Isoniazid (INH) mono-resistance and tuberculosis (TB) treatment success: analysis of European surveillance data, 2002 to 2014

Introduction: Isoniazid (INH) is an essential drug for tuberculosis (TB) treatment. Resistance to INH may increase the likelihood of negative treatment outcome. Aim: We aimed to determine the impact of INH mono-resistance on TB treatment outcome in the European Union/European Economic Area and to identify risk factors for unsuccessful outcome in cases with INH mono-resistant TB. Methods: In this observational study, we retrospectively analysed TB cases that were diagnosed in 2002–14 and included in the European Surveillance System (TESSy). Multilevel logistic regression models were applied to identify risk factors and correct for clustering of cases within countries. Results: A total of 187,370 susceptible and 7,578 INH mono-resistant TB cases from 24 countries were included in the outcome analysis. Treatment was successful in 74.0% of INH mono-resistant and 77.4% of susceptible TB cases. In the final model, treatment success was lower among INH mono-resistant cases (Odds ratio (OR): 0.7; 95% confidence interval (CI): 0.6–0.9; adjusted absolute difference in treatment success: 5.3%). Among INH mono-resistant TB cases, unsuccessful treatment outcome was associated with age above median (OR: 1.3; 95% CI: 1.2–1.5), male sex (OR: 1.3; 95% CI: 1.1–1.4), positive smear microscopy (OR: 1.3; 95% CI: 1.1–1.4), positive HIV status (OR: 3.3; 95% CI: 1.6–6.5) and a prior TB history (OR: 1.8; 95% CI: 1.5–2.2). Conclusions: This study provides evidence for an association between INH mono-resistance and a lower likelihood of TB treatment success. Increased attention should be paid to timely detection and management of INH mono-resistant TB.


Introduction
Tuberculosis (TB) causes a large degree of suffering and an estimated 1.3 million deaths per year globally, occurring mainly in less affluent countries, but also in upper-middle and high-income countries in the European Union/European Economic Area (EU/EEA) [1,2]. In Europe, there has been a steady decline in TB notification rates of ca 5% per year. Nevertheless, TB remains a considerable problem because of multidrugresistant (MDR) and extensively drug-resistant (XDR) TB [3].
The main public health response to the TB epidemic consists of early diagnosis, prevention of transmission and adequate treatment. In general, treatment is most successful when there is no resistance to any of the drugs designated for treatment of TB [4], and the drugs isoniazid (INH) and rifampicin (RIF) can be included in the treatment regimen. INH has long been an essential component of first-line treatment for active TB and an important drug in TB control because of its potent early bactericidal activity, low rate of adverse events and low cost [5]. Currently, there is no equivalent alternative available [6]. Resistance to INH is prevalent [4], with substantial geographic variation [7].
INH mono-resistance increases the likelihood of negative treatment outcome and progression to MDR TB [4,8,9]. Reported treatment outcome seems to differ by setting and region for cases with INH-resistant TB [10][11][12]. Population groups that are especially at risk of negative treatment outcome due to INH mono-resistance are children and HIV-positive patients [13][14][15]. Recently, two systematic reviews assessed treatment options for INH mono-resistant TB [8,16]. One review concluded that treatment with first-line drugs resulted in suboptimal outcome [8], whereas the other showed that extending the duration of RIF and increasing the number of effective drugs lowered the odds of unfavourable outcome [16]. An analysis of individual patient data conducted in the framework of a World Health Organization (WHO) guideline development process showed that the addition of a fluoroquinolone to a regimen of 6 months of daily RIF, ethambutol (EMB) and pyrazinamide (PZA) was associated with improved treatment success in INH-resistant cases [17]. After an evaluation of all available evidence, WHO has issued new guidelines on treatment for patients with INH mono-resistance [18]. Discussion is ongoing as to whether to maintain INH in the treatment regimen if a low degree of resistance is detected; however, there is limited data on the effect this strategy has on treatment outcome.
The European Surveillance System (TESSy), hosted by the European Centre for Disease Prevention and Control (ECDC), contains case-based information for more than 1.5 million TB cases reported by EU/EEA countries between 1995-2015 [3]. In contrast to the information from randomised controlled trials (RCTs) and cohort studies that were included in the systematic reviews, TESSy includes information on treatment outcome obtained in a programmatic setting [8,16]. These data are from a larger number of patients, are more recent and are more EU/EEA-focused than the data in the aforementioned reviews. We therefore set out to analyse this dataset to determine the current treatment outcome of INH mono-resistant TB in the EU/ EEA and to identify risk factors for unsuccessful treatment outcome in cases with INH mono-resistance.

Study population and data sources
In this observational study, we retrospectively analysed TB notification data reported to the TESSy database between 2002-14. We included pulmonary and extra-pulmonary TB cases with available information on treatment outcome and drug-susceptibility testing (DST) results for at least INH, RIF, streptomycin (STR) and EMB. Information on DST for PZA was not collected in TESSy during the study period and therefore PZA was not part of our inclusion criteria.

Operational definitions
Treatment outcome was reported 12, 24 and 36 months after the start of TB treatment. We categorised treatment outcome in accordance with the 2017 joint WHO Regional Office for Europe/ECDC surveillance and monitoring report [3]. For cases with a treatment outcome of 'still on treatment' at 12 or 24 months, the final treatment outcome reported at 24 or 36 months was used, respectively. Unadjusted data was stratified by age, although this stratification was not described in the protocol for this study. The operational definitions used in this study are listed in the Box.

Statistical analysis and modelling approach
To investigate the impact of INH mono-resistant TB compared with fully drug-susceptible TB on treatment success, we applied a multilevel logistic regression model. To adjust for the heterogeneity between countries, the model was corrected with a random intercept at the country level for the differences in the average treatment success rate between countries and with a random slope for the differences in the INH monoresistance effect on the treatment success rate at the country level. The necessity of adding a random intercept and a random slope in the model was determined using the likelihood-ratio test. The main outcome was dichotomised as unsuccessful treatment (failed, died, lost to follow-up and not evaluated) vs treatment success (cured or completed), using the final reported treatment outcome. Independent variables available in the TESSy data (age, sex, geographical origin, type of TB, microscopic confirmation, history of TB, HIV status and reporting year) were also assessed as possible

Box
Operational definitions for analysis of European tuberculosis surveillance data, 2002-2014 The following definitions were used: • INH mono-resistant TB cases: cases with resistance to INH and documented susceptibility to RIF, STR and EMB; • Cases with fully drug-susceptible TB: cases with documented susceptibility to INH, RIF, STR and EMB; • New TB cases: cases who were never previously treated for TB or who received drug treatment for less than 1 month; • Cases with a history of TB: cases who were previously treated for TB for 1 month or more (for countries who did not report information about previous treatment, the variable previous diagnosis was used as a proxy); • Geographical origin of cases, i.e. native vs foreign: based on the country of birth or, if this information was unavailable, on the citizenship of the patient; • Low TB-incidence country: country with a TB incidence rate < 10 cases per 100,000 population [31]; • High TB-incidence country: country with a TB incidence rate ≥ 10 cases per 100,000 population [31].

Table 2b
Characteristics of tuberculosis cases by isoniazid mono-resistance status, 24 EU/EEA countries, 2002-2014 (n = 194,948) the included study population. In a sensitivity analysis, cases with the outcome 'not evaluated' were excluded. As the outcomes were proportions, we used logistic regression for all models. Odds ratios (OR) with 95% confidence intervals (CI) were calculated to assess the strength of the association. All analyses were performed using the STATA (Stata/SE 14.1, StataCorp LP, Texas, United States (US)) software.

Ethical statement
This study is based on data collected on the basis of statutory notification in each EU/EEA country and reported anonymously to ECDC on the basis of decision No 2119/98/EC of the European Parliament and of the Council [19]. Therefore, informed consent from patients is not required.  EU/EEA countries fulfilled the inclusion criteria and were therefore eligible for our analysis. Of them, 7,578 (3.9%) were cases with INH mono-resistant TB ( Figure  1). Differences in the characteristics of included and excluded cases are presented in Table 1

Factors associated with unsuccessful final treatment outcome of INH mono-resistant tuberculosis cases
In the multivariable model adjusted for heterogeneity between countries, unsuccessful treatment among INH mono-resistant TB cases was associated with age ≥ median age (41 years) (OR: 1.  Table S2). In the sensitivity analysis, excluding cases with the treatment outcome 'not evaluated', no change in the associated risk factors was observed (data not shown).
INH mono-resistant and fully susceptible TB shared the same risk factors for unsuccessful treatment, except for being of foreign origin, which was associated with a higher risk for unsuccessful treatment in fully susceptible but not INH mono-resistant TB cases in the multivariable model (data not shown).

Discussion
Our retrospective study of European surveillance data, including 7,578 cases of INH mono-resistant TB and 187,370 cases of fully susceptible TB, shows that INH mono-resistance is associated with lower TB treatment success in the final TB treatment outcome. This association between INH mono-resistance and lower treatment success is in line with a previous systematic review [20]. Although this review was published in 2009, it includes many studies that were published  before the year 2000 and were mainly conducted in countries in Asia and Africa. The same applies to the reviews evaluating treatment regimens for INH monoresistant TB that were published in 2016 [16] and 2017 [8], respectively. Therefore, our analysis of more recent 2002-14 EU/EEA surveillance data represents an addition to the currently available evidence.
Studies of TB treatment outcome under routine programmatic conditions have shown inconsistent results regarding treatment outcome of INH mono-resistant TB. While studies from the US [10], Denmark [12] and Israel [21] reported treatment outcome for INH monoresistant TB as excellent, highly successful or similar to drug-susceptible TB, studies from Peru [22], Mexico [23], Georgia [24] and South Africa [11] showed poorer treatment outcome compared with fully susceptible TB. Possible explanations might be differences in the included patient populations; for example, regarding HIV prevalence [11,12], the availability of resources for patients, the accessibility of healthcare systems or the use of different treatment regimens. Our study, which pools data from 24 low HIV-prevalence and predominantly high-income European countries, shows that INH mono-resistance negatively affects treatment outcome. The study findings also highlight the need for timely identification of patients with INH mono-resistant TB, especially as rapid testing in recent years has focused more on the detection of RIF resistance as a proxy for MDR TB [6].

Figure 4
Factors associated with unsuccessful treatment among cases with isoniazid mono-resistant tuberculosis, EU/EEA, Univariable and multivariable analyses were based on multilevel logistic regression models corrected for clustering within countries using an unstructured covariance matrix. TB cases with available information for all predicting factors were included in the multivariable analysis (n = 5,759/7,578). Excluded cases have a slightly higher treatment success (74.7% vs 73.8%; p value = 0.46). Data source: TESSy.
Comparison of studies has been hampered by different definitions used for INH mono-resistance. While many studies retained all INH resistance profiles, provided that RIF resistance and therefore MDR TB was excluded [8,16], other studies required additional documented susceptibility for STR and EMB [23] or analysed INH mono-resistant and INH poly-resistant cases with additional resistance to STR or EMB separately [12]. To avoid misclassification, we have used the stricter definition requiring documented DST results for INH, RIF, STR and EMB for all cases. However, while gaining specificity in the definition of INH mono-resistance, this approach has resulted in the loss of a large number of cases for which these susceptibility testing results were not available, with considerable differences in the percentages of cases that could be included by country.
The finding that EU/EEA countries have different proportions of patients with INH mono-resistant TB still under treatment at 12 months might be a surveillance artefact related to variation in reporting procedures or the result of the use of different treatment regimens with different durations, reflecting the current lack of an agreed standard regimen for the treatment of INH mono-resistant TB cases [18,25,26]. As the 12-month outcome therefore did not seem to be the most adequate endpoint for analysis of the treatment outcome of INH mono-resistant TB, we chose a composite outcome for this study using the final documented outcome irrespective of the time to reporting (12 months, 24 months or 36 months after the start of treatment).
The factors found to be associated with a higher risk of unsuccessful treatment outcome in INH mono-resistant TB in this study-higher age, male sex, positive microscopy, positive HIV status-have been described before as associated with unsuccessful TB treatment outcome independent of drug resistance status [27,28]. Prior TB treatment has also been reported as a risk factor for unsuccessful outcome in patients with INH monoresistant TB [29]. In our study, these factors influenced TB treatment outcome regardless of the presence or absence of INH mono-resistance, indicating that they are not specific to INH mono-resistant TB, but rather are associated with lower TB treatment success in general.
The strengths of our study are the inclusion of a large number of cases and the application of a multilevel model to correct for TB clustering. This allowed to control for possible selection bias related to reporting countries and for unobserved heterogeneity between countries, thereby enhancing the generalisability of our findings. However, there are also several limitations to our study, mainly related to the use of surveillance data collected with the aim to inform TB programme management and not to evaluate clinical outcomes. As a result, data on the severity of TB disease, underlying diseases and treatment regimens were not available.
In addition, information on DST for PZA is not available in the TESSy data. In a sensitivity analysis using German notification data (40,063 TB cases, of which 1,310 were INH mono-resistant) that include information on DST for PZA, cases resistant to PZA were more frequent among INH mono-resistant TB compared with otherwise fully susceptible TB cases (5.4% vs 1.8%; p < 0.01). However, no impact of the PZA resistance status on the relationship between INH mono-resistant TB and treatment success was observed (adjusted OR: 0.8; 95% CI: 0.6-1.4; p = 0.77). Another limitation related to DST is that no information is available on the level of INH-resistance and the type of INH resistance mutations involved, which have been shown to influence treatment outcome [30].
Lastly, our study included data from only 24 of 31 EU/EEA countries; cases from seven countries were excluded due to lack of reporting treatment outcome or case-based susceptibility data (Table 1). Therefore, our data pertain only to these 24 EU/EEA countries and cannot be generalised to the whole EU/EEA without caution. Of note, 80.7% of reported cases had to be excluded due to missing information. As systematic reasons for lack of reporting within the 24 countries with included cases are not known, it is not possible for us to hypothesise how this might have affected our findings.
In conclusion, this study shows that treatment of patients with INH mono-resistant TB under routine programme conditions leads to lower treatment success compared with fully susceptible TB. The association of INH mono-resistance with negative treatment outcome highlights the need to pay increased attention to the timely identification and management of these cases to ensure treatment success for individual patients, as well as to reduce the risk for further resistance development on a population level.