Treatment Outcome of Multi-drug Resistant Tuberculosis in the United Kingdom: Retrospective-prospective Cohort Study from 2004 to 2007

Treatment outcome of multi-drug resistant tuberculosis in the United Kingdom: retrospective-prospective cohort study from 2004 to 2007. United Kingdom (UK) guidelines recommend at least 18 months treatment for patients with multidrug-resistant tuberculosis (MDR-TB). Prior to 2008, data on treatment outcome were only available at 12 months and therefore the proportion completing treatment was unknown. This retrospective-prospective cohort study reports on treatment outcomes for MDR-TB patients notified between 2004 and 2007 and examines factors associated with successful outcomes. 70.6% (144/204) completed treatment in 24 months or more, 6.9% (14) stopped treatment, 6.9% (14) died, 7.8% (16) were lost to follow up, 0.5% (1) relapsed and 4.4% (9) were transferred overseas. Following adjustment for age, being non-UK born, non-compliance and having co-morbidities, treatment with a fluoroquinolone (OR were independently associated with successful treatment outcome. Treatment completion for MDR-TB cases remains below the World Health Organization (WHO) target. Our findings support current WHO guidelines for MDR-TB treatment. The UK should consider adopting individualised regimens based on WHO recommended drugs, taking into account drug sensitivities. Improving treatment completion rates will be key to tackling further drug resistance and transmission from untreated infectious cases.


Introduction
Multidrug-resistant TB (MDR-TB) remains a threat to the global tuberculosis (TB) control effort [1].In the United Kingdom (UK), the annual number of culture confirmed cases of MDR-TB increased from 28 to 58 between 2000 and 2009 [2] and there were a total of eight extensively drug-resistant (XDR) cases reported (data unpublished).The prolonged treatment associated with MDR-TB and the often severe adverse effects of second-line antibiotics increases the challenges to achieve treatment completion.The rise in the number of MDR-TB cases has important implications for clinical management, social support and financing of TB control programmes [3].Internationally, in resource rich settings, initial empirical treatment of MDR-TB patients should be based on past drug resistance results for patients with a previous TB episode, drug resistance profiles of an identified source case, or the levels of background drug resistance in the patient's country of origin [4,5].This should be followed by individually adapted drug regimens once drug susceptibility results become available [4].
In 2008, the World Health Organization (WHO) recommended that the MDR-TB treatment regimen should ideally consist of a combination of ethambutol and pyrazinamide, an injectable agent (e.g.aminoglycosides), a fluoroquinolone and if necessary, a bacteriostatic drug should be added to give a total of at least four drugs to which resistance has not been demonstrated.Antibiotics with unknown efficacy should only be used when better options are exhausted [4].Recently published WHO guidelines recommend the inclusion of the bacteriostatics ethionomide or prothionamide and either cycloserine or ρ-aminosalicylic acid in the regimen [5].The treatment should last at least 20 months in total [5] and be supervised by directly observed therapy (DOT) [4].
In the UK, there is no recent national guidance for MDR-TB treatment [6,7].The National Institute for Health and Clinical Excellence (NICE) guidelines published in 2011, did not specifically address the treatment of MDR-TB but suggested to consult experienced clinicians who specialise in MDR-TB treatment and care [8].Data on the effectiveness of different drug combinations for MDR-TB are limited [5] and in the UK, it is currently unknown which treatment regimens are most commonly used.
For cases notified in the UK in 2010, the proportion of MDR-TB cases completing treatment was 72.1% [9] which was below the WHO and UK treatment completion targets of 75% [10] and 85%, respectively, [11] but higher than the European Union (EU) target of 70% [12].Prior to 2008, data on treatment outcome was only available at 12 months after the start of treatment and therefore it has been unclear how many cases completed treatment at 24 months.The enhanced surveillance of treatment outcome monitoring of MDR-TB cases allows treatment regimens and management to be assessed and progress towards achieving targets set by WHO [10], the EU [12] and the UK [11] to be evaluated over time.
The aims of this study were to determine the number and proportion of MDR-TB patients completing treatment who were diagnosed in the UK between 2004 and 2007, to describe the clinical characteristics of patients and to examine factors associated with a successful treatment outcome, loss to follow up and death.

Methods
All patients diagnosed with MDR-TB in England, Wales and Northern Ireland between 2004 and 2007 were included in the study.Treatment outcomes for all years were collected in 2009.Patients still on treatment at 24 months were followed up prospectively for treatment outcome on an annual basis until June 2012.All patients were followed up for a minimum of four years for relapse.
The following definitions and terms were used: MDR-TB was defined as TB resistant to at least isoniazid and rifampicin.

Treatment regimen
Drugs used for treatment were categorised in five groups based on WHO criteria [4] (Table 2).To calculate the number of effective drugs used for treatment, in accordance with current guidelines, Group 5 agents, rifabutin (not recommended due to cross-resistance with rifampicin [4]) and drugs assumed to be ineffective due to phenotypic resistance testing methods were subtracted from the number of drugs in the initial regimen.A drug change was defined as an unexpected and unplanned addition, subtraction or substitution of a drug in the treatment regimen.

Treatment outcomes
Standard treatment outcomes routinely collected for surveillance in the UK are: Completed treatment: Completed a full course of therapy within 12/24 months of starting treatment/ notification.
Lost to follow up: Defined as failure to obtain contact with the patient before the end of treatment so that treatment outcome is not known.

Data collection
Demographic and clinical characteristics of patients were obtained from the Enhanced TB Surveillance system (ETS) which is the web-based TB notification system in the UK.Questionnaires were sent by mail to treating clinics to collect further information on: treatment outcome reported at 12 and 24 months, social risk factors (current or a history of alcohol or drug misuse, homelessness, imprisonment, smoking), co-morbidities (diabetes, chronic liver or renal disease, chronic hepatitis B or C positive and receiving immunosuppressive therapy), the initial drug regimen for MDR-TB, treatment start date, details of changes in treatment, duration of treatment (planned and given for those who completed treatment), DOT (defined as direct observation of ingestion of anti-TB treatment by a health professional in the community, the home or the clinic) and non-compliance (interruption of treatment for two consecutive months or more without medical approval or non-compliance reported by medical staff).
All questionnaires were returned.Previous history of TB diagnosis was self-reported and recorded in ETS; in addition, information from previous episodes diagnosed in the UK was ascertained through a search of ETS to complete missing data.Mortality data from the Office of National Statistics was searched manually for cases that were lost to follow up or had an unknown cause of death.
TB was considered to have caused or contributed to death if this was reported in ETS and/or recorded on the death certificate; ICD10 codes A15-A19 [13].Human immunodeficiency virus (HIV) infection status was attained by record linkage as previously described [2].
Matching was not carried out on cases aged younger than 15 years (10/204; 4.9%) as HIV infection in children is reported separately.

Laboratory methods
Drug susceptibility and strain typing data for cultured isolates of Mycobacterium tuberculosis complex were available from the UK Mycobacterial Surveillance Network (MycobNet).Drug susceptibility testing (DST) was carried out using the proportion or the resistance ratio method [14].MDR-TB cases, notified in the UK (excluding Scotland) between 2004 and 2007, were identified by matching laboratory isolates to case reports in ETS [2], including those who subsequently developed MDR -TB during treatment.Cases of laboratory cross-contamination were excluded.Drugs with borderline resistance were considered to be resistant.The number of additional drugs to which MDR isolates were resistant was based on the resistance profile prior to the initiation of MDR-TB treatment.

Statistical analysis
The demographic and clinical characteristics of cases, drug resistance and the treatment regimen, management and outcome were described.Logistic regression modelling was used to calculate odds ratios for factors associated with a successful treatment outcome.All variables independently associated with treatment outcome in the univariate analysis (p<0.2) were considered in the multivariable model to evaluate the effect of drug treatment on outcome.A likelihood ratio test was used to investigate whether interactions between the different drug classes should be included in the final model.Individual co-morbidities were not considered in the multivariate analysis due to co-linearity with co-morbidity.
Ciprofloxacin is no longer recommended for use and therefore we did not include it as a variable in the model as it is no longer relevant for consideration in future treatment of MDR-TB.
Outcome categories were based on criteria by Ditah et al. ( 2007) [15] but were modified for the study population of drug-resistant, rather than fully sensitive TB cases (Box 1).Neutral outcomes (n=10) and patients who did not initiate treatment because they were diagnosed post mortem (n=3) were excluded from this part of the analysis.
Additional analyses using a chi-square test were undertaken to determine factors associated with the following adverse outcomes: (i) loss to follow up (included all MDR-TB cases) and (ii) death (included all patients with a known vital status at 24 months and excluded those transferred overseas or lost to follow up).Statistical analyses were carried out using Stata version 10.0.

Factors associated with treatment success
Results of the univariate analysis are shown in tables 3 and 4. In the multivariable analysis patients receiving a fluoroquinolone or a bacteriostatic drug were more likely to have a successful treatment outcome compared to those who did not (Table 5).Treatment with an injectable agent did not have a significant effect on treatment outcome after adjusting for treatment with a fluoroquinolone and a bacteriostatic drug.No significant interactions were detected and all other factors remained significantly associated with treatment outcome, apart from having resistance to five or more drugs.Exploratory analyses were carried out to try to explain the relationship between resistance to five or more drugs and a successful outcome.Firstly, resistance to five or more drugs was added to the model, a Patients were coded as yes if they had "yes" for any social risk factor and no if they had "no" for every social risk factor included.b Patients were coded as yes if they had "yes" for any comorbidity and no if they had "no" for every co-morbidity included.If data were missing for one or more co-morbidities, they were coded as missing.HIV: human immunodeficiency virus; TB: tuberculosis; UK United Kingdom.
following adjustment for all factors associated with a poor treatment outcome (Table 5), and remained significant.Each treatment was then added separately and only adjustment for bacteriostatic drug treatment led to loss of statistical significance.

Factors associated with treatment stopped, mortality and loss to follow up
Reasons given for stopping treatment in 14 patients included 'non-compliance', resulting in the clinician's decision to discontinue treatment, 'patient's choice', 'pregnancy', 'side effects' and 'spontaneous recovery'.
Of the 14 patients who died, six died prior to starting treatment and three of these were diagnosed post mortem.Compared to cases who were known to be alive at the end of treatment, death was found to be strongly associated with having any co-morbidity (p<0.0005), and in particular with HIV (p<0.0005),diabetes (p=0.002) or chronic renal disease (p=0.002).
Only being a new entrant to the UK (11/12 were in the UK ≤2 years prior to diagnosis, p=0.030, with six returning home) was associated with being lost to follow-up.Reasons given for completing treatment within 12 months were that 'the patient improved', 'it was the recommendation at the time' or 'it was initially planned and the patient was followed up instead'.

Discussion
The proportion of MDR-TB cases notified between 2004 and 2007 completing treatment in the UK was 70.6%.This was higher than the EU/European Economic Area (EEA) average of 30.9% [16] for 2007 MDR-TB cases and most other low incidence resource-rich countries [17][18][19].This completion rate met the EU target of 70% [12] but was still below the WHO target of 75% [10] and the UK Chief Medical Officers action plan goal of 85% [11].The treatment completion rate for MDR-TB cases in the UK has improved in recent years with 80% and 72% completing treatment for cases notified in 2009 [20] and 2010 [9], respectively.
Treatment with a fluoroquinolone or a bacteriostatic drug were statistically significantly associated with achieving treatment success, which provides further evidence to support the recent WHO recommendations to include drugs belonging to Groups 2, 3 and 4 in a treatment regimen for MDR-TB [5].These findings have potential implications for the development of future national guidelines and the UK should consider adopting individualised regimens, based on the drug classes recommended by WHO for treatment of MDR-TB cases, taking into account DST results.
While the majority of patients appeared to have appropriate treatment according to WHO guidelines, approximately a quarter of patients had a substandard regimen with too few effective drugs or shorter treatment duration than required.DST results were not always used to ensure administration of effective individualised regimens.However, we note that DST results for most drugs other than isoniazid and rifampicin are less accurate [21] and therefore it is possible that these patients still received effective treatment.
The majority of patients, even those with social risk factors and those hospitalised, did not receive DOT although it is recommended for all MDR-TB patients [4].Therefore greater use of DOT remains important until 85% treatment completion is achieved.Mortality rates of 6.4% in our study, were similar to those observed for all TB cases in the UK in the same period (6.2%, 2004-2007) (data unpublished) and to other low incidence countries [16], although higher than expected for a population where most cases are 15 to 44 years old [22].Similarly to other studies, death was significantly associated with presence of HIV infection [23][24][25][26][27], chronic renal disease [28,29] and diabetes [28,30].Treatment of HIV co-infected patients is complicated due to a high tablet burden, increased drug side effects [31] and opportunistic infections.TB patients with chronic renal disease may also experience more treatment side effects and some TB drugs can directly damage kidney function [32].Anti-TB treatment can worsen glycaemic control in patients with diabetes [30].These problems are intensified in MDR-TB where treatment duration is prolonged and drug options are limited.
The association between successful treatment outcome and fluoroquinolone or bacteriostatic drug use, has been shown previously [21,27,[33][34][35][36][37].The relative infrequency of resistance to fluoroquinolones in our study further supports their use in MDR-TB treatment.Consistent with previous reports [38,39], ciprofloxacin was not shown to be an effective agent in the univariate analysis, supporting its recent exclusion from the list of recommended TB drugs [4].
Similarly to findings in resource-rich countries, where the majority of MDR-TB cases are imported, we detected high proportions of streptomycin and ethambutol resistance [17,18,21,40].In contrast to other studies [41][42][43][44][45][46][47], the treatment success of UK MDR-TB patients is not affected by the number of additional drugs to which isolates are resistant, which may reflect the local availability of alternative second line antibiotics.The association between resistance to a greater number of drugs and a successful treatment outcome was not significant following adjustment for treatment with a bacteriostatic agent.A possible explanation for this is that those with fewer treatment options are more likely to receive a bacteriostatic which leads to treatment success.
Our study has several limitations.The small sample size limited our ability to detect the effect of individual antibiotics on treatment outcome or significant interactions.The initial drug resistance profile affects the choice of antibiotic used and therefore this may have confounded associations between antibiotics used and treatment outcome.
The treatment outcomes in the UK differ from the standard WHO definitions [48], which means that it is difficult to compare outcomes directly with other countries.For example, in the absence of bacteriological or radiological data at the end of treatment we were not able to determine whether patients who completed treatment had been successfully cured.The relapse rate in the UK however is low, despite lack of evidence of cure [49,50] and during the study period only one case relapsed and was appropriately categorised as unsuccessful.
The treatment outcome classification used in the statistical analysis was based on an approach by Ditah et al. [15] which also differs from other studies.Deaths where TB is incidental to death are usually classified as an unsuccessful outcome but we chose to exclude them from the analysis as the eventual outcome, for example had the patients not died for another reason, was unclear.The UK however, has a strong vital registration system and we are therefore confident that these deaths were not caused by TB.
The partly retrospective study design prevented time to event analysis as we did not have the dates for all outcome categories.Data sources used to complete the questionnaire may not have been as accurate or complete as they would have been in an entirely prospective cohort.This may be particularly true for variables that can vary in definition such as DOT or variables relying on comprehensive notes such as changes in treatment.Future prospective studies or randomised control trials will likely provide stronger evidence for the association between individual drugs and treatment outcome, as well as allow for the investigation of the role of treatment duration on treatment completion or cure.

Public health and clinical implications
Since 2011, WHO guidelines for MDR treatment regimens recommend the inclusion of an oral bacteriostatic drug in combination with a fluoroquinolone and an injectable agent.This study, in addition to a recent meta analysis of 9,153 patients [37], supports these guidelines and therefore, provided the future MDR-TB population remains similar to our study population, we recommend that a bacteriostatic drug should be considered an important part of all MDR-TB treatment regimens in the UK, taking into account drug susceptibility.However due to side effects associated with bacteriostatic drugs their use should be managed with care.
During our study period the WHO guidelines in use did not recommend one fluoroquinolone over another but moxifloxacin was more widely used, as many clinicians believe that it may be more potent.Current WHO guidelines [5] recommend the use of later generation fluoroquinolones such as moxifloxacin and levofloxacin and these have also recently been shown to be significantly associated with successful treatment outcomes [37].
The failure to take account of drug sensitivity results appropriately as shown in our study, could reflect a lack of experience in treating MDR-TB, possibly due to its rarity in the UK.If geographical considerations prevent all cases being managed in specialist centres, outcomes may be improved by advice from clinicians in the national web-based MDR-TB advisory service hosted by the British Thoracic Society [51].
Since the majority of cases who were lost to follow up returned to their countries of origin, efforts should be made to engage with national TB programs overseas at an early stage in treatment to ensure optimised continuation of management.Alternatively, patients should be supported to complete treatment in the UK, especially if they are returning to resource-poor countries where TB treatment and, in particular, the supply of effective second line antibiotics may not be guaranteed.Referring detainees prior to deportation to a TB service dedicated to improving health in mobile populations, such as TBNet (part of the Migrant Clinicians Network, USA), has been shown to result in high treatment completion rates [52] and this option should be explored for the UK.

Conclusion
Our findings are in line with the international guidance for the use of a bacteriostatic drug in addition to an injectable agent and a fluoroquinolone for the treatment of MDR-TB.It is important to continue to monitor treatment outcomes of MDR-TB patients to improve treatment management policy.Further research should evaluate the role of DOT among MDR-TB patients in the UK.Patients should be given psychosocial support to improve treatment compliance.
Treatment stopped: Patient found to have stopped treatment (by choice) or for any other reason not mentioned below.

Table 4
Univariate analysis of demographic and clinical characteristics, social risk factors and comorbidities associated with successful treatment outcome in patients with multidrug-resistant tuberculosis, United Kingdom, 2004-2007 (n=191) Interruption of treatment for two consecutive months or more without medical approval or non-compliance reported by medical staff.
b Not estimable.c CI: confidence interval; HIV: human immunodeficiency virus; OR: odds ratio; TB: tuberculosis; UK: United Kingdom.

Table 5
Multivariate analysis of factors associated with successful treatment outcome (n=182) CI: confidence interval; OR: odds ratio; UK: United Kingdom.