Drug induced liver injury (DILI) may be different in the East compared to the West due to differing disease prevalence, prescribing patterns and pharmacogenetic profiles. To review existing literature on causative agents of DILI in the East compared to the West, a comprehensive literature search was performed on electronic databases: MEDLINE/PubMed, Embase, Cochrane Library and China National Knowledge Infrastructure without language restrictions. Studies which involve patients having DILI and reported the frequency of causative agents were included. A random effects model was applied to synthesize the current evidence using prevalence of class-specific and agent-specific causative drugs with 95% confidence intervals. Of 6,914 articles found, 12 showed the distribution of drugs implicated in DILI in the East with a total of 33,294 patients and 16 in the West with a total of 26,069 DILI cases. In the East, the most common agents by class were anti-tuberculosis drugs (26.6%), herbal and alternative medications (25.3%), and antibiotics (15.7%), while in the West, antibiotics (34.9%), cardiovascular agents (17.3%), and non-steroidal anti-inflammatory drugs (12.5%) were the commonest. For individual agents, the most common agents in the East were isoniazid-rifampicin-pyrazinamide (25.4%), phenytoin (3.5%), and cephalosporin (2.9%) while in the West, amoxicillin-potassium clavulanate combination acid (11.3%), nimesulide (6.3%), and ibuprofen (6.1%) were the commonest. There was significant heterogeneity due to variability in single-centre compared to multi-centre studies. Differences in DILI in the East versus the West both in drug classes and individual agents are important for clinicians to recognize.
Drug induced liver injury (DILI) is defined as liver damage due to drugs, herbal medications or supplements. The pathogenesis of DILI is extremely heterogenous and causes vary from intrinsic hepatotoxicity, mitochrondrial toxicity, to immune mediated liver damage related to variations in drug metabolism, genetic differences and HLA susceptability [
DILI has the potential to be fatal with a wide spectrum of liver damage that ranges from mild abnormality to liver fibrosis to acute liver failure. Severe events are relatively rare, but can be catastrophic, particularly once jaundice occurs. This is reflected in Hy’s law [
The mainstay of treatment is withdrawal of the offending agent. In mild to moderate cases, this is followed by eventual resolution of the liver damage, but in severe cases, liver transplantation may be necessary.
There are very few studies on the incidence of DILI worldwide. The main difficulty is to acertain the demoninator, i.e., the number of individuals receiving a culprit drug. True prevalence studies are prospective community based studies with well characterised reporting systems, low dropout and loss to follow up. Consequently, there are very few population based studies.
So far the ‘best’ available population based survey was performed in Nievre, France, 2002 where the local population had only a few hospitals for their healthcare, and almost all the doctors participated in the study using a standard protocol following the International Consensus Meeting Proposal. The annual incidence rate of DILI was 13.9±2.4 per 100,000 inhabitants [
Is drug induced liver injury different in the East compared to the West? One may speculate that Asians have a different pharmacogenetic and immunological profile, and may therefore handle drugs differently. An additional problem in Asia is the widespread use of herbal and alternative medicines, the safety of which is poorly defined.
Consequently, there is an unfilled knowledge gap with regards to differences in Eastern compared to Western DILI with regards to causative agents. We aimed to determine whether agents causing DILI in the East was different from the West by performing meta-analyses of studies of DILI reported in the East, and those in the West, with regards to the most frequently reported agents. Since the reporting frequencies of DILI are most affected by reporting bias, we took measures to reduce bias.
A comprehensive literature search was performed on electronic databases: MEDLINE/PubMed, Embase, Cochrane Library and China National Knowledge Infrastructure until 31 March 2016. The specific concepts used in the search strategy were “drug induced liver injury”, “incidence” and “prevalence”. We searched these terms in combination with the conducted literature search by Medical Subject Headings or Emtree terms, and free text terms. There were no restrictions on language. All the bibliography listed in review papers and included publications were also checked.
Two investigators (E.X.S.L. and S.G.L.) independently screened for eligible studies based on pre-defined eligibility criteria. Inclusion criteria were cross sectional or cohort studies which involve patients having DILI and which reported the frequency of the causative agents. For studies that had published duplicate results with accumulating numbers of patients, only the most recent or complete reports were included. Exclusion criteria were studies which sampled less than 200 subjects, did not provide sufficient information, technical reports, editorials, letters to the editor, and case reports. Any discrepancies regarding whether articles met selection criteria were resolved by consensus.
The following data were extracted (E.X.S.L. and S.G.L.) from the included studies: 1) study characteristics (publication year, country of population, causality assessment criteria and study design); and 2) frequency of class-specific and agent-specific causative drugs. To reduce bias, frequency of agent specific causative drugs were only included if they were present in at least three studies.
The quality of each study was evaluated by two independent investigators (Z.Q. and E.C.). Quality of sampling (i.e., source of sampling and sampling methods) and quality of measurement (i.e., causality assessment criteria and consistency of criteria application) were assessed to determine the study quality. Any disagreement in quality assessment was resolved by discussion and consensus.
A random effects model was applied to synthesize the current evidence using prevalence of class-specific and agent-specific causative drugs with 95% confidence interval (CI). To assess the potential heterogeneity, we calculated the I2 for each of analysis. We also planned a sub-group analysis on single-center and multicenter study to explore the potential source of heterogeneity, as it was possible that single-center studies may lead to higher risk of bias compared to multi-center studies. Potential small-study effects and publication bias was assessed by a funnel plot. The estimated prevalence was plotted against the inverse of the standard error of the prevalence as a measure of precision reflecting the effect size. The Egger’s test was conducted with the null hypothesis that symmetry exists in the funnel plot. Statistical analyses were performed using Comprehensive Meta Analysis 3.3 (Biostat, Englewood, NJ, USA) and StataMP 15.1 (StataCorp LLC, College Station, TX, USA).
A total of 6,914 studies were found using the search strategy but after exclusions only a total of 28 studies (
In
In
In
While anti-TB agents are also antibiotics, they were classed separately since they were often quite specific and used in combination.
In
Almost all the forest plots showed significant statistical heterogeneity. One source of heterogeneity was whether the data source was a multi-centre or single centre study. It is possible that single centre studies may lead to higher risk of bias compared to multi-centre studies. There were significant differences in prevalence for individual drugs. This was especially true for Western studies in particular, such as the case of acetaminophen where multi-centre studies reported an event rate of 7% (95% CI, 0.9– 37.1%) compared to 67.5% (95% CI, 51.7–80.1%) for single centre studies (
Studies of DILI have reported varying frequency of implicated agents. Almost all of these studies come from the West where there are well established networks and infrastructure for DILI. From our systematic review, we can conclude that the agents commonly implicated in the East are quite different to those from the West. Ranking of number of causative events show that in the West, amoxicillin-clavulanate, nimesulide and ibuprofen are the common agents implicated in DILI while in the East, INH_RIF_PZA, phenytoin and cephalosporins are the commonly reported. INH_RIF_PZA is common in Eastern DILI, but is less frequent in the West.
Examining the event frequencies of classes of DILI agents are also revealing. Antibiotics 34.9%, are the most frequent class of agent in the West but only the third most frequent 15.7% in the East, and is reflected in the lower frequency by almost half. An increasingly important DILI group is herbs and supplements, for which Eastern studies show a frequency of 25.3% compared to Western studies showing a significantly lower frequency of 6.7%.
The herbs implicated are also likely to be different in the West compared to the East. In a review of herbal hepatotoxicity, Teschke and Eickhoff [
By using the frequency of reported DILI comparing Western reports to that of Eastern reports we are able to obtain an overview of the common drugs implicated and a pooled estimate of their frequency. Nonetheless, the study’s main limitation is that of reporting bias. Potential sources of bias were reduced by examining subgroups using multicentre rather than single centre studies. These were sources of heterogeneity with particular drugs such as acetaminophen. In the case of acetaminophen, this may be due to a classification problem since it may be considered a drug causing liver toxicity and tends to be treated separately from DILI as this has been excluded from the prospective DILIN study in the USA [
The certainty of pooled estimates are reflected by the narrowness of the CIs. Although this can be a little subjective, we can be more certain when CIs are relatively narrow. In this regard, Western reports of antibiotics show an event frequency of 34.9% (95% CI, 25.4–45.1%) but when examining single agents of antibiotics, amoxicillin-clavulanate 13.1% (95% CI, 10.6–16.2%), and nitrofurantoin 5.1% (95% CI, 3.3–7.8%) provide more reliable estimates while flucloxacillin 12.2% (95% CI, 4.9–27.5%), and minocycline 5.9% (95% CI, 1.4–21.0%) have wider CIs. In other drug classes of Western DILI, anti-TB drugs 8.2% (95% CI, 6.3– 10.5%) and anti-neoplastic drugs 4.7% (95% CI, 2.8–7.7%) have reasonably narrow CIs. With regards to Eastern DILI by drug class, NSAIDs 4.8% (95% CI, 2.2–8.2%) has reasonable CIs and for the single agents, carbamazepine 1.8% (95% CI, 0.7–4.7%) has reasonable CIs. Nonetheless, despite the wider CIs in some instances, the overall findings and rankings of both classes of agents and individual agents provides a picture of differences in DILI between East and West.
Could the differences between Western and Eastern DILI be explained by pharmacogenetic differences? One well studied field is TB drug related DILI. A meta analysis by Cai et al. [
The strength of our study is that it provides for the first time frequencies and ranking of DILI comparing East and West, despite the limitations of the quality of the DILI reports, differences in disease prevalence and patterns of drug prescribing. Overall, the data syntheses show differences in DILI reporting and prevalence between East and West, both in drug class and individual agents. How should we use this information? The first clinical utility is that clinicians should be aware that certain classes and individual agents are more susceptible in the East to DILI, and monitor such patients more carefully, and the second is that certain classes of agents should be avoided (herbs and supplements) if they provide no clear benefit but come with increased risk, and finally drugs that are potentially hepatotoxic in the West, may not be so in the East. In the first instance, TB drug related DILI in the East has a 26.6% frequency amongst all DILI (but only 8.2% in the West, see
As discussed previously, limitations of our study are that the reported differences in prevalence of DILI may be confounded by the frequency of the disease in the countries concerned and consequently the frequency of the drugs prescribed, and not just explained by pharmacogenomics differences. It is beyond the scope of our study to examine such limitations, which can only be addressed by prospective collection of data in country specific databases. An example of differences in disease prevalence and incidence that could impact DILI reporting is that of TB. The World Health Organisation Tuberculosis Global Report [
In conclusion, this study has shown clear differences in frequency of drugs implicated in DILI reported in Western studies and compared to Eastern studies, both by drug class (antibiotics are frequent DILI in the West while anti-TB drugs are frequent in the East), and for specific agents (Amoxicillin-clavulanate related DILI is the common in the West while INH_RIF_PZA is the common in the East). Knowledge of such differences and their pooled estimates of frequency, now provides clinicians of more precise dangers of DILI during prescribing, and to advise patients of potential DILI when taking herbs and supplements. This field is also an area where research into genetic polymorphisms drug handling are much needed.
Overall concept and study design: SGL
Search, retrieval and evaluation of articles: SGL, EXSL
Analysis of quality and statistics: QZ, EC
Manuscript draft and editing: SGL, EXSL, QZ, EC
Final approval: SGL, EXSL, QZ, EC
Supplementary material is available at Clinical and Molecular Hepatology website (
Western DILI by class of agents: forest plots. CI, confidence interval; TB, tuberculosis; CVD, cardiovascular; NSAIDs, non-steroidal anti-inflammatory drugs; DILI, drug induced liver injury.
Eastern DILI by class of agents: forest plots. CI, confidence interval; TB, tuberculosis; CVD, cardiovascular; NSAIDs, non-steroidal anti-inflammatory drugs; DILI, drug induced liver injury.
Western DILI by single agents: forest plots. CI, confidence interval; INH_RIF_PZA, in the East was isoniazid-rifampicin-pyrazinamide; DILI, drug induced liver injury.
Eastern DILI by single agents: forest plots. CI, confidence interval; INH_RIF_PZA, in the East was isoniazid-rifampicin-pyrazinamide; DILI, drug induced liver injury.
Funnel plot for antibiotics in the East.
Funnel plot for antibiotics in the West.
INH_RIF_PZA, in the East was isoniazid-rifampicin-pyrazinamide; TB, tuberculosis; NA, not available; CVD, cardiovascular; NSAID, non-steroidal anti-inflammatory drug.
confidence interval
cardiovascular disease
drug induced liver injury
drug metabolizing enzyme
isoniazid-rifampicin-pyrazinamide
non-steroidal anti-inflammatory drugs
odds ratio
tuberculosis
World Health Organisation-Uppsala Monitoring Center
Flowchart of the identification, screening, eligibility, and included studies. CKNI, China Knowledge Network CKNI Database; DILI, drug induced liver injury.
(A) DILI by drug class of Eastern studies. Summary pooled estimates are shown as ES and 95% CIs. Where I2 and
(A) DILI by individual agents in Eastern studies. Summary pooled estimates are shown as ES and 95% CIs. Where I2 and
Characteristics of the 28 selected studies
No | Study | Country/region | Causality assessment for DILI | Time periods | Publication type (report, journal article, abstract) | Hospital-/registrybased | Drug classification type (single agent or class) | Total number of DILI records | Included acetaminophen |
---|---|---|---|---|---|---|---|---|---|
Studies from the West | |||||||||
1 | Andrade et al. [ |
Spain | CIOMS/RUCAM | Apr 1994 to Aug-04 | Prospective | Prospective Regional registry | Both | 461 | Yes 13 cases |
Clinical judgement, CIOMS by 3 independent experts | |||||||||
2 | Bessone et al. [ |
Latin America | CIOMS/RUCAM | 2011 to 2014 | Prospective | Latin American registry | Single | 206 | Yes but 0 cases |
3 | Björnsson et al. [ |
Iceland | CIOMS/RUCAM | Mar 2010 to Feb 2012 | Prospective | National | Both | 96 | No (specifically excluded) |
4 | Chalasani et al. [ |
USA | DILIN methods | Sep 2004 to May 2013 | Observational longitudinal | National | Both | 899 | No (specifically excluded) |
5 | Sgro et al. [ |
France | Global imputability score I | Nov 1997 to Nov 2000 | Population based | Regional | Both | 34 | Yes (1 died: prolonged acetaminophen) |
6 | De Valle et al. [ |
Swedish outpatient clinic | CIOMS | 1995 to 2005 | Retrospective review of case records | 1 university hospital outpatients urban area | Both | 77 | Not mentioned |
7 | de Abajo et al. [ |
UK | Description like RUCAM but not specifically stated | 1994 to 1999 | Population based case control (5,000 controls) | GP research database in the UK | Both | 128 | Yes: 12 from acetaminophen |
8 | Ibáñez et al. [ |
Spain | Description like RUCAM but not specifically stated | 1992 to 1998 | Population based prospective | 12 hospitals collaborating network | Both | 107 | Yes: 12 from acetaminophen |
9 | Carey et al. [ |
US Mayo (inpatients) | CIOMS | 1998 to 2006 | Retrospective search with codes | Inpatient visits at Mayo Hospital | Both | 40 | Yes: 40 had DILI, 27 from acetaminophen |
11 | Hussaini et al. [ |
Cornwall England | CIOMS | 1998 to 2004 | Retrospective analysis | Rural population | Both | 28 | Not mentioned |
12 | Meier et al. [ |
Switzerland | CIOMS | Jan 1996 to Dec 2000 | Medical records review | Pharmacoepidemiological databases | Both | 88 | Yes |
13 | Sabaté et al. [ |
Barcelona Spain 12 hospitals | Jaundice, ALT AST | Jan 1993 to Dec 1999 | Multi centre prospective case control | 12 hospitals | Drugs | 126 | Yes |
2,700,000 | |||||||||
14 | Sistanizad Peterson. [ |
Tasmania Australia | CIOMS | Jun 2008 to July 2009 | Retrospective | 1 major hospital serving 250,000 people | Drugs | 17 | Yes |
15 | Galan et al. [ |
US | Description like RUCAM but not specifically stated | 1993 to 2002 | Retrospective review of records | 1 tertiary care hospital | Both | 32 | Yes but 0 cases |
16 | Vega et al. [ |
Delaware, US | CIOMS | 2014 | Prospective | DILN | Both | 23 | Yes |
Studies from the East | |||||||||
17 | Suk et al. [ |
Korea | RUCAM | May 2005 to May 2007 | Prospective | 17 referral hospitals (nationwide) | Class | 371 | Yes |
18 | Kwon et al. [ |
Korea | WHO-UMC | Jan 2007 to Dec 2008 | Retrospective-registry of spontaneous reports of adverse drug reactions | 9 regional pharmacovigilence centres in Korea (nationwide) | Both | 567 | Yes |
19 | Zhou et al. [ |
China | Various | 1994 to 2011 | Retrospective review of electronic and manual searches | Multiple centres | Class | 24,112 | Yes |
20 | Ou et al. [ |
China | CIOMS | Jan 2011 to Dec 2014 | Retrospective review of inpatient records | Inpatients - 1 hospital | Both | 361 | Yes |
21 | Lee et al. [ |
Taiwan | ICD–9 code (case cross over comparison of diagnosis) | 1997 to 2004 | Retrospective | Population based database (insurance) | Both | 4,857 | Unclear |
22 | Takikawa et al. [ |
Japan | DDW-J 2004 | Jan 1997 to Dec 2006 | Retrospective | 29 facilities (nationwide) | Class | 1,676 | Yes |
23 | Huang et al. [ |
Taiwan | RUCAM | Unclear | Retrospective | 6 medical centres across Taiwan | Class | 1,099 | Yes |
24 | Sobhonslidsuk et al. [ |
Thailand | ICD-10 (toxic liver disease) | 2009 to 2013 | Retrospective | Population based database (nationwide) (DILN Taiwan) | Single | 589 | Yes |
25 | Rathi et al. [ |
India | RUCAM | 2014 to 2015 | Prospective | 1 tertiary care hospital in metropolitan India | Class | 82 | Unclear |
26 | Bektas et al. [ |
Turkey | Various | Unclear | Retrospective | Single centre | Class | 170 | Unclear |
27 | Jaiprakash et al. [ |
India | AST/ALT | Jul 2006 to Jul 2007 | Retrospective | 1 tertiary care hospital in rural South India | Class | 65 | Unclear |
28 | Devarbhavi et al. [ |
India | RUCAM | 1997 to 2008 | Retrospective | Inpatients - 1 hospital | Both | 244 | Yes |
DILI, drug induced liver injury; CIOMS, Council for International Organizations of Medical Sciences; RUCAM, Roussel Uclaf Causality Assessment Method; GP, general practice; ALT, alanine aminotransferase; AST, aspartate transferase; WHO-UMC, World Health Organisation-Uppsala Monitoring Center; ICD, International Classification of Diseases; DDW-J, The Digestive Disease Week Japan.
Summary of ranks by class and single agents by class (with all studies)
Rank | West | No. of studies | DILI event | Total DILIs | Prevalence (%), range | East | No. of studies | DILI event | Total DILIs | Prevalence (%), range |
---|---|---|---|---|---|---|---|---|---|---|
1 | Antibiotics | 15 | 1,167 | 3,613 | 34.9% (25.4%, 45.1%) | Anti-TB | 4 | 563 | 2,340 | 26.6% (13.1%, 42.9%) |
2 | CVD agents | 6 | 392 | 2,868 | 17.3% (7.8%, 29.5%) | Herbal and sup | 8 | 914 | 4,164 | 25.3% (12.5%, 40.6%) |
3 | Psychotropic | 7 | 161 | 1,512 | 13.1% (6.8%, 21.0%) | Antibiotics | 9 | 554 | 4,380 | 15.7% (9.0%, 23.9%) |
4 | NSAIDs | 10 | 307 | 3,252 | 12.5% (6.8%, 19.8%) | Psychotropic | 5 | 251 | 2,665 | 8.2% (4.4%, 12.8%) |
5 | Herbal and sup | 4 | 184 | 2,094 | 6.7% (1.2%, 16.0%) | NSAIDs | 5 | 256 | 3,666 | 4.8% (2.2%, 8.2%) |
DILI, drug induced liver injury; TB, tuberculosis; CVD, cardiovascular; NSAIDs, non-steroidal anti-inflammatory drugs.
Summary of ranks by class and single agents by single agents
Rank | West | No. of studies | DILI event | Total DILIs | Prevalence (%), range | East | No. of studies | DILI event | Total DILIs | Prevalence (%), range |
---|---|---|---|---|---|---|---|---|---|---|
1 | Amoxicillin- clavulanate | 15 | 356 | 2,977 | 11.3% (8.4%, 14.8%) | INH_RIF_PZA | 4 | 348 | 1,270 | 25.4% (8.3%, 47.7%) |
2 | Nimesulide | 4 | 77 | 1,485 | 6.3% (0.87%, 15.9%) | Phenytoin | 4 | 25 | 1,270 | 3.5% (0.6%, 8.2%) |
3 | Ibuprofen | 5 | 68 | 1,389 | 6.1% (2.8%, 10.4%) | Cephalosporin | 3 | 17 | 1,241 | 2.9% (0.0%, 10.1%) |
4 | INH_RIF_PZA | 8 | 109 | 2,027 | 4.6% (3.2%, 6.2%) | Carbamazepine | 3 | 61 | 1,241 | 1.3% (0.4%, 2.7%) |
5 | Diclofenac | 8 | 75 | 2,588 | 3.7% (1.8%, 6.2%) | Valproate | 3 | 37 | 909 | 0.3% (0.0%, 1.8%) |
DILI, drug induced liver injury; INH_RIF_PZA, isoniazid-rifampicin-pyrazinamide.
Comparison of herbal DILI between East and West
East | West |
---|---|
Tu San Qi ( |
Lu Cha ( |
Shou Wu Pian ( |
Heliotropoium ( |
Chai Hu ( |
Herbalife ( |
Xiao Chai Hu Tang ( |
Greater Celandine ( |
Bai Xian Pi ( |
Kava ( |
Chi R Yun ( |
Chaparral ( |
Huang Qin ( |
Lycodium similiaplex ( |
Long Dan Xie Gan Tang (n=17) | Germander ( |
Yin Chen Hao ( |
Hydroxcut ( |
Kudzu ( |
Jin Bu Huan ( |
Adapted from Teschke et al. [
DILI, drug induced liver injury.