9, P < 0.05), suggesting less severe liver injury in the DRESS group with no mortality (only one severe case).

Discussion

Our study showed that ATT was the major cause for DILI, accounting for 37.5% of cases with a high mortality of 20%. Other drugs implicated in mortality were CAM and neuropsychiatry drugs. CAM-related DILI constituted 10.4% with significant risk for chronicity, severe injury, and mortality. Another salient feature of CAM-induced DILI was the high mortality of 27.8% in patients with jaundice, supporting Hy's law. There was no consistent identifiable risk profile for DILI; however, time trends with serum bilirubin, INR, and serum creatinine would help in predicting mortality in these patients. In our study, the overall mortality was 13.5% and was higher with ATT (20%) and CAM (21.4%). Mortality in patients with DILI varied between 10% and 17.3% in various studies^{2,3,5,6,13,22, 23, 24} Data from Mumbai showed a mortality of 15.8%, of which 70% were secondary to ATT.²⁵ A population-based study from Spain showed a mortality of 22.7% due to ATT.²⁶ In our series also, mortality due to ATT was significantly higher than the non-ATT group (20% vs 9.6%). The incidence of ATT-induced DILI is higher in developing countries where factors such as chronic liver disease, indiscriminate use of drugs, malnutrition, and more advanced TB are higher.^8,10,18In our study, the mean duration for normalization of liver tests was 24.3 ± 23 days, which is comparable with data from Mumbai (range, 3–50 days).²⁵

Similar to other studies, neuropsychiatric and antibiotic/antifungal drugs were other common drugs implicated in DILI. Sulphamethoxazole-trimethoprim and amoxicillin-clauvulanic acid were common antibiotics implicated in our study. In Europe and the United States also, antibiotics are common causes for DILI.^{2, 3, 4, 5,26, 27, 28} CAM constituted a minor percentage (10.4%) in our study but had higher morbidity, in terms of the number of days for recovery and mortality (21.4%). A higher proportion of CAM-related DILI is underreported because of heterogeneous presentation, concomitant intake of different drugs, and difficulty in characterization regards to their contents and dosage.^{11,28, 29, 30} DRESS was seen in 13 patients, and the implicated drugs were anticonvulsants, sulphasalazine, and amoxicillin-clauvulanic acid. The mortality rates are comparatively lower in patients with DRESS due to the early identification of illness because of the skin lesions and partly due to the good response to corticosteroids. There was no mortality due to DILI in patients with DRESS in our study, and this was most probably due to occurrence of less severe forms of disease in such patients.

There was no statistically significant difference in the pattern of liver injury with morbidity or mortality. Hepatocellular type of liver injury was the most commonly observed type, followed by mixed and cholestatic injury, in comparison with other studies.^2,3,5,22 Severe liver injury was seen in 37.8%, 16.67%, and 28.5% in hepatocellular, mixed, and cholestatic groups, respectively. Among nonsurvivors in our study, the majority had hepatocellular pattern of injury. Similar to our data, a study from Mumbai showed a higher number of fatal cases with the hepatocellular type with no significant difference in mortality.²⁵ A study from Spain showed a higher mortality rate with the hepatocellular type of injury, depending on the type of drug involved and presence of jaundice.²⁶ In a study from China, cholestatic and mixed injury had higher mortality rate, whereas data from the UK showed higher mortality in the hepatocellular type of liver injury.^26,27,30,31 It is pertinent to note that the pattern of liver injury can change from one type to the other over the course of the disease. The degree of elevation of liver enzymes has poor correlation with severity of liver disease.³⁰ Another notable feature was the presence of high mortality (27.8%) in those with jaundice, demonstrating Hy's law. Hy's law has been corroborated in several studies, including those from India, and our results are comparable with previous studies.^6,13,32

Age, gender, BMI, and comorbidities form an important factor in the course of disease.^2,8,13,27 Older age and female gender were considered as predisposing factors for DILI due to pharmacodynamic changes such as drug metabolism by CYP450 enzymes, drug binding and distribution, and reduced glutathione synthesis activity.^33,34 The median age in our study was 45.8 years, and the male to female ratio was 1:1.07. In contrast to previous studies, older age and female gender were not associated with increased risk of DILI in our study. Similar results were shown in the Spanish and Danish studies.^26,28 Clinical symptoms such as anorexia, vomiting, jaundice, and abdominal pain were studied in relation to morbidity. These symptoms merely reflect adverse general condition to liver injury and did not differ among various types of liver injury and among nonsurvivors.^6,10,25,31 In our study, there was significantly lower albumin (3 ± 0.4 vs 3.6 ± 0.5 gm/dl) among nonsurvivors. An Indian study had shown a similar corelation with BMI and albumin when survivors versus nonsurvivors with ATT-induced DILI were compared.³³

Early detection of abnormalities in liver function and timely discontinuation of the drugs are very important in the management of DILI. Many studies have defined predictors of mortality in patients with DILI, which include older age, high MELD score, high bilirubin, high INR, hepatocellular type of liver injury, presence of ascites, and encephalopathy.^{2,6,10,13,26,30,33} However, the critical level of these parameters that would indicate poor outcome needs validation. In one of the recent studies from India, laboratory variables at one week predicted mortality better than those at initial recognition of DILI.²⁵ Hence, close monitoring with serial evaluation of laboratory parameters is of paramount importance. In our study, MELD score of >28, creatinine of >1.35 mg/dl, INR of >1.97, and bilirubin of >15.6 mg/dl were associated with a poor outcome. Some studies have shown that the MELD score is more sensitive than other scoring systems used in assessing the severity and prognosis of patients in ALF.^19,20 Similar clinical models to predict mortality were studied in the past.^6,35 The presence of these parameters and ascites should alert the treating physician about the high risk of mortality.

The limitation of our study was the difficulty to analyze the outcome according to the type of treatment, and the reasons are already explained. Rechallenge was tried in most of ATT cases but due to heterogeneity in its response, it was omitted in the final analysis. Owing to lower events of death and small sample size, regression analysis could not be performed.

To conclude, ATT is the most common cause for DILI in our study. Mortality due to DILI was higher in the ATT and CAM group. Our study showed that the severity of liver injury determines the outcome of patients with DILI rather than the pattern of liver injury. The outcome is less favorable in those with marked jaundice, high creatinine, lower albumin, and coagulopathy.

CRediT authorship contribution statement

N. Sunil Kumar: Collection of data of DILI case, Data analysis, Follow-up, Writing - original draft. Bhavith Remalayam: Collection of data of DILI cases, Follow-up. Varghese Thomas: Conceptualization, Collection of data of DILI cases, Follow-up, Detailed examination of cases before inclusion, Writing - original draft, Writing - review & editing. Thazhath M. Ramachandran: Writing - original draft, Data analysis. K. Sunil Kumar: Collection of data, Writing - original draft.

Conflicts of interest

The authors have none to declare.

Funding

None.

Appendix A. Supplementary data

The following is the Supplementary data to this article: