Introduction

Tuberculosis (TB) remains a significant health issue worldwide. Resistance to anti-TB drugs makes treatment difficult. Isoniazid (H) resistance without concomitant rifampicin resistance is a common problem. Inadequate management of isoniazid-resistant TB (Hr-TB) creates the ideal circumstances for amplification of resistance to multidrug resistance (MDR, resistance to at least isoniazid and rifampicin) which is much harder to treat. Detection of isoniazid resistance is often delayed or overlooked entirely, and even once identified there is very limited evidence to direct treatment regimen choice in such cases.

Worldwide, Hr-TB without MDR-TB is estimated at 9.5% of all cases; for new cases it is 8.1% and for previously treated cases it is 14% [1]. According to the National Surveillance Study in Peru (2005–2006), Hr-TB without MDR-TB rate in new cases is 6.3% and with previous treatment it is 6.7% [2].

Timely detection of this and other types of resistance is important; the World Health Organization (WHO) approved the End-TB Strategy in May 2014 one of the mainstays of which is the universal use of rapid drug susceptibility testing (DST) for early resistance detection [3]. In Peru these measures were adopted earlier; in 2002 drug susceptibility testing for first-line and second-line anti-TB drugs using the proportion method on Middlebrook 7H10 agar (APP) was implemented; by 2006 implementation and decentralization of nitrate reductase assay was initiated [4]; by 2008 MODS was validated [5]; and by 2010 the GenoType MTBDRplus method was validated by the National Institute of Health (Instituto Nacional de Salud, INS) [6]. This development of rapid drug susceptibility testing (DST) allowed for universal access since 2011, which in turn facilitated early resistance detection not only for rifampicin but also for isoniazid [7].

There is no international consensus for the treatment of Hr-TB, due to a lack of evidence; remarkably, however, most of the emphasis is on regimens that only include first-line drugs. The British Thoracic Society (BTS) recommended in 1998 2 months of streptomycin (S), rifampicin (R), pyrazinamide (Z), and ethambutol (E) followed by 7 months of rifampicin and ethambutol (2SRZE/7RE) when H resistance is diagnosed before treatment initiation, and 2RZE/10RE if the patient had already started treatment [8], a position maintained in the recommendations from the 2016 NICE clinical guidelines [9]. In 2003 the ATS (American Thoracic Society), IDSA (Infectious Diseases Society of America), and CDC (Center of Disease Control and Prevention), recommended HREZ and then RZE after H-resistance identification, for a total of 6 months [10].

In the 2008 Management Guidelines, the World Health Organization (WHO) recommended R,Z,E for 6–9 months, with addition of levofloxacin (Lfx) for extensive disease [11]; however, the 2010 Guidelines state that the most effective regimen for this type of resistance is not known, and recommend that for cases which had already started treatment and have known or suspected H resistance, HRE be maintained for 5–7 months, whilst acknowledging that the level of evidence supporting this or any other recommendation is insufficient [12]. In the 2014 Management Guidelines, the WHO again recommended R,Z,E for 6–9 months with caution, both if resistance to E and Z is not known, due to risk of multidrug resistance, and in case of resistance to H alone, due to potential for R resistance amplification; based on expert opinions, Lfx may be added as well [13]. In the latest 2016 Management Guidelines, the WHO makes no recommendations about Hr-TB management, due to lack of evidence [14]; new specific recommendations are expected in 2018.

In Peru, the 2006 Technical Standard for Tuberculosis Control recommended 2 months of SHREZ for H resistance alone, followed by 1 month of HREZ, and then 5 months of HRE twice a week (2SHRZE/1HRZE/5H₂R₂E₂), and in case of H and S resistance, kanamycin (Km) was to be added to the regimen 2KmRZE/1Km₃RZE/6RZE [15]. The 2010 Peru Technical Standard indicated RZE or RZE plus ciprofloxacin (Cpx) for 9 months, depending on the timing of diagnosis [16]; that same year, based on the 2008 WHO Management Guidelines, Cpx was replaced by Lfx in all treatment regimens [11]. The treatment regimen in the 2013 Technical Standard was two months of Lfx, R, E, Z for the intensive phase, followed by seven months of Lfx, R, E (2LfxRZE/7LfxRE)[17], which was progressively implemented. This regimen was designed based on review and analysis of susceptibility results from 12,311 M. tuberculosis isolates obtained from 2007 to 2009, in which 98.9% of H resistant non-MDR strains diagnosed by rapid DST were found to be susceptible to at least 3 agents of a CpxRZE regimen [18]. Prior to this 2013 recommendation, in some cases an aminoglycoside was added to the regimen, at the discretion of the treating clinician but with no clear justification recorded.

Unlike treatment for susceptible and MDR TB, there are few studies and no clinical trials which assess different regimens for Hr-TB without MDR-TB, and the recent review by Gegia [19] shows that first-line treatment only is inadequate; here we report treatment outcomes for a cohort of patients with Hr-TB treated under programmatic conditions in Peru with a standardized nine month quinolone containing regimen.

Methods

Results

In the RNTR, 1973 cases of H-resistant TB without MDR-TB were identified. A total of 623 cases were excluded; 1350 started treatment with LfxREZ (n = 1143) or LfxREZ plus an injectable drug (n = 207); of these, 403 cases had no treatment conclusion date, so treatment outcome could not be evaluated; these cases were designated as not evaluated. There were 947 cases with treatment initiation and termination dates (791 and 156 with and without use of injectable drug, respectively) where treatment outcome was evaluated (evaluated cases) (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart for case inclusion.

https://doi.org/10.1371/journal.pone.0206658.g001

Evaluation cohort vs. not evaluated cases

To determine whether the evaluation cohort is representative of patients treated for Hr-TB, the characteristics of evaluated and not evaluated groups were compared. Table 1 shows these results; there were no significant differences in gender, comorbidities (HIV, diabetes), or availability of results from rapid DST and APP DST. Patients without a treatment outcome date, and thus not evaluated, were more likely to have been treated outside of Lima-Callao (p = 0.02) and have started treatment in 2013/14 (p<0.05). The 15–34 age group was also slightly over-represented amongst the non-evaluated (p = 0.03).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of evaluated and not evaluated groups.

https://doi.org/10.1371/journal.pone.0206658.t001

Treatment outcomes

Table 2 shows treatment outcomes and distribution based on study variables. Treatment success, a composite of cure and treatment completion, was recorded for 77.2% of cases. Although the percentages of patients experiencing failure (1.3%) and death (1.9%) were very low, the loss to follow-up was high (19.6%), with no significant variation over the years. Therefore, the unsuccessful treatment outcome in 22.8% was driven by loss to follow-up.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of evaluated cases (n = 947) based on treatment outcomes.

https://doi.org/10.1371/journal.pone.0206658.t002

Women experienced greater treatment success than men (85.1% vs. 73.6%), with a large gender differential in dropout and failure rates for men vs. women (22.7% vs. 12.9% and 1.7% vs. 0.3%, respectively), with similar death rates (2.0% in men and 1.7% in women). Loss to follow-up was twice as frequent amongst patients with HIV co-infection (39.0 vs. 18.4% amongst HIV-uninfected).

Amongst patients for whom H resistance was only diagnosed with the conventional agar proportion method–those without an available rapid H DST–the percentage of loss to follow-up (23.9% vs. 18.7%), failure (1.9% vs. 1.1%) and death (3.1% vs. 1.6%) was higher than for patients with available rapid DST results. Use of a second-line injectable (vs. non-use) was associated with a higher percentage of loss to follow-up (25.6% vs. 18.5%) and death (7.1% vs. 0.8); among treatment failure cases there were no patients who received second-line injectable treatment. Fig 2 shows treatment outcomes by year of treatment initiation. Treatment success was recorded for 80.1% in 2012, 75.2% in 2013, and 76.1% in 2014.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Treatment outcome based on treatment initiation year.

https://doi.org/10.1371/journal.pone.0206658.g002

The univariate logistic regression analysis (Table 3) showed decreased treatment success to be associated with male gender, lack of an available rapid DST for H resistance and use of a second-line injectable drug. In the multivariate logistic regression analysis, in addition to the previous variables, treatment initiation in 2014 was also independently associated with a less favorable treatment outcome.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Univariate and multivariate logistic regression analysis of variables that may affect treatment success.

https://doi.org/10.1371/journal.pone.0206658.t003

Discussion

This study has demonstrated that the LfxREZ regimen indicated for Hr-TB in Peru results in a favorable outcome in 77.2% of treated patients, with the dominant reason for lack of treatment success due to a high proportion of loss to follow-up. Loss to follow-up has been a programme-wide problem in Perù, not specific to Hr-TB; between 2008–2010 the loss to follow up was 10% for drug-susceptible TB and 18–20% for MDR TB [7]. Variables that diminished treatment success were male gender, lack of available rapid DST for H resistance, use of adjunctive second-line injectable drug therapy and treatment initiation in 2014.

A strength of this operational research study is the usage of nationwide programmatic data from the National Resistant Tuberculosis Registry and inclusion of all cases approved for treatment initiation for isoniazid-resistant TB. In addition, the strict microbiological inclusion criterion depended upon proven isoniazid resistance at the national mycobacteria reference laboratory, which is overseen for EQA by the supranational reference laboratory.

The registry includes a significant section (403/1350, 30%) of patients for whom no treatment outcomes (“not evaluated”) had been recorded. To assure that exclusion of these patients from the evaluation cohort would not generate a major bias, their characteristics were analyzed and found to show no important differences regarding demographic data and disease characteristics. It has only been shown that compared with evaluated patients, “not evaluated” status was more common in the provinces than in Lima and were more frequent in 2013 and 2014 than in 2012. In the evaluation cohort, loss to follow-up was more common in the two more recent years of study. Despite the apparent demographic and clinical similarity of the “not evaluated” group to the “evaluated” group, there remains a significant concern that the evaluated group may not accurately represent the entire cohort of LfxREZ treated patients. From our available data we are unable to exclude the possibility that deaths, treatment failures and loss to follow-up may have been much greater in the “not evaluated” group, which remains an important limitation of this analysis.

Treatment success broadly consisted of half cured and half patients who had completed treatment. Although failures and deaths were very few, the high proportion of loss to follow-up is striking, and it is important to acknowledge the possibility that there may have been more patients who failed treatment or died amongst this group that were not captured or recognized by the TB programme.

Men had less favourable outcomes and more loss to follow up than women, as is seen for TB in general, regardless of drug-susceptibility. In Peru, TB mainly affects men and the age group 15–44, and 95% of cases are seen in the lower socioeconomic strata, where men are the ones who mainly work; this may in turn affect treatment adherence and lead to subsequent unfavorable outcomes, since work or school are important reasons for non-adherence [21].

Lack of an available rapid DST for H resistance, and thus reliance only on the APP DST result with longer turnaround times, may delay initiation of an adequate treatment and thus explain lower favorable outcomes. Current implementation in Peru of universal rapid DST, which includes isoniazid susceptibility testing–MODS and Genotype MTBDRplus–has been a key policy for treatment success, with a standardized regimen initiated without delay (Obregon Boltan G. et al. Introduction of rapid drug susceptibility tests and treatment outcomes for multidrug-resistant TB in Peru, 2010–2015. In Press Int J Tuberc Lung Dis 2018). A drawback to the Xpert MTB/RIF test, implemented at large-scale in several countries, although not yet in Peru, is the lack of information about isoniazid resistance. Without these rapid DSTs, patients included in this analysis could not have been identified; most patients would have started 2RHZE4RH, which in a recent systematic review [19] was associated with failure in 11% and relapse in 10% of cases, and with additional acquired resistance to rifampicin (acquired MDR-TB) in 8% of cases. It may be inferred that this source of MDR-TB has been shut off with the implementation of rapid DST in Peru.

In 16% of cases, a second-line injectable drug was added to the standardized regimen (LfxREZ). Among the 18 deceased patients, 61% received an injectable drug, which suggests that the use may have been indicated for cases with a greater clinical deterioration (as this was not algorithm-based but based on medical decision).

Treatment initiation year was identified as a variable that decreased favorable outcome in 2014 more than in 2013 and 2012; the multivariate analysis shows that this effect persists regardless of injectable drug use and additional resistance profiles. For the most part, the effect is due to increased loss to follow-up; this is a significant gap that the programme must urgently address.

In common with any other operational research study which takes advantage of available programmatic data, there are certain limitations to take into consideration. This is a single-cohort study, with no control group.

In order to be able to define treatment outcome, treatment initiation and termination dates are needed; healthcare facilities are required to report these dates, and if this is not performed adequately, not all cases can be evaluated. This situation was more common in the provinces than in Lima-Callao. Thirty percent of patients were excluded as “not evaluated”; however, no significant differences that compromise generalizability were seen between the groups, and therefore the study is considered representative of Peru. The National Resistant Tuberculosis Registry includes no treatment information prior to diagnosis of drug-resistance. As treatment cards are not available, there is no opportunity for source data verification.

In Peru, cases treated for drug-resistance are required to have monthly cultures throughout the course of treatment [16], however this is not performed in all cases, particularly when cough has resolved or become non-productive, which limits the determination of “cured” status. There are no defined criteria for treatment outcome with this type of resistance, in particular for cure and failure; therefore WHO criteria were adapted for this study [20]. Given the high loss to follow-up the true failure rate is difficult to assess. An additional limitation of this analysis is that the information system does not record data on acquired resistance or post-treatment relapse.

The presence of H resistance is associated with unfavorable treatment outcomes. In a prospective observational study conducted in Lima, Peru, on 1039 cases between 2010 and 2011, 8% of cases had H resistance alone. These cases had a higher risk of failure (2%) and death (5%) than susceptible cases [22].

In other retrospective studies in which treatment regimens were assessed, treatment outcomes are highly variable. Menzies et al performed a meta-analysis based on studies conducted between 1965 and 2008, with very variable regimens, and highoighted the complete lack of randomized controlled clinical trials. In cohorts that used 2HREZS/1HREZ/5HRE with a continuation phase of daily, or 2 or 3 times a week administration, the failure rates were 18–44%. Other regimens which included R use only in the initial phase had a relapse and failure rate of up to 70%. Unfavorable outcomes were found to be related to use of R exclusively during the initial phase, and it is recommended that at least 4 effective drugs be used during the intensive phase and 3 during the continuation phase [23].

Gegia et al expanded the meta-analysis mentioned above with studies conducted up to March 2015; the use of first-line drugs to treat H resistance was evaluated. Cases which used the WHO recommended regimen for new patients (2HREZ/4HR) showed an 11% failure rate, a 10% relapse rate, and development of MDR-TB in 8%. Among those who received the WHO regimen recommended for previously treated patients (2HREZS/1HREZ/5HRE), failure rate was 6%, relapse rate was 5%, and 3% developed MDR-TB. Therefore, the development of better regimens is recommended [19].

There is a distinct paucity of data on the use of fluoroquinolones for treatment of Hr-TB though there are some encouraging early signals. In an evaluation of 40 patients conducted in Denmark from 2002 to 2007 with H resistance determined by BACTEC 460, the REZ regimen plus fluoroquinolone for an average of 277 days had a favorable outcome in 90% of cases [24].

Further evidence for the use of fluoroquinolones instead of isoniazid comes from three large clinical trials using a fluoroquinolone (moxifloxacin or gatifloxacin) with the aim of reducing treatment for drug-susceptible TB to 4 months [25–27]. Treatment outcomes achieved by fluoroquinolone regimens were non-inferior compared with those seen with conventional treatment (2RHZE4RH).

Lee et al presented treatment outcomes with fluoroquinolone use (Lfx and Mfx) in 75 cases of H resistance as determined by DST using proportion method. This regimen had R (94.7%), E (90.7%), Z (82.7%), and S (2.7%) with a favorable treatment outcome in 97.3% and only 1.3% failures [28].

Overall the use of fluoroquinolones improves treatment outcomes over non-fluoroquinolone containing regimens, but studies are scarce. Current evidence suggests that regimens for H resistant TB should at least include R throughout treatment, should include at least 4 effective drugs for the induction phase and should probably last over 6 months, perhaps with additional benefit from prolonging to 9 months. A forthcoming systematic review, undertaken to inform new WHO guidance on Hr-TB management, should shed further light on the minimum requirements for an effective regimen.

The present study evaluated a much greater number of cases (n = 947) than previously reported; favorable outcome was significantly affected by loss to follow-up, which was greater than in other studies, but failure was minimal (1.2%).

In case of no resistance to the drugs included in the treatment, the regimen indicated in Peru for treatment of isoniazid resistant TB has a high treatment success rate and does not require the addition of an injectable drug to the regimen. In programmatic conditions, with the use of rapid DST which includes an isoniazid susceptibility test, the regimen 2LfxRZE/7LfxRE has shown strong performance in patients who remained in the programme. The regimen was derived from a detailed prior analysis of national drug-resistance epidemiology in strains resistant to isoniazid but susceptible to rifampicin. Improved adherence and sustained access to rapid DST may serve to further improve treatment success. In places where the resistance profile is known in strains resistant to isoniazid but not to rifampicin, the use of rapid DST with a rapid isoniazid test is recommended, along with the use of the 2LfxRZE/7LfxRE regimen, with particular attention to minimize loss to follow-up during treatment.

Supporting information

Acknowledgments

This research was made possible with support from the United States Agency for International Development (USAID) through the TREAT TB Cooperative Agreement (AID-GHN-A-00-08-00004). The contents are the responsibility of the authors and do not necessarily reflect the views of USAID or the United States Government. Additional support was provided through the Structured Operational Research and Training Initiative (SORT IT), a global partnership led by UNICEF/UNDP/World Bank/WHO Special Programme for Research and Training in Tropical Diseases (TDR) based at the World Health Organization.