Epidemiology of Isoniazid Resistance Mutations and Their Effect on Tuberculosis Treatment Outcomes

Abstract

Isoniazid resistance is highly prevalent in Vietnam. We investigated the molecular and epidemiological characteristics and the association with first-line treatment outcomes of the main isoniazid resistance mutations in Mycobacterium tuberculosis in codon 315 of the katG and in the promoter region of the inhA gene. Mycobacterium tuberculosis strains with phenotypic resistance to isoniazid from consecutively diagnosed smear-positive tuberculosis patients in rural Vietnam were subjected to Genotype MTBDRplus testing to identify katG and inhA mutations. Treatment failure and relapse were determined by sputum culture. In total, 227 of 251 isoniazid-resistant strains (90.4%) had detectable mutations: 75.3% in katG codon 315 (katG₃₁₅) and 28.2% in the inhA promoter region. katG₃₁₅ mutations were significantly associated with pretreatment resistance to streptomycin, rifampin, and ethambutol but not with the Beijing genotype and predicted both unfavorable treatment outcome (treatment failure or death) and relapse; inhA promoter region mutations were only associated with resistance to streptomycin and relapse. In tuberculosis patients, M. tuberculosis katG₃₁₅ mutations but not inhA mutations are associated with unfavorable treatment outcome. inhA mutations do, however, increase the risk of relapse, at least with treatment regimens that contain only isoniazid and ethambutol in the continuation phase.

INTRODUCTION

With 8.8 million cases notified and 1.4 million deaths in 2010, tuberculosis (TB) remains a major burden to global health (1). In addition to rifampin, isoniazid is an important drug in first-line anti-TB treatment (2). Mycobacterium tuberculosis strains resistant to at least both rifampin and isoniazid are referred to as multidrug resistant (MDR). Both multidrug resistance and resistance to isoniazid without concomitant rifampin resistance are associated with poor response to first-line treatment (3, 4). Whereas rifampin resistance is usually encoded in a part of the rpoB gene, the mechanism of resistance to isoniazid is more complex, with mutations conferring resistance in several genomic loci, such as katG, inhA, ahpC, and, potentially, ndh (5–8). Mutations in codon 315 of the katG gene (katG₃₁₅) and in the promoter region of the inhA gene are by far the most common. katG₃₁₅ mutations occur in 50 to 95% of isoniazid-resistant strains (6, 9, 10), whereas 20 to 42% of such strains have mutations in the promoter region of the inhA gene (6, 10, 11), depending on the geographic region studied.

Isoniazid is activated by the enzyme catalase peroxidase, encoded by katG (12). katG mutations lead to high-level isoniazid resistance (to ≥1.0 μg/ml in 7H10 agar) (13). The inhA gene encodes an enoyl acyl carrier protein reductase involved in fatty acid synthesis. These fatty acids are the target of the active derivative of isoniazid. inhA mutations usually lead to low-level isoniazid resistance (resistant to 0.2 μg/ml in 7H10 agar) (13, 14).

katG₃₁₅ mutations have been shown to be associated with MDR-TB and TB transmission (15, 16). Such mutations were more frequent among patients infected with Beijing genotype strains (17, 18), which are common in East Asia, including Vietnam (19, 20), and related to drug resistance (20, 21), as well as to relapse in various areas (22, 23). These differences in isoniazid resistance-conferring mutations may also be related to other characteristics of M. tuberculosis strains, as well as to treatment outcomes. There are, however, very few studies on the mutations underlying resistance to anti-TB drugs and treatment outcome. We therefore studied the epidemiology of katG and inhA mutations in M. tuberculosis isolates and the clinical characteristics of the respective patients in Vietnam, where the prevalence of smear-positive TB was 197/100,000 in 2006-2007 (24) and resistance to isoniazid is common (16 to 25% in new patients) (25). For this we assessed M. tuberculosis genotype and TB treatment outcomes in association with katG and inhA mutations in a prospective, population-based study.

MATERIALS AND METHODS

Study subjects.

The study area consisted of three adjacent rural districts in Tiengiang Province in the Mekong River Delta in Southern Vietnam. Details of the study have been described elsewhere (26).

From 1 July 2005 to 30 June 2007, all patients aged ≥15 years, resident in the study area and registered for treatment of smear-positive pulmonary TB, were eligible for inclusion after provision of written informed consent. Excluded were patients who received treatment for more than 2 weeks before registration. Ethical clearance for the study was obtained from the ethical health committee of the Ho Chi Minh City Council.

According to the guidelines of the Vietnam National TB Control Program (27) patients with no history of treatment with anti-TB drugs for >1 month (i.e., new TB cases) were treated with 2 months of daily streptomycin (S), isoniazid (H), rifampin (R), and pyrazinamide (Z), followed by daily ethambutol (E) and isoniazid for 6 months (2SHRZ/6EH). Previously treated patients were given all five drugs (SHRZE) daily for 2 months and then four drugs (HRZE) for one more month, followed by RHE 3 days per week for 5 months. All doses were given under directly observed treatment as long as patients were given rifampin, irrespective of treatment phase or regimen. Drug susceptibility testing was done later, and results were not used to modify treatment regimens. Treatment adherence was confirmed from the treatment cards kept at the district tuberculosis unit (DTU).

Study design.

The purpose of the present study was to quantify possible associations between isoniazid resistance-conferring mutations in the M. tuberculosis strain isolated before treatment and M. tuberculosis genotype and patient characteristics among all TB patients, and between pretreatment isoniazid resistance-conferring mutations and treatment outcomes (treatment failure and relapse) among new TB cases. New TB patients were followed up during standard first-line treatment with sputum smear microscopy at months 3, 5, and 8 and with sputum culture at the end of treatment (after 8 months, or after 5 months if the month 5 smear was positive). Participants whose sputum smear and culture were negative for M. tuberculosis at the end of treatment were visited by study staff twice thereafter, at around 9 and 18 months after treatment completion, or later if not encountered. In addition, data were collected during this period on study participants reporting with TB symptoms at any of the study clinics. Participants who had any complaints suggesting recurrent TB during these visits, or when they themselves consulted a participating clinic, provided two sputum specimens for smear and culture. The data were also collected on any intermediate TB treatment elsewhere, and on causes of death among the study patients based on clinic reports, death certificates and interviews with family members.

Laboratory methods.

Sputum specimens were kept refrigerated and transported to Pham Ngoc Thach Hospital in Ho Chi Minh City within 72 h after collection. They were decontaminated and liquefied with 1% N-acetylcystine–2% NaOH, inoculated on modified Ogawa medium, and incubated at 37°C. Cultures were examined for growth after 1, 2, 4, 6, and 8 weeks of incubation. Cultures with no growth after 8 weeks were reported as negative. M. tuberculosis was identified using the niacin and the nitrate tests.

Drug susceptibility testing (DST) was performed using the proportion method on Löwenstein-Jensen (LJ) medium (28). Criteria for drug resistance were ≥1% of the CFU grown at 28 or 42 days compared to the drug-free control medium at the following drug concentrations: isoniazid, 0.2 μg/ml; rifampin, 40 μg/ml; streptomycin, 4 μg/ml; and ethambutol, 2 μg/ml (28). All isoniazid-resistant M. tuberculosis strains that were isolated were subjected to testing by GenoType MTBDRplus that combines detection of M. tuberculosis complex with detection of mutations in the 81-bp hot spot region of rpoB, at codon 315 of the katG gene and in the inhA promoter region (15). All baseline and follow-up isolates from patients with positive follow-up M. tuberculosis cultures were subjected to molecular typing by spoligo and variable number of tandem repeats (VNTR) typing. Bacterial DNA was extracted from positive cultures using an earlier described method (29). Spoligotyping was performed according to the internationally standardized method (30), and VNTR typing was done using15 loci (31).

Definitions.

Previously treated patients were those who received 1 month or more of anti-TB drugs in the past. Cure was defined as a negative sputum smear examination and culture in the last month of treatment and on at least one previous occasion, and treatment failure was defined as any positive sputum smear or culture at 5 months or later during treatment. Treatment completion was defined as having completed treatment without meeting the criteria for being classified as cure or failure.

Recurrent TB was defined as any case of positive smear and/or culture during the follow-up period among the cured patients (32). We defined a case of recurrent TB as relapse if the initial and follow-up M. tuberculosis isolates had identical spoligotypes and VNTR patterns, or if the VNTR patterns differed by ≤1 locus, and as reinfection if otherwise (31). Unfavorable treatment outcome was referred to as treatment failure or death and related to the treatment period only.

Genotypes were based on spoligotyping. The Beijing genotype was defined as any isolate without direct-repeat spacers 1 to 34 and the presence of at least three of the spacers from 35 to 43 (33). Other genotypes, including the East African-Indian (EAI) genotype that is predominant in Vietnam, were defined as described by Brudey et al. (34).

Data analysis.

Data were double entered in Epi-Info (version 6.04; Centers for Disease Control and Prevention, Atlanta, GA); discrepancies were corrected based on the raw data. Analyses were performed in Stata (version 10SE; Stata Corp., College Station, TX).

For comparison of categorical variables we used the chi-squared and two-sided Fisher exact tests as appropriate. Associations of katG₃₁₅ or inhA mutations with explanatory variables before start of treatment were expressed as odds ratios (ORs); confounding effects were investigated by multivariable logistic regression modeling. In the analysis of treatment failure, mutations in katG and inhA were assessed by multivariable logistic regression as explanatory variables, along with covariates that showed confounding effects, potentially including age, sex, residence, resistance to other drugs, the M. tuberculosis genotype, pretreatment smear grading and the extent of chest X-ray abnormalities, and treatment adherence. Only variables that showed confounding effects for the association between resistance mutations and the outcome were retained in the final model. Since patients who died during treatment may reflect treatment failures, we repeated this analysis taking failure or death as unfavorable treatment outcome. For the association with relapse, we did a similar analysis using multivariable Cox' proportional hazard modeling. P values for contribution to multivariate models, including interaction, were based on the likelihood ratio test. All tests were done at the 5% significance level.

RESULTS

After excluding 151 patients (Fig. 1), pretreatment data were available for analysis for 1,213 (88.9%) of 1,364 registered patients. Of these, 924 were male (76.2%); the mean age was 50 years (standard deviation [SD] = 18.3; range, 15 to 102). There were 1,102 (90.9%) new patients and 111 (9.1%) patients previously treated for TB.

Fig 1.

Schematic presentation of enrollment of the study population. TB, tuberculosis.

Of 1,213 M. tuberculosis pretreatment isolates, 69 (5.7%) were monoresistant to isoniazid, 146 (12.4%) were monoresistant to streptomycin, 128 (10.6%) were resistant to isoniazid and streptomycin, and 47 (3.9%) were multidrug resistant (see Table S1 in the supplemental material). Monoresistance to isoniazid was more frequent among previously treated patients than among new TB patients (12.6% versus 5.0%, P < 0.05), whereas the proportion of other monoresistance patterns did not significantly differ between previously treated and new patients.

Isoniazid resistance-conferring mutations.

Of the 251 (20.7%) phenotypically isoniazid-resistant M. tuberculosis strains, 227 (90.4%) exhibited mutations by GenoType MTBDRplus testing; 171 (75.3%) had mutations or no reaction on wild-type (WT) probes in katG₃₁₅, including 167 (97.7%) with katG S315T1 mutations. Sixty-four (28.2%) had mutations in the inhA promoter region, including 61 (95.3%) involving inhA C15T mutations. Only 8 of 227 (3.5%) strains with a katG₃₁₅ mutation had an additional mutation in the inhA promoter region (Table 1).

Table 1.

Results of isoniazid resistance mutations detected by MTBDRplus test among 227 TB patients with phenotypic resistance to isoniazid in Vietnam

Mutation(s)		Frequency
katG	inhA	No. of patients	% Total
WT (315) absent		171	75.3
MUT1 (S315T1)		167	73.6
MUT2 (S315T2)		0	0
	WT (−15/−16) absent	56	24.7
	WT (−8) absent	5	2.2
	MUT1 (C15T)	61	26.9
	MUT2 (A16G)	0	0
	MUT3A (T8C)	0	0
	MUT3B (T8A)	3	1.3
MUT1 (S315T1)	MUT1 (C15T)	8	3.5

Characteristics for katG₃₁₅ mutations.

There were no significant associations between the probability of having a strain with a katG₃₁₅ mutation and the patient's district, type of residence, age or presence of mutations in the inhA promoter region. However, katG₃₁₅ mutations were significantly more frequent among women (odds ratio [OR] 1.4), among patients previously treated for TB (OR 5.3), among strains that were resistant to rifampin (OR 27.7), streptomycin (OR 16.9), or ethambutol (OR 62.1), and among strains that belonged to the Beijing genotype (OR 3.2).

In a multivariable model katG₃₁₅ mutations remained associated (adjusted OR [OR^adj]; 95% confidence interval [CI]) with previous TB treatment (2.6; 1.4 to 4.8), resistance to rifampin (4.5; 1.8 to 11.1), ethambutol (12.0; 2.7 to 54.4) or streptomycin (15.0; 9.4 to 23.9), as well as with female sex (1.9; 1.2 to 3.0), but not with the Beijing genotype (1.3; 0.8 to 2.2) or EAI genotype (1.7; 0.9 to 3.0) compared to all other genotypes together (Table 2). When leaving resistance to other drugs than isoniazid out of the model, the Beijing genotype (2.6; 1.7 to 4.1) but not the EAI genotype was associated with katG₃₁₅ mutation. When adding resistance to only one of the three other drugs to the model, the Beijing genotype was still significantly associated with katG₃₁₅ mutations after adjustment for rifampin (2.2; 1.4 to 3.4) or ethambutol resistance (2.5; 1.6 to 4.0), whereas this association disappeared after adjustment for streptomycin resistance (1.5; 0.9 to 2.5).

Table 2.

Univariable and multivariable associations with patient characteristics, genotype, and anti-TB drug resistance for katG codon 315 mutations at the start of treatment (baseline)

Characteristic or parameter	No. of patients	katG mutations		OR (95% CI)a		P
		No.	%	Crude	Adjusted
Total no. of subjects	1,213
Sex						0.009
Male	924	120	13	1	1
Female	289	51	17.7	1.4 (1.0–2.05)	1.9 (1.2–3.0)
Age (yrs)						0.349
<25	119	16	13.5	1	1
25 to 49	528	79	15	1.1 (0.6–2.0)	1.7 (0.8–3.7)
≥50	566	76	13.4	1.0 (0.6–1.8)	1.7 (0.8–3.7)
History of TB						0.002
New	1,102	126	11.4	1	1
Previously treated	111	45	40.5	5.3 (3.4–8.2)	2.6 (1.4–4.8)
District
Cailay	440	60	13.6	1
Caibe	418	57	13.6	1.0 (0.7–1.5)
Chauthanh	355	54	15.2	1.1 (0.8–1.7)
Residence
On waterway	457	67	14.7	1
On provincial road	624	91	14.6	1.0 (0.7–1.4)
On national road	132	13	9.9	0.6 (0.3–1.2)
Genotype family						0.203
East African-Indian	461	40	8.7	0.9 (0.6–1.5)	1.7 (0.9–3.0)
Beijing	407	98	24.1	3.0 (2.0–4.6)	1.3 (0.8–2.2)
Other	345	33	9.6	1	1
Rifampin resistance						<0.001
No	1,163	132	11.4	1	1
Yes	50	39	78	27.7 (13.1–58.6)	4.5 (1.8–11.1)
Streptomycin resistance						<0.001
No	889	36	4.1	1	1
Yes	324	135	41.7	16.9 (10.7–26.7)	15.0 (9.4–23.9)
Ethambutol resistance						<0.001
No	1,184	145	12.3	1	1
Yes	29	26	89.7	62.1 (17.2–224.4)	12.0 (2.7–54.4)
InhA mutation						<0.001
No	1149	163	14.2	1	1
Yes	64	8	12.5	0.9 (0.4–1.9)	0.2 (0.1–0.5)

^a“Adjusted” means adjusted for all other variables in the model. CI, confidence interval; OR, odds ratio.

Characteristics for inhA promoter region mutations.

In univariate analysis, inhA mutations were associated with previous TB treatment (OR 2.7), with resistance to rifampin (OR 3.2), streptomycin (OR 3.1), or ethambutol (OR 4.0) and with living in one of the districts (Caibe, OR 2.4), but not with sex, age, residence, or genotype. In multivariate analysis also including resistance to other drugs, inhA mutations remained significantly associated with resistance to streptomycin (OR^adj 4.4; 95% CI = 2.4 to 8.1) and previous TB treatment (OR^adj 2.5; 95% CI = 1.2 to 5.4), and near-significantly associated with the Caibe district (OR^adj 2.1; P = 0.066) but not with resistance to rifampin or ethambutol (Table 3). There was significant interaction between previous TB treatment and streptomycin resistance: inhA mutations were significantly associated with streptomycin resistance among new TB patients (OR^adj 6.5, P < 0.001) but not among previously treated patients (OR^adj 0.36, P > 0.05).

Table 3.

Univariable and multivariable associations with patient characteristics, genotype, and anti-TB drug resistance for the inhA promoter region mutations at the start of treatment (baseline)

Characteristic or parameter	No. of patients	inhA mutations		OR (95% CI)a		P
		No.	%	Crude	Adjusted
Total	1,213
Sex						0.184
Male	924	46	5	1	1
Female	289	18	6.2	1.3 (0.7–2.2)	1.5 (0.8–2.7)
Age group (yrs)						0.453
<25	119	4	3.4	1	1
25 to 49	528	30	5.7	1.7 (0.6–5.0)	1.9 (0.6–5.9)
≥50	566	30	5.3	1.6 (0.6–4.7)	1.7 (0.6–5.1)
History of TB						0.025
New	1,102	51	4.6	1	1
Previously treated	111	12	11.7	2.7 (1.4–5.2)	2.5 (1.2–5.4)
District						0.066
Cailay	440	14	3.2	1	1
Caibe	418	31	7.4	2.4 (1.3–4.7)	2.1 (1.1–4.1)
Chauthanh	355	19	3.4	1.7 (0.9–3.5)	1.3 (0.6–2.8)
Residence
On waterway	457	25	5.5	1
On provincial road	624	32	5.1	0.9 (0.6–1.6)
On national road	132	7	5.3	1.0 (0.4–2.3)
Genotype family						0.276
East African-Indian	461	21	4.6	0.7 (0.4–1.4)	1.0 (0.5–2.0)
Beijing	407	22	5.4	0.9 (0.5–1.6)	0.6 (0.3–1.2)
Other	345	21	6.1	1	1
Rifampin resistance						0.496
No	1,163	57	4.9	1	1
Yes	50	7	14	3.2 (1.4–7.4)	1.6 (0.5–5.4)
Streptomycin resistance						<0.001
No	889	31	3.5	1	1
Yes	324	33	10.2	3.1 (1.9–5.2)	4.4 (2.4–8.1)
Ethambutol resistance						0.122
No	1,184	59	5	1	1
Yes	29	5	17.2	4.0 (1.5–10.8)	3.3 (0.8–14.3)
katG mutation						<0.001
No	1,042	56	5.4	1	1
Yes	171	8	4.7	0.9 (0.4–1.8)	0.2 (0.1–0.6)

^a“Adjusted” means adjusted for all other variables in the model.

Predictors of treatment failure.

Of 1,102 new TB patients, 51were excluded due to loss of data (30), reinfection (7), defaulting (9), transfer-out (3), and changed treatment regimen because of side effects (2). Furthermore, we excluded 41 patients who died during treatment, leaving 1,010 new patients for this analysis (Fig. 1). Of these, 21 (2.1%) had a treatment failure (Table 4).

Table 4.

Multivariable associations for treatment failure or treatment failure and death during treatment combined

Resistance	Failurea					Failure or deathb
	Total	No.	%	Adjusted OR (95% CI)	P	Total	No.	%	Adjusted OR (95% CI)	P
Isoniazid resistance					0.370					0.046
Susceptible	834	7	0.8	1		866	39	4.5	1
Any katG mutation	116	12	10.3	3.2 (0.8–12.8)	0.102	122	18	14.8	3.0 (1.4–6.8)	0.007
InhA mutations only	42	1	2.4	1.0 (0.1–10.0)	0.974	43	2	4.7	1 (0.2–4.5)	0.994
No mutations	18	1	5.6	1.8 (0.1–24.1)	0.657	20	3	15.0	3.4 (0.9–13.4)	0.081
Rifampin resistance					0.001					<0.001
No	986	12	1.2	1		1027	53	5.2	1
Yes	24	9	37.5	7.6 (1.7–34.1)		24	9	37.5	6.3 (2.3–17.1)
Streptomycin resistance					0.191					0.348
No	760	7	0.9	1		794	41	5.2	1
Yes	250	14	5.6	2.3 (0.7–8.2)		257	21	8.2	0.7 (0.3–1.5)
Ethambutol resistance					0.056
No	997	15	1.5	1		1038	56	5.4	–
Yes	13	6	46.2	5.9 (1.0–32.3)		13	6	46.2

^aThe model for the OR and P values includes the following covariates: district, isoniazid resistance, rifampin resistance, streptomycin resistance, and ethambutol resistance.

^bThe model for the OR and P values includes the following covariates: isoniazid resistance, rifampin resistance, and streptomycin resistance.

In univariate analysis, the risk of treatment failure was significantly increased for isoniazid-resistant strains having at least a katG₃₁₅ mutation (OR 13.6; 95% CI = 5.3 to 35.4) but not for isoniazid-resistant strains having inhA mutations only (OR 2.9; 95% CI = 0.3 to 14.0) or no mutations (OR 6.9; 95% CI = 0.8 to 59.6). After multivariable adjustment for district and resistance to rifampin, streptomycin, or ethambutol, the association between katG₃₁₅ mutations and treatment failure (OR^adj; 95% CI) was no longer significant (3.2; 0.8 to 12.8; Wald test, P = 0.102). Similarly, neither inhA mutations only (1.0; 0.1 to 10.1) nor isoniazid resistance without mutations detectable by the MTBDRplus assay (1.8; 0.1 to 24.1) showed significant association with treatment failure (Table 4). There was no significant difference in failure between patients with katG₃₁₅ and patients with inhA mutated strains. Isoniazid resistance mutations showed no association with treatment adherence, pretreatment smear grade, and the extent of pretreatment abnormalities, and none of these variables confounded the observed associations between resistance mutation and treatment failure.

When the 41 patients who had died during treatment were included in the analysis and failure or death (62 of 1,051 patients; 5.9%) was combined as unfavorable treatment outcome, its risk was significantly increased for isoniazid-resistant strains having at least a katG₃₁₅ mutation (OR 3.7; 95% CI = 2.0 to 6.7) or no mutations detectable by the MTBDRplus assay (OR 3.7; 95% CI = 1.1 to 13.3) but not for inhA mutations only (OR 1.0; 95% CI = 0.2 to 4.4). In a multivariate model adjusting for covariates that appeared to confound this association (resistance to rifampin or resistance to streptomycin) unfavorable treatment outcome remained significantly associated with the katG₃₁₅ mutations (OR^adj 3.0; 95% CI = 1.4 to 6.8; Wald test P = 0.007) but not with no detectable mutations (OR^adj 3.4; 95% CI = 0.9 to 13.4; P = 0.081) (Table 4).

Predictors of relapse.

For this analysis, we included all 984 new patients who were smear and culture negative at the end of treatment and available for follow-up (Fig. 1). We observed 31 cases of recurrent TB, of which 9 were classified as reinfections and 22 (2.2%) were classified as relapse. There were no relapses among the 17 participants with isoniazid-resistant isolates that did not display any mutation in the MTBDRplus assay. The three strains that displayed both a katG₃₁₅ and an inhA mutation were included in the katG₃₁₅ mutation category.

In univariate analysis, both katG₃₁₅ and inhA mutations were strongly associated with relapse; hazard ratios (HR) were 6.7 (95% CI = 2.6 to 16.9) and 8.3 (95% CI = 2.6 to 26.4), respectively (Fig. 2). Relapse was also significantly more frequent among participants harboring strains that were of the Beijing genotype (HR 6.2) or streptomycin resistant (HR 4.0) and among those with MDR-TB (HR 7.4).

Fig 2.

Inverted survival curve for tuberculosis relapse cases among 984 patients after first line TB treatment, isoniazid resistance-conferring mutations. Log-rank test, P < 0.001. Solid black line, cases due to katG codon 315 mutated strains. Interrupted black line, cases due to inhA promoter region-only mutated strains. Gray line, cases due to isoniazid-susceptible strains. y axis, proportion of relapse cases. INH, isoniazid.

After multivariable adjustment for genotype and resistance to streptomycin or rifampin, relapse remained strongly associated with katG₃₁₅ mutations (HR^adj 4.3; 95% CI = 1.4 to 13.6, P = 0.013) and inhA mutations (HR^adj 8.7; 95% CI = 2.5 to 30.0, P = 0.001) compared against isoniazid-susceptible strains (Table 5). The relapse rate did not differ significantly between strains with katG₃₁₅ mutations, and strains with inhA mutations. We found no significant interactions.

Table 5.

Multivariable associations for relapse in 967 new TB patients^a

Resistance	No. of relapses	Incidence per 100 person-years of follow-up	Adjusted hazard ratio	95% CI	P
Total	22
Isoniazid resistance					0.004
Susceptible	10	0.78	1
Any katG mutations	8	5.14	4.3	1.4–13.6	0.013
inhA mutations only	4	6.34	8.7	2.5–30	0.001
Rifampin resistance	2	9.21	1.2	0.3–5.7	0.821
Streptomycin resistance	12	3.5	1.1	0.4–3.1	0.900
Genotype					<0.001
Beijing		3.46	5.1	1.9–14
Non-Beijing		0.58	1

^aVariables in the model: isoniazid resistance mutations, Beijing genotype, streptomycin resistance, and rifampin resistance. A total of 17 patients with isoniazid-resistant strains for which no mutation was found were excluded from the analysis.

DISCUSSION

In this population-based study conducted in Vietnam, katG₃₁₅ mutations occurred in 75% of the isoniazid-resistant strains and were more often found in strains resistant to rifampin, streptomycin, or ethambutol. In contrast, the inhA promoter region mutations were less frequent among isoniazid-resistant strains and only associated with streptomycin resistance. Follow-up of new TB patients on standard first-line treatment showed that katG₃₁₅ and inhA promoter region mutations were both strongly associated with relapse. katG₃₁₅ mutations showed a 3-fold but nonsignificant association with treatment failure, while inhA promoter region mutations showed no such association at all. Unfavorable treatment outcome was, however, significantly associated with katG₃₁₅ mutations, as previously reported by Tolani et al. (35). This was expected because these mutations confer high-level isoniazid resistance (13). The independent association between katG₃₁₅ and failure was nonsignificant, probably because of small numbers, which is supported by the finding that the association was of similar magnitude but now significant when treatment failure and death were combined. We found no associations between katG₃₁₅ or inhA mutations and the Beijing genotype.

A strong association between katG₃₁₅ and inhA mutations and relapse may be caused by the 8-month 2SHRZ/6EH regimen used for new TB patients in Vietnam. This means that not only with high-level but also with low-level isoniazid resistance the supplementation of isoniazid with only ethambutol in the continuation phase of treatment is not effective in sterilizing M. tuberculosis. Interestingly, the association between isoniazid resistance and relapse was most pronounced for inhA mutations. Since the catalase-peroxidase release is a component of the bacterial oxyR response, this helps the bacteria to survive inside macrophages (36). Hence, the probability of survival of bacteria with an inhA mutation inside macrophages is higher than for the katG₃₁₅ mutant strains because they still have full catalase-peroxidase expression. Whether the increased risk of relapse, in particular for inhA mutations, also exists with the World Health Organization (WHO)-recommended 6-month regimen (2RHEZ/4RH) remains to be studied.

In the univariate analysis katG₃₁₅ mutations were strongly associated with the Beijing genotype. Although in accordance with previous studies (9, 17, 18), this association completely disappeared after adjustment for streptomycin resistance. In our study 47% of the Beijing strains were resistant to streptomycin of which 48% also had katG₃₁₅ mutations, suggesting that in the Beijing strains streptomycin resistance and katG₃₁₅ mutations are often present simultaneously. This correlation is not unexpected; streptomycin resistance was also associated with the Beijing genotype, MDR-TB and increased transmission in the same study area (37). The association between streptomycin resistance and high-level isoniazid resistance and MDR, especially among Beijing strains, needs further study. It might be related to a specific combination of low-fitness cost mutations conferring these resistances and the strain's genetic background. In addition, yet-unknown compensatory mutations may contribute to the strain's fitness.

Our findings also have consequences for the choice of the standard first-line regimen in Vietnam. In line with WHO recommendations, the 8-month regimen for the treatment of new TB patients should be replaced by the 6-month regimen, including rifampin in the continuation phase.

There were limitations to our study. We only tested isolates for mutations that showed phenotypic resistance to isoniazid and may have missed genotypically isoniazid-resistant isolates. However, drug susceptibility testing was done by an internationally recognized reference laboratory that has consistently shown high concordance rates in proficiency testing therefore this risk is small. We did not determine the MICs of isoniazid for the M. tuberculosis isolates. We also did not use other genotyping methods to assess the type of mutations conferring isoniazid resistance other than those included in the MTBDRplus test. However, the MTBDRplus test covers ≥90% of the isoniazid mutations in M. tuberculosis isolates in Vietnam, as shown previously (15). HIV testing was not routinely performed for all patients. However, the HIV infection prevalence in Vietnam is estimated to be 0.4% of the adult population, with substantially lower prevalence in rural provinces than in major cities. We collected no data on clinical characteristics known to predict treatment failure or relapse, such as the presence of comorbidities or cavities on the chest X-ray. Since it is unlikely that these would be associated with specific isoniazid resistance-conferring mutations before treatment, we do not expect that this resulted in uncontrolled confounding of the observed associations.

In conclusion, isoniazid resistance was most frequently due to mutations in the katG₃₁₅ gene, and these mutations were associated with multidrug and polydrug resistance, whereas inhA mutations were less frequent and were only associated with streptomycin resistance. Both katG₃₁₅ and inhA mutations increased the risk of relapse. Our results also suggest that in Vietnam the 8-month regimen should be discontinued and be replaced by the WHO-recommended 6-month regimen for the treatment of new TB patients.