Recurrent tuberculosis in the Netherlands – a 24-year follow-up study, 1993 to 2016

Background Not all treated tuberculosis (TB) patients achieve long-term recovery and reactivation rates reflect effectiveness of TB treatment. Aim We aimed to estimate rates and risk factors of TB reactivation and reinfection in patients treated in the Netherlands, after completed or interrupted treatment. Methods Retrospective cohort study of TB patients with available DNA fingerprint data, registered in the Netherlands Tuberculosis register (NTR) between 1993 and 2016. Reactivation was defined as an identical, and reinfection as a non-identical Mycobacterium tuberculosis strain in sequential episodes. Results Reactivation rate was 55/100,000 person-years (py) for patients who completed, and 318/100,000 py for patients who interrupted treatment. The risk of reactivation was highest in the first 5 years after treatment in both groups. The incidence rate of reactivation was 228/100,000 py in the first 2 years and 57/100,000 py 2–5 years after completed treatment. The overall rate of reinfection was 16/100,000 py. Among those who completed treatment, patients with male sex, mono or poly rifampicin-resistant TB and a previous TB episode had significantly higher risk of reactivation. Extrapulmonary TB was associated with a lower risk. Among patients who interrupted treatment, directly observed treatment (DOT) and being an undocumented migrant or people experiencing homelessness were associated with a higher risk of reactivation. Conclusions Both patients who completed or interrupted TB treatment should be considered as risk groups for reactivation for at least 2–5 years after treatment. They patients should be monitored and guidelines should be in place to enhance early detection of recurrent TB.


Introduction
Globally, there is an increase in treatment success rates of tuberculosis (TB), with an estimated 83% treatment success of new patients in 2016 [1]. Treatment success is defined as the proportion of patients successfully treated or cured and is a proxy for the effectiveness of TB treatment. Unfortunately, not all TB patients treated successfully achieve long-term recovery [2]. Some patients may experience a re-occurrence of the disease, which is referred to as recurrent TB and can be caused by either a reactivation or a reinfection [2][3][4]. The rate of reactivated TB can be considered as a more accurate indicator for the quality and effectiveness of TB treatment and care on an operational level than treatment success.
DNA fingerprinting of the bacterial isolates has enabled the differentiation between recurrent TB in endogenous reactivation and exogenous reinfection [5,6]. Reactivation is defined as re-inflammation with an identical Mycobacterium tuberculosis strain as isolated in the initial episode of the disease [3,5]. Reinfection is the infection with a different M. tuberculosis strain than isolated in the previous episode(s) of the disease [3,5]. Higher rates of reactivation and reinfection were reported in countries with a high incidence of TB compared with countries with a low TB incidence. To illustrate, although with different time frames for follow-up, reactivation rates of 2.7-4.7% have been reported in countries with a high TB incidence compared with 0.3-1.0% in countries with a low TB incidence [6][7][8][9]. Likewise, higher reinfection rates of 1.5-2.8% have been found in countries with a high incidence of TB compared with 0.1-0.3% in countries with a low incidence of TB [4,[8][9][10][11][12][13][14][15][16][17][18][19][20].
Even though risk factors for reactivation of TB in countries with a low TB incidence have been described, previous studies have had limited power to explore the findings over an extended time period. Additionally, only a few studies explored the effect of interrupted treatment on the risk of reactivation [6,13].
The aim of this study was to estimate the rate of reactivation and reinfection of TB in patients diagnosed and Table 1 Characteristics during the first episode of the disease of all confirmed culture-positive tuberculosis patients with an available DNA fingerprint in both episodes, the Netherlands, 1993 to 2016 (n = 15,970)

Characteristics
Level Total cohort Reactivation Reinfection n Column % a n Row % a n Row % a treated in the Netherlands in the period 1993-2016 after treatment completion and treatment interruption. We used data of more than 16,000 notified patients with culture-positive TB. Additionally, risk factors for reactivation and reinfection among patients who completed the treatment were analysed.

Study design and setting
Data for this retrospective cohort study of TB patients notified during the period 1993-2016 were obtained from the database of the Netherlands Tuberculosis register (NTR) [21]. We retrieved demographic data (sex, age, date of birth, country of birth, year of entry in the Netherlands), previous TB, year of diagnosis of previous TB episode, type of TB drug resistance, bacterial culture results, DNA fingerprint results, risk factors for TB, HIV status and other co-morbidities, treatment regimen, treatment outcome and duration, and directly observed treatment (DOT). The variables for previous TB and treatment outcome were defined according to recommendation of the European Centre for Disease Prevention and Control (ECDC) and the World Health Organization (WHO) [22]. Patients with drug-susceptible TB were treated with a standardised regimen according to WHO recommendations. DOT was used in patients with a perceived high risk of non-adherence or drug resistance [23].

DNA fingerprinting
Since 1993, the National Institute for Health and Environment (RIVM) performs nationwide DNA fingerprinting of M. tuberculosis positive isolates. Until 2009, restriction fragment length polymorphism (RLFP) was used to identify the specific insertion sequences IS6110. From 2009 onwards, the 24-locus variable number of tandem repeats (VNTR) has been used. Isolates from 2004 until 2009 were re-typed retrospectively by the VNTR method. The analysis for both IS6110 RFLP and VNTR typing was conducted using Bionumerics software, version 5.0 (Applied Maths, Saint-Martens-Latem, Belgium).

Definitions
In the Netherlands, a patient with bacteriologicallyconfirmed TB or with a clinical diagnosis requiring TB treatment is notified to the NTR. Patients with recurrent TB more than 2 months after the end of a previous treatment are registered as previously treated patients. We defined recurrent TB as a new episode of active TB in a patient, 2 months or longer after completing the treatment of the first recorded episode in the database, either due to reactivation or reinfection. For this study, reactivation was defined as two TB episodes caused by an identical M. tuberculosis strain. A reinfection was defined as two TB episodes caused by different M. tuberculosis strains. For the main analysis, fingerprints with one or more bands difference in the RFLP or one or more repeats difference in the VNTR were considered as different M. tuberculosis strains signifying reinfection. For the sensitivity analysis, fingerprints with more than two differences in bands in the RFLP or in repeats in the VNTR were considered as different M. tuberculosis strains, while differences of up to two bands/repeats were considered reactivation. The follow-up period for patients with a recurrence was the time between the date of the end of treatment of the first episode of the disease until the date of diagnosis of the second episode. For patients without a recurrence, the follow-up period was the time between the end of treatment until 31 December 2016 expressed in the number of person-years. Bars indicate the 95% confidence interval. Underlying data are in the Supplement Table S2.

Study population
Data on all culture-positive TB patients with a known DNA fingerprint pattern registered from 1993 until 2016 in the NTR were included. To identify patients with recurrent TB episodes, we traced potential duplicate cases in the NTR by sex, date of birth, country of birth, year of diagnosis, previous TB episode and year of entry in the Netherlands. For patients reported after 2005, the NTR record number of the previous episode was registered in the NTR database and could also be used for matching. In addition, for patients reported after 2010, recurrent episodes were recorded in the RIVM laboratory data system, which allowed us to match recurrent entries with NTR data. For all potential duplicate cases before 2005, RIVM experts checked if the RIVM laboratory numbers recorded in the NTR database applied to the same individual in the RIVM laboratory register.
For the cohort analysis, we excluded patients with a treatment failure, patients who completed their treatment after 31 December 2016, patients with an unknown treatment outcome or unknown treatment duration, patients who continued treatment outside the Netherlands, patients who died and all patients with multi-drug resistant TB or resistance against pyrazinamide or ethambutol.

Statistical analysis
We calculated reactivation and reinfection rates from the number of recurrent cases identified per 100,000 person-years followed-up (py), stratified according to treatment outcome i.e. completed treatment vs interrupted treatment. We calculated incidence rates for the first 2 years of follow-up, for 2-5 years of follow-up, and for > 5 years of follow-up. We performed univariate and multivariable Cox regression analysis to determine risk factors for reactivation and reinfection and calculated the unadjusted hazard ratios (HR) of potential risk factors. We performed a sensitivity analysis with the alternative definition of reactivation and reinfection. Characteristics considered as potential risk factors for reactivation and/or reinfection were: sex, age group, incidence in country of birth, DOT, type of TB, drug resistance, alcohol or drug use, experiencing homelessness or being an undocumented migrant, HIV or other comorbidities (such as diabetes mellitus), immunosuppressive medication (such as tumor necrosis factor (TNF)-alpha blockers use), previous TB treatment, TB treatment regimen, TB treatment outcome, treatment duration and adverse events following TB medication. We performed a stratified analysis of risk for reactivation among patients who completed the treatment and among those who interrupted the treatment. Adjusted HR were calculated for all variables with a p value of ≤ 0.20 in the univariate analysis. We used a stepwise backward elimination method for a multivariable model. In the final model, we considered risk factors with a p value of < 0.05 statistically significant and performed in SPSS version 25 for Mac (Chicago, United States). Mid-P exact test to calculate incidence rates and the 95% confidence interval (CI) and Chi-square test to compare rates were performed in Open Epi 3.01.

Ethical approval
Permission to use the NTR data for this study was obtained from the NTR registration committee. The second step in the matching process to verify the potential duplicate cases was done by the NTR data manager at RIVM, who is authorised to access person-identifiable data. Verification of the RIVM laboratory numbers in the RIVM laboratory register was done by authorised RIVM laboratory staff. Only anonymised data were provided to the researchers for the analysis. In the Netherlands, ethical approval is not needed for the use of anonymised data from the NTR.  Table 1.

Incidences
Using the main definition of reactivation as infection with a strain with an identical DNA fingerprint, a total of 141 (0.9%) patients were identified with a reactivation and 31 (0.2%) with a reinfection. Five patients had a recurrent infection with a strain with one or two differences in the DNA fingerprint. The risk of reactivation (or rate ratio) was 5.8 times higher (95% confidence interval (CI): 4.0-8.4) in patients who interrupted their treatment (incidence: 318/100,000 py,) compared with patients who completed their treatment successfully in the first episode of the disease (reactivation rate: 55/100,000 py). The reinfection rate (16/100,000 py) was lower than the overall rate of reactivation (71/100,000 py). The median time to reactivation was 1.1 years and to reinfection was 5.5 years. (Supplement Table S1 shows the incidence rates for reactivation and reinfection).
The highest incidence of reactivation was in the first 2 years after finishing treatment, both for patients who completed or interrupted their treatment. In that period, the incidence of reactivation was 228 per 100,000 py (95% CI: 178-288) for patients who completed their treatment and 1,798 per 100,000 py (95% CI: 1,227-2,548) for patients who interrupted their treatment. After 2−5 years after treatment, the incidence remained high for both patients who completed their treatment in the first episode of the disease (incidence: 57/100,000 py; 95% CI: 37-84) and those who interrupted their treatment (incidence: 432/100,000 py; 95% CI: 219-769) (Figure 1 and Supplement Table  S2 show incidence rates by duration of follow-up). Table 2 shows the risk factors for reactivation in patients who completed their treatment in the first episode of disease. In the multivariable analysis, males (adjusted (a) HR: 1.9; 95% CI: 1.2-2.9), patients with mono or poly rifampicin (R)-resistant TB (aHR: 6.6; 95% CI: 1.6-27.3) and patients with a previous TB episode before 1993 or outside the Netherlands (aHR: 3.2; 95% CI: 1.7-5.9) had a significant higher risk of reactivation. Extrapulmonary TB was associated with a lower risk of reactivation (aHR: 0.5; 95% CI: 0.3-0.9). Patients with comorbidity or who were HIV-positive also had a higher risk for reactivation, but the p value of the aHR was more than 0.05. provided in the supplement material. b Adjusted for treatment outcome, type of TB, drug resistance, homelessness/being undocumented and country of birth. c Three patients with unknown sex (excluded in the multivariable analysis). d TB incidence of country of birth for patients who were born outside the Netherlands. e Until 2005 drug sensitivity was only recorded when drug resistance was detected, hence all cases without registered drug-resistance were categorised as "probably susceptible". Statistical significant risk factors in the multivariate analysis are shown in bold.
In the sensitivity analysis with reactivation defined as an infection with a strain with > 2 differences in the DNA fingerprint, patients who were HIV-positive also had a significant higher risk for reactivation among patients who completed the TB treatment (Supplement Table  S4 displays risk factors for reactivation after treatment completion (reactivation <2 differences in the DNA fingerprint)). The other associations remained the same.

Risk factors for reinfection
Patients with mono-or poly-resistance against rifampicin (aHR: 11.9; 95% CI: 1.6-88.3) and patients who were experiencing homelessness or undocumented (aHR: 2.7; 95% CI: 1.2-5.8) had a significant higher risk of reinfection. Patients who were born in the Netherlands (aHR: 0.2; 95% CI: 0.1-0.7) or had extrapulmonary TB (adj. HR: 0.2; 95% CI: 0.0-0.6) on the other hand, had a significant lower risk of reinfection (Table 3). In retrospect, the one case with a reinfection with mono-or poly-resistance against rifampicin differed only one repeat in one locus in the VNTR. When repeating the analysis with the less strict definition of reinfection (≥ 2 or more differences in the RFLP or VNTR), the association with both treatment interruption and mono-or poly-resistance against rifampicin disappeared (Supplement Table S5 shows risk factors for reinfection defined as >2 differences in the DNA fingerprint).

Discussion
The aim of this study was to estimate the rate of reactivation and reinfection of TB patients diagnosed and treated in the Netherlands in the period 1993-2016 after completed and interrupted treatment and to identify risk factors for reactivation and reinfection. The overall rate of reactivation was 55 per 100,000 py in patients who completed their treatment successfully and 318 per 100,000 py for patients who interrupted their treatment. The rate of reinfection was 16 per 100,000 py. In our study we found 0.9% of patients with a reactivation and 0.2% with a reinfection. This is similar to studies from other countries with low TB-incidence for which, although with shorter time frames for follow-up, reactivation rates were reported of 0.3-1.0% [6][7][8][9] and reinfection rates of 0.1-0.3% [4,[8][9][10][11][12][13][14][15][16][17][18][19][20].
Despite the low proportion of patients that developed recurrent TB during the median follow-up period of 15 years, relatively high rates of reactivation were found in the first 5 years after treatment. In the first 2 years after treatment completion, the reactivation rate was 228 per 100,000, while 2−5 years after the completion of TB treatment, the rate of reactivation after treatment completion was still 57 per 100,000. These rates are higher than 50 per 100,000 which is the defined cutoff for a risk group for TB in the Netherlands [23] that should be targeted for active case finding. In particular, patients with sputum smear-or culture-positive pulmonary TB, males, patients with mono or poly drug-resistance against rifampicin and patients who experienced a previous TB episode had a high incidence of reactivation. Presently, ex-TB patients are not considered as a risk group in the Netherlands and not targeted for further screening or follow-up [23]. In view of our findings, this should be reconsidered, at least for patients with risk factors for recurrence. Among interrupters, in the first 2 years after treatment cessation the reactivation rate is extremely high with a rate of 1,798 per 100,000 py; it is evident that this group is most important to find and restart treatment.
The proportion of patients presenting reactivation by timing of treatment interruption are quite similar between those who were treated less than 3 months and those treated less than 6 months (Table 1). This was unexpected and is possibly because the registered duration of treatment did not reflect if dosages were taken each day, as this is not captured in the registration. As expected, we note a smaller proportion of reactivation in patients receiving treatment for at least 7 months. With respect to risk factors for reactivation and reinfection, we showed that males had a higher risk of reactivation after completing treatment, which is also observed in other studies [9,11,24,25]. The WHO has reported a male to female ratio of 1.7:1 among global TB notifications in 2015 [26]. This can explain the higher risk of reactivation in males in the current study.
Among patients who interrupted their treatment in the first episode, patients who received medication under DOT and were undocumented migrants or people experiencing homelessness had a higher risk of reactivation. A possible explanation for this finding is the fact that all patients get treatment support from dedicated TB nurses, but DOT is offered to selective patients with a perceived high risk to be non-adherent to treatment, patients who have recurrent TB or patients with multidrug-resistant (MDR) TB [23]. Those who interrupt TB treatment despite daily supervision and intensive support are likely to have been more non-adherent than other patients. This was also observed in the US [10]. This study confirms need for intensive patient support given on daily basis, to reduce the risk for recurrent TB.
Several studies found a strong association between reactivation of TB and drug resistance, especially multidrug resistance and isoniazid resistance [10,14,17]. We excluded MDR TB patients from our cohort analysis because they are treated with individualised regimens and of note, we did not find any recurrent cases among the excluded patients with MDR TB. We did not find an association of isoniazid mono-or poly-resistance with reactivation and/or reinfection. However, we did find that patients with mono/poly rifampicin resistance had a higher risk of reactivation after completing treatment and of reinfection. The three patients with recurrent rifampicin-resistant TB in our cohort were diagnosed initially before 2006 and all received a standard TB regimen. This may explain the high reactivation rate. Since 2007, patients with mono or poly rifampicin-resistant TB disease in the Netherlands are treated with an appropriate regimen containing second-line anti-TB drugs, based on the sensitivity pattern of the isolated strain as recommended by national and international guidelines [23,27].
The association of mono or poly rifampicin-resistant TB with reinfection can be explained by a too-strict definition of reinfection (≥ 1 loci difference in fingerprint results). When we repeated the analysis defining reinfection as more than two differences in the RFLP or VNTR, the association of rifampicin resistance with reinfection disappeared. After all, genetic drift does appear in some of the strains and finding a difference of one or even more than one VNTR loci may still mark infection by the same strain [28]. The strict definition of reinfection may have caused misclassification of reactivation. When we redefined five cases with potential misclassification of reinfection into reactivation and repeated the risk analysis for reactivation, we found the same associations with the risk for reactivation and the association with HIV infection became stronger (Supplement Table S4).
Several studies have reported a higher rate of recurrent TB among patients with HIV [10,14]. Reduced immunity due to the reduction of CD4 + T-cells may explain the higher risk of recurrent TB in patients with an HIV coinfection [28]. Studies from countries with a low incidence of TB, England and Wales, Spain and the US, also found an association between HIV and recurrent TB, but they did not distinguish between a reactivation and reinfection [10,29,30]. This is the first study in a low TB incidence country that studied the association between HIV, other comorbidity and reactivation. We found a significant association when we used the less strict definition of reactivation.
Patients born in the Netherlands had a lower risk of reinfection in this study. This reflects the infection pressure typical to the TB epidemic in the Netherlands. In the Netherlands, the TB incidence among foreign-born residents is more than 20 times higher than among the native population [23]. It has been shown that migrants from high endemic countries have a higher risk of TB in their host country for at least 10 years after arrival [31]. We therefore assume that the increased risk of reinfection is attributable to a high transmission in this population group.
Extrapulmonary TB was associated with a lower risk of both reactivation and reinfection. Previous studies have also found a similar association but did not differentiate between a reactivation and reinfection [4,[6][7][8]. The lower risk of reinfection among patients with extrapulmonary TB was neither described before nor explained in the literature. Therefore, more studies are needed to confirm and explain this association. However, extrapulmonary TB is known as a milder and more paucibacillary form of TB, what makes the disease easier to treat [32]. Preliminary data from a study into shorter treatment for minimal TB in children (SHINE trial) have shown that in children with non-severe TB, a standard TB treatment regimen with a continuation phase of 4 months is equivalent to a shorter regimen with a continuation phase of 2 months [33]. This could explain also why a lower risk of reactivation is found in adult patients with extrapulmonary TB. Our study showed that even after interrupting treatment, only 2% of the patients with extrapulmonary TB had a reactivation compared with patients with pulmonary TB (6%) (Supplement Table S3). This suggests that for patients with mild forms of extrapulmonary TB, shorter treatment regimens could be considered.
Our study had some limitations. Firstly, there was likely an underestimation of the patients with recurrent TB because we did not actively follow the patients included in the cohort. Therefore, we do not know if patients have died, or have left the country between the first and second episode of disease. Moreover, TB disease in patients with recurrent TB is not always confirmed with culture. We included only confirmed culture-positive patients with available DNA fingerprint data, which might have caused an underestimation of the reactivation rate. The low number of recurrent events may have resulted in unstable estimates and wide confidence intervals for the risk analysis for some of the factors explored. Nevertheless, strengths of this study are that it had a much larger follow-up period (maximum of 24 years) compared with previous studies (maximum of 18 years) [13] and larger cohort with available DNA fingerprint data in the first and second episode of the disease compared with other studies where the maximum was 8,084 patients [13]. This made it possible to determine risk factors for both a reactivation and reinfection.

Conclusions
In conclusion, during the first 2 years after completion of TB treatment reactivation rates are higher than 200 per 100,000 py and 65 times higher in patients who interrupt their treatment. Informing patients about the risk of recurrence and regular monitoring and follow-up of those with a higher risk of recurrence is important, at least in the first 2─5 years after treatment. Studies into shortening the TB treatment in the milder form of (extrapulmonary) TB should be undertaken to reduce the burden of treatment in the patient.