NAT2 Polymorphisms and Susceptibility to Anti-Tuberculosis Drug-Induced Liver Injury: a Meta-Analysis

NAT2 polymorphisms and susceptibility to anti-tuberculosis drug-induced liver injury: a meta-analysis

P-Y. Wang,*† S-Y. Xie,* Q. Hao,‡ C. Zhang,* B-F. Jiang† * Binzhou Medical University, Yantai, † School of Public Health, Shandong University, Jinan, ‡ Binzhou Central Hospital, Binzhou, China

SUMMARY

BACKGROUND AND OBJECTIVE: Although a series of slow acetylators and the risk of ATLI. The OR for NAT2 studies have evaluated the potential association between slow acetylators compared with rapid acetylators was N-acetyltransferase 2 (NAT2) polymorphisms and the 4.697 (95%CI 3.291–6.705, P < 0.001). Subgroup anal- risk of anti-tuberculosis drug-induced liver injury (ATLI), yses indicate that both Asian and non-Asian cases with the results have generally been controversial and inade- slow acetylators develop ATLI more frequently, which is quate, mainly due to limited power. The present meta- similar to patients with slow acetylators receiving first- analysis sought to resolve this problem. line combination treatment. On comparing NAT2 inter- DESIGN: PubMed, Embase and Web of Science were mediate acetylators with rapid acetylators, the OR for searched using the following key words: ‘N-acetyltrans- ATLI was 1.261 (95%CI 0.928–1.712, P = 0.138). ferase 2’ or ‘NAT2’ and ‘polymorphism’ and ‘tuberculo- CONCLUSIONS: This meta-analysis showed that tuber- sis’ or ‘TB’ and ‘hepatotoxicity’ or ‘liver injury’. Crude culosis patients with slow acetylators had a higher risk odds ratios (ORs) with 95% confidence intervals (CIs) of ATLI than other acetylators. Screening of patients for were summarised in forest plots and set out in a table. the NAT2 genetic polymorphisms will be useful for the RESULTS: A total of 14 studies, comprising 474 cases clinical prediction and prevention of ATLI. and 1446 controls, were included in the meta-analysis. KEY WORDS: N-acetyltransferase 2; polymorphism; tu- A significant association was observed between NAT2 berculosis; drug-induced liver injury; meta-analysis

TUBERCULOSIS (TB) remains a major global public notype varies remarkably with ethnic origin. It has health problem. Isoniazid (INH), rifampicin (RMP), been demonstrated that NAT2 genetic polymor- ethambutol (EMB) and pyrazinamide (PZA) are phisms play an important role in the effi ciency of used in combination for ⩾6 months as fi rst-line the enzyme for the metabolism of several carcino- anti-tuberculosis treatment. However, TB patients genic arylamines and drugs.37–40 To date, a series frequently face an increasing occurrence of adverse of studies have evaluated the potential association anti-tuberculosis drug reactions during long-term between NAT2 and the risk of ATLI, but results have treatment. One of the most serious adverse drug reac- generally been inadequate and controversial.17–36 tions is anti-tuberculosis drug-induced liver injury Earlier studies suggested that cases with fast ace- (ATLI), associated with high morbidity and mortal- tylators are more susceptible to developing ATLI,39,40 ity, as well as with increased treatment costs.1,2 The while more recent studies report that cases with incidence of ATLI ranges from 1% to 47%, depend- slow acetylators develop ATLI more often than fast ing on the anti-tuberculosis drugs used, race, age, sex, acetylators.17,19,25–27 nutrition status, alcohol intake, severity of TB, hu- As the sample size of each of the above-mentioned man immunodefi ciency virus (HIV) infection status, studies is relatively small, a meta-analysis was there- and presence of hepatitis B and C virus.3–16 fore necessary to solve the problem of inadequate In the last few years, an increasing number of statistical power and controversial results based on ac- studies have suggested that genetic polymorphisms cumulated data. The present study is a meta-analysis may be associated with susceptibility to ATLI.17–36 of all relevant studies published up to 1 August 2011 NAT2 is located on chromosome 8p21.3–23.1, and carried out to examine the relationship between codes for a phase II xenobiotic metabolising enzyme. NAT2 genetic polymorphisms and susceptibility to The frequency of NAT2 alleles and acetylator phe- ATLI in pulmonary TB patients.

Correspondence to: B-F Jiang, School of Public Health, Shandong University, 44 Wenhua Xi Road, Jinan, Shandong 250012, Peoples’ Republic of China. Tel/Fax: (+86) 531 883 821 4188. e-mail: [email protected]; or S-Y Xie, Binzhou Medical University, Yantai, P R China. e-mail: [email protected] Article submitted 28 May 2011. Final version accepted 29 October 2011. 590 The International Journal of Tuberculosis and Lung Disease

SUBJECTS AND METHODS The following characteristics were extracted from each study if available: 1) name of fi rst author; 2) pub- Search strategy lication year; 3) country or region of origin; 4) num- The PubMed, Embase and Web of Science databases ber of cases and controls, and numbers of cases and were searched using the following key words: ‘N- controls of different NAT2 polymorphism distribu- acetyltransferase 2’ or ‘NAT2’ and ‘polymorphism’ tion data; and 5) anti-tuberculosis drugs used. and ‘tuberculosis’ or ‘TB’ and ‘hepatotoxicity’ or ‘liver injury’. The search was restricted to case-control stud- Statistical analysis ies published in English and Chinese up to 1 August Based on complete distribution data on NAT2 poly- 2011. All of the studies were reviewed, and a manual morphism in cases and controls, the pooled ORs and search of the original studies and references was per- their 95% confidence intervals (CIs) were calculated formed to identify additional relevant articles. and displayed as forest plots to assess the strength of Inclusion and exclusion criteria association between NAT2 genetic polymorphisms Study quality was assessed according to Little’s pro- and susceptibility to ATLI in pulmonary TB patients. posed checklist for reporting and appraising studies of Subgroup analysis was estimated by race-specific ORs genotype prevalence and gene-disease associations.41 (separately for Asian and non-Asian subjects) or drug- These items included 1) analytic validity of geno- specific ORs (separately for INH+RMP+PZA+EMB, typing; 2) selection of study subjects; 3) confounding, INH+RMP or INH). including population stratifi cation; and 4) statistical The heterogeneity of the study results was explored issues.41 A study in which all or most of the afore- using both the Cochrane Q statistic and estimating mentioned criteria are satisfi ed would be graded as the I2 test.43 In case of heterogeneity (P < 0.05 for high quality. Inclusion criteria for this meta-analysis the Q-test, I2 > 50%), the random-effects model were the following: 1) case-control studies: cases were (DerSimonian Laird) was chosen. Otherwise, the fixed- TB patients with ATLI, while controls were TB pa- effect model (Mantel-Haenszel method) was used tients without ATLI; 2) genetic detection method using to calculate the pooled OR. The meta-analysis was polymerase chain reaction with restriction fragment performed using the ‘metan’ STATA command (Stata length polymorphism (PCR-RFLP); 3) clearly defi ned Corp, College Station, TX, USA). inclusion and exclusion criteria for study subjects; A sensitivity analysis was performed to assess the 4) explicit diagnostic criteria for TB and ATLI; and stability of the results. Random-effects model, fi xed- 5) correct reporting statistical analysis and study re- effects model and subgroup analysis were used to sults, such as odds ratios (ORs). Animal studies and examine any differences. In addition, a single study pure therapy studies without complete distribution included in the meta-analysis was removed each data on NAT2 polymorphisms were excluded. time to reflect the influence of the ORs. Publication In our analysis, study subjects were adult patients bias was assessed using Begg and Egger’s formal sta- newly diagnosed with active TB, with lesions of TB tistical test (statistical significance was defined asP < clearly visible on chest X-ray or computed tomogra- 0.10).44,45 The meta-analysis was performed using phy, or positive sputum smear or culture results for the the ‘metabias’ STATA command. detection of mycobacteria. To avoid confusion and ren- der the diagnosis of ATLI reliable, exclusion criteria included 1) abnormal serum alanine transferase (ALT), RESULTS aspartate transaminase (AST) or bilirubin levels or Characteristics of included studies symptoms related to abnormal liver function, such as A total of 17 papers were found to meet the inclu- jaundice, before anti-tuberculosis treatment; 2) alcohol- sion criteria.17–20,23–28,30–36 The study by Huang et al. related liver disease or habitual alcohol drinking; and was excluded due to overlap with another study;20 3) any other hepatic or systemic diseases that may the first study, published in 2002, was selected for cause liver dysfunction, such as carriers of the hepati- its more detailed NAT2 polymorphism distribution tis B or C virus, heart failure, respiratory failure, etc. data.19 The study by Roy et al. was excluded due to According to the criteria established by an inter- the absence of complete NAT2 polymorphism dis- national consensus meeting organised by the Council tribution data.18 The study by Possuelo et al. was for International Organizations of Medical Sciences also excluded, as it included HIV and HCV pa- (CIOMS), ATLI was defined as an increase in liver bio- tients.28 Overall, 14 studies, including 474 cases and chemical parameters more than two times the upper 1446 controls, were included in the meta-analysis. limit of normal during anti-tuberculosis treatment.42 The patient characteristics were described in the stud- Data extraction ies selected, and the studies satisfi ed most of the qual- Two investigators independently extracted data using ity assessment criteria specifi ed by Little et al.41 The standardised forms. If they could not come to an main fi ndings of each study included in the meta- agreement, a third investigator adjudicated. analysis, and related information, are summarised in NAT2 gene relative to drug-induced liver injury 591

Table Main characteristics of studies included in the meta-analysis

First author, Sample No. reference Year Country size Cases Controls Drugs 1 Ohno17 2000 Japan 77 14 63 INH+RMP 2 Huang19 2002 Taiwan 224 33 191 INH+RMP+PZA+EMB 3 Vuilleumier24 2006 Switzerland 89 8 81 INH 4 Shimizu23 2006 Asian 42 10 32 INH+RMP 5 Cho25 2007 Korean 132 18 114 INH+RMP+PZA+EMB 6 Higuchi26 2007 Japan 100 18 82 INH+RMP 7 Bozok27 2008 Turkey 100 30 70 INH+RMP+PZA+EMB 8 Lee30 2010 Taiwan 140 45 95 INH+RMP+PZA+EMB 9 Bose32 2011 India 218 41 177 INH+RMP+PZA+EMB 10 Kim34 2009 Korea 226 67159 INH+RMP+PZA+EMB 11 Yamada33 2009 USA 170 23 147 INH 12 An35 2011 China 208 101 107 INH+RMP+PZA+EMB 13 Khalili31 2011 Iran 50 14 36 INH+RMP+PZA+EMB 14 Sotsuka36 2011 Japan 144 52 92 INH+RMP+PZA+EMB

INH = isoniazid; RMP = rifampicin; PZA = pyrazinamide; EMB = ethambutol.

Appendix Table A.1,* which shows the main charac- On comparing slow NAT2 acetylators with other teristics of the 14 studies, including first author, year acetylators (rapid plus intermediate acetylators), of publication, country of origin, numbers of cases C ochran Q statistics and I2 estimates indicated pres- and controls and anti-tuberculosis drugs used. ence of heterogeneity (heterogeneity χ2 = 23.33, P = The sample sizes of the 14 studies ranged from 42 0.038 and I2 = 44.3%); therefore, both fi xed- and to 226. Almost all of the subjects were newly diag- random-effects models were performed; the OR was nosed and on treatment for pulmonary TB (age range respectively 3.732 (95%CI 2.906–4.793) and 3.969 38–73 years). Patients were seen at regular intervals (95%CI 2.747–5.736) in the fi xed- and random- and were questioned about their symptoms and ad- e ffects models. verse reactions to anti-tuberculosis drugs. In the case- Of the 14 studies, 11 were conducted among Asians. control studies, cases were patients with ATLI, and A subgroup analysis for different regions (Asian vs. controls were patients without ATLI. Individuals ho- non-Asian) was performed. On comparing slow NAT2 mozygous for rapid NAT2 acetylator alleles (NAT2*4, acetylators with rapid acetylators, the combined ORs NAT2*11A, NAT2*12A, NAT2*12B, NAT2*12C, for Asians and non-Asians were respectively 4.88 NAT2*13) were classifi ed as rapid acetylator pheno- (95%CI 3.349–7.111, P < 0.001) and 3.72 (95%CI type; individuals homozygous for slow acetylator 1.318–10.469, P = 0.013; Figure 1). alleles were classifi ed as slow acetylator phenotypes, Patients received INH+RMP+PZA+EMB in nine and heterozygous individuals (one rapid and one slow studies, while in the fi ve other publications they re- NAT2 allele) were classifi ed as intermediate acety- ceived only INH or INH+RMP. A subgroup analysis lator phenotypes.46 for different treatment combinations was therefore performed. The combined ORs for INH+RMP, INH+ NAT2 acetylators and susceptibility to ATLI RMP+PZA+EMB and INH were respectively 34.303 When all studies were pooled in the meta-analysis, a (95%CI 10.413–113.000, P < 0.001), 4.092 (95%CI signifi cant association was observed between slow 2.777–6.028, P < 0.001) and 2.362 (95%CI 0.520– NAT2 acetylators and the risk of ATLI (Figure 1). As 10.729, P = 0.266) on comparing slow NAT2 acety- Cochrane Q statistics and the I2 estimate indicated lators with rapid acetylators (Figure 2). little heterogeneity (heterogeneity χ2 = 15.08, P = This meta-analysis also evaluated the risk for 0.303 and I2 = 13.8%), the fi xed-effects model was s everal major NAT2 genotypes and susceptibility to chosen; the OR for slow NAT2 acetylators compared ATLI. There were statistically signifi cant associa- with rapid acetylators was 4.697 (95%CI 3.291– tions between NAT2*5/7, NAT2*6/6, NAT2*6/7 and 6.705, P < 0.001, Figure 1). On comparing interme- NAT2*7/7 and the risk of ATLI (Appendix Table A.2). diate NAT2 acetylators with rapid acetylators, the fi xed-effects model used for Q statistics and estimated Sensitivity analysis and publication bias I2 showed no evidence of study heterogeneity (het- The results in the random- and fi xed-effects models as erogeneity χ2 = 9.15, P = 0.762 and I2 = 0.0%), well as the subgroup analysis were similar. On classi- while the OR for ATLI was 1.261 (95%CI 0.928– fying NAT2 acetylators as rapid, intermediate and slow 1.712, P = 0.138). for comparison, study heterogeneity visibly decreased. A single study with a small sample size and exces- sively wide CI involved in the meta-analysis was de- * The Appendix is available in the online version of this article at leted each time; the corresponding ORs were similar http://www.ingentaconnect.com/content/iuatld/ijtld/2012/00000016/ to the previous results (data not shown), indicating 00000005/art00006 that our results were statistically robust. There was 592 The International Journal of Tuberculosis and Lung Disease

Figure 1 Association between NAT2 polymorphisms and the risk of ATLI. ORs for NAT2 slow acetylators compared to rapid acetylators, and ORs by subgroup analysis for Asian and non-Asian populations. OR = odds ratio; CI = confidence interval; ATLI = anti-tuberculosis drug-induced liver injury. no evidence of publication bias according to Begg’s NAT2 acetylator phenotype (P < 0.001). Moreover, (P = 0.240) and Egger’s tests (P = 0.213). subgroup analyses indicate that both Asians and non- Asians with slow acetylators develop ATLI more fre- DISCUSSION quently, similarly to patients with slow acetylators receiving fi rst-line combination treatment. This fi nd- Principal findings ing is biologically plausible. This meta-analysis provides compelling evidence for Different laboratories recently focused on molecu- an increased risk of ATLI associated with the slow lar epidemiology studies to examine the association

Figure 2 Drug-speciﬁc ORs for NAT2 slow acetylators compared to rapid acetylators by subgroup analysis. OR = odds ratio; CI = confidence interval; INH = isoniazid; RMP = rifampicin; PZA = pyrazinamide; EMB = ethambutol. NAT2 gene relative to drug-induced liver injury 593

Figure 3 Diagram of interplay between drugs and enzymes that causes anti-tuberculosis drug-induced hepatotoxicity (based on Roy et al.18). INH = isoniazid; RMP = rifampicin; GSH = glutathione. between polymorphisms in different drug-metabolising ies, more studies with different drug treatments may enzymes and risk of ATLI. A proposed model of the be necessary for further progress in this area. ATLI is described in Figure 3.18,21 According to Fig- ure 3, NAT2 is one of the main enzymes involved Strengths and limitations of the meta-analysis in anti-drug metabolism in the liver. Diminution or Several limitations of this meta-analysis should be disturbance of NAT2 activity could result in the mentioned. First, the analysis was limited by the rela- a ccumulation of precursors such as hydrazine and tively small number of available studies and the geno- acetyldiazene in the liver, leading to hepatotoxicity; types of outcomes reported in the studies. the effi ciency of this process depends partially on the Second, our results were based on unadjusted esti- polymorphic alleles present in the individual.18,21 Slow mates. If raw data on all subjects had been available acetylators were acetylated more slowly, not only in and adjustments by other covariants, including envi- the parent component but also in immediate pre- ronmental factors and other gene polymorphisms, such cursors of toxic intermediates. Slow acetylators may as cytochrome P450 2E1 (CYP2E1), were taken into therefore indirectly critically increase the accumula- account, our results would have been more accurate. tion of toxic metabolites.21 Third, limiting the study to English or Chinese lan- Previous discrepant results regarding acetylator guage articles may potentially lead to language bias. status and ATLI may be due to the different pheno- In addition, studies with no statistically signifi cant re- typing methods used.17–36 It is diffi cult to compare sults often have a lower likelihood of being published. the accuracy of different NAT phenotyping methods Although available statistical approaches for publica- or of different cut-off points using the same pheno- tion bias (Begg’s and Egger’s tests) did not indicate typing method.17–36 In this meta-analysis, the PCR- clear evidence of bias, it is still diffi cult to rule out po- RFLP method was performed in selected studies; this tential publication bias in the meta-analysis. enabled us to distinguish possible NAT2 polymor- Despite these limitations, our meta-analysis had phisms and classify NAT2 acetylator types more several strengths. In genetic association studies, sam- accurately. ple size and statistical power are often of particular Signifi cantly increased risks in slow acetylators importance for rare conditions such as adverse drug were observed in patients receiving INH+RMP and reactions.34 To overcome the limitations in the statis- INH+RMP+PZA+EMB compared to patients re- tical signifi cance of single studies involving a relative ceiving INH alone. This suggests that potentially dif- small number of subjects, more suffi cient numbers of ferent mechanisms occur with different drug treat- cases and controls were pooled from different stud- ments. However, it should be noted that the total ies, which signifi cantly increased the statistical power sample size of patients receiving INH+RMP was of the analysis. In addition, when presence of hetero- only 219 in three studies, while that of patients re- geneity was detected, random-effects model and sub- ceiving INH alone was 259 in only two publications. group analysis were used, and heterogeneity was con- Due to the relatively small number of available stud- siderably reduced. For the sensitivity analysis, a single 594 The International Journal of Tuberculosis and Lung Disease study involved in the meta-analysis was deleted each to antitubercular medicines and assessment of risk factors. Ann time; the results remained similar, indicating the sta- Pharmacother 2004; 38: 1074–1079. bility of our results. 17 Ohno M, Yamaguchi I, Yamamoto I, et al. Slow N-acetyltransferase 2 genotype affects the incidence of isoniazid and In conclusion, this meta-analysis showed that TB rifampicin-induced hepatotoxicity. Int J Tuberc Lung Dis 2000; patients with a slow acetylator genotype had a higher 4: 256–261. risk of ATLI than patients with rapid or intermediate 18 Roy B, Chowdhury A, Kundu S, et al. Increased risk of anti- acetylators. Screening of patients for the NAT2 ge- tuberculosis drug-induced hepatotoxicity in individuals with netic polymorphisms can prove clinically useful for glutathione S-transferase M1 ‘null’ mutation. J Gastroenterol Hepatol 2001; 16: 1033–1037. the prediction and prevention of ATLI. 19 Huang Y S, Chern H D, Su W J, et al. Polymorphism of the N-acetyltransferase 2 gene as a susceptibility risk factor of Acknowledgements anti-tuberculosis drug-induced hepatitis. Hepatology 2002; 35: 883–889. This work was supported by grants from the Foundation of 20 Huang Y S, Chern H D, Su W J, et al. Cytochrome P450 2E1 Shandong Natural Science (no. ZR2009CL005) and the Scientific genetic polymorphism and the susceptibility to anti-tuberculosis Research Foundation of Shandong Educational Committee (no. drug-induced hepatitis. Hepatology 2003; 37: 924–930. J09LF11, J08LG15). 21 Roy B. Anti-tuberculosis drug induced hepatotoxicity and g enetic polymorphisms in drug-metabolizing genes. Curr P harmacogenomics 2003; 1: 101–109. References 22 Hussain Z, Kar P, Husain S A. Anti-tuberculosis drug-induced 1 Yee D, Valiquette C, Pelletier M, et al. Incidence of serious side hepatitis: risk factors, prevention and management. Indian J effects from fi rst-line anti-tuberculosis drugs among patients Exp Biol 2003; 41: 1226–1232. treated for active tuberculosis. Am J Respir Crit Care Med 23 Shimizu Y, Dobashi K, Mita Y, et al. DNA microarray genotyp- 2003; 167: 1472–1477. ing of N-acetyltransferase 2 polymorphism using carbodimide 2 Lundkvist J, Jonsson B. Pharmacoeconomics of adverse drug as the linker for assessment of isoniazid hepatotoxicity. Tuber- reactions. Fundam Clin Pharmacol 2004; 18: 275–280. culosis (Edinb) 2006; 86: 374–381. 3 Parrish A G, Robson S C, Trey C, et al. Retrospective survey of 24 Vuilleumier N, Rossier M F, Chiappe A, et al. CYP2E1 geno- drug-induced liver disease at Groot Schuur Hospital, Cape type and isoniazid-induced hepatotoxicity in patients treated for Town—1983–1987. S Afr Med J 1990; 77: 199–202. latent tuberculosis. Eur J Clin Pharmacol 2006; 62: 423–429. 4 Wu J C, Lee S D, Yeh P F, et al. Isoniazid-rifampin-induced 25 Cho H J, Koh W J, Ryu Y J, et al. Genetic polymorphisms of hepatitis in hepatitis B carriers. Gastroenterology 1990; 98: NAT2 and CYP2E1 associated with anti-tuberculosis drug- 502–504. i nduced hepatotoxicity in Korean patients with pulmonary tu- 5 Farrell G C. Drug-induced hepatic injury. J Gastroenterol berculosis. Tuberculosis (Edinb) 2007; 87: 551–556. Hepatol 1997; 12: S242–S250. 26 Higuchi N, Tahara N, Yanagihara K, et al. NAT2 6A, a haplo- 6 Durand F, Bernuau J, Pessayre D, et al. Deleterious infl uence of type of the N-acetyltransferase 2 gene, is an important bio- pyrazinamide on the outcome of patients with fulminant or marker for risk of anti-tuberculosis drug-induced hepatotoxic- subfulminant liver failure during antituberculous treatment in- ity in Japanese patients with tuberculosis. World J Gastroenterol cluding isoniazid. Hepatology 1995; 21: 929–932. 2007; 13: 6003–6008. 7 Singh J, Garg P K, Thakur V S, et al. Anti-tubercular treatment 27 Bozok Cetintas‚ V, Erer O F, Kosova B, et al. Determining the induced hepatotoxicity: does acetylator status matter? Indian J relation between N-acetyltransferase-2 acetylator phenotype Physiol Pharmacol 1995; 39: 43–46. and anti-tuberculosis drug-induced hepatitis by molecular bio- 8 Pande J N, Singh S P N, Khilnani G C, et al. Risk factors for logic tests. Tuberk Toraks 2008; 56: 81–86. hepatotoxicity from antituberculosis drugs: a case-control study. 28 Possuelo L G, Castelan J A, de Brito T C, et al. Association of Thorax 1996; 51: 132–136. slow N-acetyltransferase 2 profi le and anti-TB drug-induced 9 Acharya S K, Dasarathy S, Kumer T L, et al. Fulminant hepati- hepatotoxicity in patients from southern Brazil. Eur J Clin tis in a tropical population: clinical course, cause, and early Pharmacol 2008; 64: 673–681. predictors of outcome. Hepatology 1996; 23: 1448–1455. 29 Sun F, Chen Y, Xiang Y, Zhan S. Drug-metabolising enzyme 10 Wong W M, Wu P C, Yuen M F, et al. Anti-tuberculosis drug- polymorphisms and predisposition to anti-tuberculosis drug- related liver dysfunction in chronic hepatitis B infection. Hepa- induced liver injury: a meta-analysis. Int J Tuberc Lung Dis tology 2000; 31: 201–206. 2008; 12: 994–1002. 11 Dossing M, Wilcke J T, Askgaard D S, et al. Liver injury during 30 Lee S-W, Chung L S-C, Huang H-H, et al. NAT2 and CYP2E1 anti-tuberculosis treatment: an 11-year study. Tubercle Lung polymorphisms and susceptibility to fi rst-line anti-tuberculosis Dis 1996; 77: 335–340. drug-induced hepatitis. Int J Tuberc Lung Dis 2010; 14: 622– 12 Devoto F M, Gonzalez C, Iannantuono R, et al. Risk factors 626. for hepatotoxicity induced by anti-tuberculosis drugs. Acta 31 Khalili H, Fouladdel S, Sistanizad M, Hajiabdolbaghi M, Azizi Physiol Pharmacol Ther Latinoam 1997; 47: 197–202. E. Association of N-acetyltransferase-2 genotypes and anti- 13 Ungo J R, Jones D, Ashkin D, et al. Anti-tuberculosis drug- t uberculosis induced liver injury; fi rst case-controlled study i nduced hepatotoxicity. The role of hepatitis C virus and the from Iran. Curr Drug Saf 2011; 6: 17–22. human immunodefiiency virus. Am J Respir Crit Care Med 32 Bose P D, Sarma M P, Medhi S, et al. Role of polymorphic 1998; 157 (6 Pt 1): 1871–1876. N-acetyltransferase 2 and cytochrome P4502E1 gene in 14 Jasmer R M, Saukkonen J J, Blumberg H M, et al. Short-course anti-tuberculosis treatment-induced hepatitis. J Gastroenterol rifampin and pyrazinamide compared with isoniazid for latent Hepatol 2011; 26: 312–318. tuberculosis infection: a multicenter clinical trial. Ann Intern 33 Yamada S, Tang M, Richardson K, et al. Genetic variations of Med 2002; 137: 640–647. NAT2 and CYP2E1 and isoniazid hepatotoxicity in a diverse 15 Singh J, Arora A, Garg P K. Anti-tuberculosis treatment-induced population. Pharmacogenomics 2009; 10: 1433–1445. hepatotoxicity: role of predictive factors. Postgrad Med J 34 Kim S H, Kim S H, Bahn J W, et al. Genetic polymorphisms of 1995; 71: 359–362. drug-metabolizing enzymes and anti-TB drug-induced hepati- 16 Shakya R, Rao B S, Shrestha B. Incidence of hepatotoxicity due tis. Pharmacogenomics 2009; 10: 1767–1779. NAT2 gene relative to drug-induced liver injury 595

35 An H R, Wu X Q, Wang Z Y, et al. The associations of poly- 41 Little J, Bradley L, Bray M S, et al. Reporting, appraising, and morphism of N-acetyltransferase 2 gene is associated with integrating data on genotype prevalence and gene-disease as- anti-tuberculosis drug-induced hepatotoxicity in tuberculosis sociations. Am J Epidemiol 2002; 156: 300–310. patients. Zhonghua Yu Fang Yi Xue Za Zhi 2011; 45: 36–40. 42 Benichou C. Criteria of drug-induced liver disorders: report of 36 Sotsuka T, Sasaki Y, Hirai S, et al. Association of isoniazid- an International Consensus Meeting. J Hepatol 1990; 11: 272– m etabolizing enzyme genotypes and isoniazid-induced hepato- 276. toxicity in tuberculosis patients. In Vivo 2011; 25: 803–812. 43 Higgins J P, Thompson S G, Deeks J J, et al. Measuring incon- 37 Smith C A, Smith G, Wolf C R. Genetic polymorphisms in sistency in meta-analyses. BMJ 2003; 327: 557–560. x enobiotic metabolism. Eur J Cancer 1994; 30A: 1921–1935. 44 Begg C B, Mazumdar M. Operating characteristics of a rank 38 Hein D W. Molecular genetics and function of NAT1 and correlation test for publication bias. Biometrics 1994; 50: 1088– NAT2: role in aromatic amine metabolism and carcinogenesis. 1101. Mutat Res 2002; 506–507: 65–77. 45 Egger M, Davey S G, Schneider M, et al. Bias in meta-analysis 39 Mitchell J R, Thorgeirsson U P, Black M, et al. Increased detected by a simple, graphical test. BMJ 1997; 315: 629–634. i ncidence of isoniazid hepatitis in rapid acetylators: possible 46 García-Closas M, Malats N, Silverman D, et al. NAT2 slow relation to hydranize metabolites. Clin Pharmacol Ther 1975; acetylation and GSTM1 null genotypes increase bladder can- 18: 70–79. cer risk: results from the Spanish Bladder Cancer Study and 40 Yamamoto T, Suou T, Hirayama C. Elevated serum amino- meta-analyses. Lancet 2005; 366: 649–659. transferase induced by isoniazid in relation to isoniazid acetylator phenotype. Hepatology 1986; 6: 295–298. NAT2 gene relative to drug-induced liver injury i

APPENDIX

Table A.1 Main findings of each study included in the meta-analysis

First author, Sample No. reference Year Country size Main findings 1 Ohno17 2000 Japan 77 Compared with rapid acetylators, the relative risk was 4.0 (95%CI 1.94–6.06) for intermediate and 28.0 (95%CI 26.0–30.0) for slow acetylators 2 Huang19 2002 Taiwan 224 Slow acetylators had significantly higher serum aminotransferase levels than rapid acetylators. Logistic regression showed that slow acetylator status (OR 3.66, 95%CI 1.58–8.49, P = 0.003) was an independent risk factor for anti- tuberculosis drug-induced hepatitis. It was adjusted for age and body mass index 3 Vuilleumier24 2006 Switzerland 89 No significant influence of NAT2 polymorphism 4 Shimizu23 2006 Asian 42 Slow acetylator patients had a significantly higher risk of INH-induced hepatotoxicity than rapid acetylator patients (P < 0.05) 5 Cho25 2007 Korean 132 Regarding NAT2, slow acetylators had a higher incidence of hepatotoxicity than rapid acetylators (36.8% vs. 9.7%, P = 0.005) and there was a 3.8-fold risk of hepatotoxicity for slow acetylators compared to rapid acetylators 6 Higuchi26 2007 Japan 100 Statistical analysis revealed that the frequency of a variant haplotype, NAT2*6A, was significantly increased in TB patients with hepatotoxicity compared with those without (P = 0.001, OR 3.535) 7 Bozok27 2008 Turkey 100 The difference was statistically significant when the control and study groups were compared for their acetylator status. The proportion of slow acetylators was much higher in the study group 8 Lee30 2010 Taiwan 140 Slow acetylators defined by NAT2 genotypes had a higher risk of hepatotoxicity than rapid acetylators (OR 3.28, 95%CI 1.53–7.06) 9 Bose32 2011 India 218 There was a higher prevalence of slow NAT2 acetylator genotypes in DIH (70.73%) compared to non-DIH (44.63%, P < 0.05) 10 Kim34 2009 Korea 226 Genetic variants in the promoter and exons of NAT2 increase the risk of anti- tuberculosis DIH by modifying acetylation phenotypes and/or gene expression of NAT2 11 Yamada33 2009 USA 170 There was evidence of a trend for increasing risk of hepatotoxicity across the rapid, intermediate and slow acetylator groups (P = 0.08) 12 An35 2011 China 208 Patients with slow acetylator genotype (OR 4.74, 95%CI 2.42–9.28) had a significantly higher risk of anti-tuberculosis drug-induced hepatotoxicity than those with rapid or intermediate acetylator genotypes 13 Khalili31 2011 Iran 50 Among Iranian tuberculosis patients, anti-tuberculosis-induced hepatotoxicity was more frequent in slow acetylators than fast acetylators 14 Sotsuka36 2011 Japan 144 The risk of side-effects, such as hepatic disorder, may rise in patients with a slow acetylator phenotype

CI = confidence interval; OR = odds ratio; INH = isoniazid; DIH = drug-induced hepatitis.

Table A.2 Meta-analysis results of NAT2 genotype and ATLI

Genotype n OR* 95%CI* I2, %* OR† 95%CI† I2, %† Model NAT2*4/4 9 0.502 0.358–0.704 0.00 1.000 — — Fixed NAT2*4/5 6 0.730 0.487–1.095 0.00 1.371 0.702–2.678 0.00 Fixed NAT2*4/6 8 0.702 0.358–1.380 57.20 1.090 0.474–2.503 52.50 Random NAT2*4/7 9 0.628 0.414–0.953 0.00 1.059 0.644–1.741 0.00 Fixed NAT2*5/7 7 3.473 1.671–7.218 0.00 4.988 1.696–14.671 0.00 Fixed NAT2*6/6 9 2.910 1.936–4.374 2.10 4.633 2.713–7.911 0.00 Fixed NAT2*6/7 8 4.350 2.716–6.966 0.00 5.729 3.211–10.220 0.00 Fixed NAT2*7/7 8 2.660 1.340–5.279 34.10 3.004 1.387–6.507 44.70 Fixed

* Comparing each genotype with all the other types. † Comparing each genotype with the same wild type (NAT2*4/*4). ATLI = anti-tuberculosis drug-induced liver injury; OR = odds ratio; CI = confidence interval. ii The International Journal of Tuberculosis and Lung Disease

RÉSUMÉ

CADRE : Bien que toute une série d’études aient évalué On a observé une association significative entre le fait l’association potentielle entre le polymorphisme de la d’être acétyleur lent de NAT2 et le risque d’ATLI. L’OR N-acétyltransférase 2 (NAT2) et le risque de complica- des acétyleurs lents de NAT2 comparés avec les acé- tions hépatiques dues aux médicaments antituberculeux tyleurs rapides est de 4,697 (IC95% 3,291–6,705 ; P < (ALTI), les résultats sont généralement controversés et 0,001). Les analyses des sous-groupes indiquent que inadéquats, principalement en raison des limitations du tant les cas asiatiques que non-asiatiques acétyleurs lents pouvoir statistique. Cette méta-analyse a été menée afin développent plus fréquemment une ATLI et sont sem- de résoudre ce problème. blables aux patients acétyleurs lents traités par une com- SCHÉMA : On a fait des recherches sur PubMed, Em- binaison de première ligne. Lorsqu’on compare les acé- base et Web of Science en utilisant les mots-clés suivants : tyleurs intermédiaires de NAT2 avec les acétyleurs « N-acétyltransférase 2 » ou « NAT2 » et « polymor- rapides, l’OR de ATLI est de 1,261 (IC95% 0,928– phisme » et « tuberculose » ou « TB » et « hépatotoxi- 1,712 ; P = 0,138). cité » ou « atteinte hépatique ». On a combiné les odds CONCLUSIONS : Cette méta-analyse a démontré que ratio (OR) bruts et les intervalles de confiance (IC) à chez les patients TB acétyleurs lents, le risque d’ATLI 95% dans des schémas de Forest et on les a exposés est plus élevé que chez les autres acétyleurs. Le dépistage dans des tableaux. des patients atteints d’un polymorphisme génétique de RÉSULTATS : On a pu inclure dans la méta-analyse au NAT2 sera utile pour la prédiction et la prévention cli- total 14 études comportant 474 cas et 1446 contrôles. nique de l’ATLI.

RESUMEN

OBJETIVO: Aunque en diversos estudios se ha evaluado lentos de la NAT2 y el riesgo de ATLI. El OR de los ace- la posible asociación entre los polimorfismos de la tiladores lentos de la NAT2 fue 4,697 en comparación N-acetiltransferasa 2 (NAT2) y el riesgo hepatotoxicidad con los acetiladores rápidos (IC95% 3,291 a 6,705; P < inducida por los medicamentos antituberculosos (ATLI), 0,001). El análisis por subgrupos indicó que los grupos los resultados suelen ser controvertidos e inadecuados, de asiáticos y los grupos no asiáticos con acetiladores sobre todo por su insuficiente potencia estadística. El lentos sufrían con mayor frecuencia de ATLI, en igual presente metanálisis se realizó con el propósito de su- medida que los grupos con acetiladores lentos que re- perar esta deficiencia. cibían un tratamiento con asociaciones a base de medi- MÉTODO: Se llevó a cabo una búsqueda en las bases de camentos de primera línea. Cuando se compararon los datos PubMed, Embase y la Web of Science con las si- acetiladores intermedios con los acetiladores rápidos de guientes palabras clave: ‘N-acetyltransferase 2’ o ‘NAT2’ la NAT2, el OR de ATLI fue 1,261 (IC95% 0,928 a y ‘polymorphism’ y ‘tuberculosis’ o ‘TB’ y ‘hepatotoxi- 1,712; P = 0,138). city’ o ‘liver injury’. Los cocientes de posibilidades brutos CONCLUSIÓN: Con este metanálisis se puso de mani- (OR) con intervalos de confianza del 95% (IC95%) se fiesto que los pacientes tuberculosos que son acetila- resumieron en diagramas de bosque y se detallaron en dores lentos de la NAT2 se encuentran en mayor riesgo cuadros. que los demás tipos de acetiladores de presentar una RESULTADOS: Se incluyeron en el metanálisis 14 estu- ATLI. La detección sistemática de los polimorfismos dios que comprendían 474 casos y 1446 testigos. Se ob- genéticos de la NAT2 en los pacientes será útil en la pre- servó una correlación significativa entre los acetiladores dicción y la prevención de ATLI.