First-line tuberculosis drug resistance patterns and associated risk factors in Germany, 2008-2017

Background Drug-resistant tuberculosis (TB), especially multidrug-resistant TB (MDR-TB), poses a threat to public health. While standard surveillance focuses on Rifampicin and/or Isoniazid resistance, little is known about other resistance patterns. This study aims to identify predominant drug resistance (DR) patterns in Germany and risk factors associated with them in order to inform diagnostic and treatment strategies. Methods Case-based TB surveillance data notified in Germany from 2008–2017 were utilized to investigate DR and MDR-TB patterns for Isoniazid (H), Rifampicin (R), Pyrazinamide (Z), Ethambutol (E), and Streptomycin (S). Predominant patterns were further analyzed stratified by sex, age, country of birth, prior TB, and disease site. Multivariable logistic regression was conducted to determine risk factors associated with any resistance, MDR-TB, and complete HRZES resistance. Results 26,228 cases with complete DST results were included in the study, among which 3,324 cases had any DR (12.7%). Four patterns were predominant, representing about ¾ of all cases with any resistance (S: 814 [3.1%]; H: 768 [2.9%]; HS: 552 [2.1%]; Z: 412 [1.6%]). High proportions of S and H resistances were found among both German and foreign-born populations, especially those born in Eastern Europe, and were unexpectedly high among children (H: 4.3%; S: 4.6%). Foreign-born cases had significantly higher proportion of any resistance (16.0%) and MDR-TB (3.3%) compared to German-born cases (8.3% and 0.6%). Of 556 MDR-TB cases, 39.2% showed complete HRZES resistance. Logistic regression revealed having prior TB and being foreign-born as consistently strong risk factors for any DR, MDR-TB, and complete HRZES resistance. Conclusions DR patterns observed in Germany, particularly for MDR-TB were more complex than expected, highlighting the fact that detailed drug-testing results are crucial before incorporating HRZES drugs in MDR-TB treatment. Furthermore, the relatively high rate of H-resistance in Germany provides strong rationale against the use of only H-based preventive therapy for LTBI.

Introduction Tuberculosis (TB) is the ninth leading cause of death worldwide and the leading cause of death from a single infectious agent, exceeding even HIV/AIDS [1]. The World Health Organization 109 cases (3.0%), of which 4 were reported as having XDR-TB [15]. Even within the context of Germany's low TB incidence, however, it is important to remain attentive to epidemiological changes in drug-resistant TB in order to inform clinical decision-making on appropriate drug regimens [16].
As TB treatment has shifted from standardized to more individualized approaches [17][18][19] information on drug resistance patterns within the population has become ever more relevant and important. Furthermore, due to the high prevalence of MDR-TB among R-resistant cases, R resistance is considered as surrogate for MDR-TB and amongst the WHO-endorsed rapid drug susceptibility testing (DST) tools, one of the most frequently utilized tools tests only for R resistance [20][21][22][23]. However, this approach is questioned by some experts and, hence, additional DST for H, especially in low-incidence settings, is recommended [24,25]. Here too, information on drug resistance patterns is crucial to inform the utilization of different resistance diagnostic methods.
To our knowledge, with the exception of one study conducted by the European Centre for Disease Prevention and Control, covering a period from 2007 to 2012, there is currently no analysis available on drug resistance patterns in low TB incidence countries in Europe, such as Germany [16]. Other studies who have addressed this question either limited their analysis to certain population groups [26][27][28] or were conducted using surveillance data from high TB incidence countries with very different TB epidemiology [29][30][31]. As a result, this study aims to analyze German national TB surveillance data for the predominant drug resistance patterns for the five standard anti-TB drugs (Isoniazid-H, Rifampicin-R, Pyrazinamide-Z, Ethambutol-E, Streptomycin-S; HRZES) and the factors associated with them, in order to inform diagnostic and treatment strategies for TB patients in low incidence countries such as Germany. The objectives of this study are to determine (i) frequency and distribution of drug resistance patterns over a 10-year period and (ii) demographic factors associated with specific drug resistance patterns.

Data source and collection
TB is a notifiable disease in Germany [32]. District health authorities report, via their respective state health departments case-based data electronically to the national surveillance database (SurvNet@rki, [33]) of the Robert Koch Institute (RKI). The data comprises of, among others, information on case demographics, site of disease, prior TB disease, bacteriological findings, DST results, and treatment outcome.

Case inclusion and definitions
For study inclusion, a case had to meet the following criteria: fulfil the RKI reference case definition for active TB disease [34], be notified between the years 2008 to 2017, be culture positive, and have been tested for all five HRZES drugs and have DST results available. The cut-off date for the data extraction was March 1 st , 2018. As S was considered and used as a first-line TB drug until 2011 covered by the study period, it was included in the analyses to reflect drug resistance dynamics in Germany. Z was also included, despite inconsistent international guidelines regarding the reliability and quality assurance of its DST [35, 36] as interlaboratory tests in Germany revealed good performance; Z has become increasingly important in newer therapy regimens [37,38]. Information on second-line drugs could not be included in this analysis due to insufficient data completeness.
Cases <15 years were categorized as children; cases between 15 and 59 years were categorized as adults, and cases > = 60 years as elderly. Any drug resistance was defined as resistance against at least one of the HRZES drugs, any R and any H resistance was defined as resistance to at least R or H, respectively. MDR-TB was defined as being resistant against at least H and R. A case was considered a previous case, when a prior diagnosis of TB disease was notified.

Data analysis
For the purposes of this study, the following variables were extracted from the national surveillance database: age, sex, country of birth, prior TB diagnosis (yes/no; used as surrogate for prior treatment due to a much higher level of data completeness), main site of disease (pulmonary vs. extrapulmonary), presence of MDR-TB (yes/no), individual DST results for HRZES (resistant vs. susceptible). For children <15 years of age, country of birth (born in Germany vs. born outside Germany) of the father and/or mother were also extracted.
Drug resistance patterns and their distribution were first analyzed descriptively and presented as case counts and proportions stratified by sex, age-group, country of birth, prior TB, and site of disease. Trends in HRZES monoresistances and MDR-TB, including MDR-TB with additional resistances to first line drugs, were analyzed using the Chi 2 test for linear trends. For all other analyses, data were pooled over the entire study period. Case counts were compared using the Chi 2 test and the corresponding p-values specified.
Further exploratory analysis was then conducted using univariable logistic regression. Finally, multivariable logistic regression analyses were performed to determine the association between age, sex, country of birth, previous TB, and site of disease (independent variables) and the presence of any resistance, MDR-TB, and resistance against all five drugs (dependent variables). The variable age was categorized into the following three groups before inclusion into the models: children (<15 years), adults (15-59 years) and elderly (60+ years). This was conducted in order to elucidate the expected differences in drug resistance patterns between the different age groups due to birth cohort effects as well as effect modification by age. Cases with missing values for any of the included variables were excluded from the analyses. A p-value of <0.01 was considered significant for the descriptive analyses, while 95% confidence intervals were calculated for the logistic regression analyses. All data were analyzed in RStudio Version 1.0.153 [39]. All data were collected in accordance with the German 'Protection against Infection Act' ('Infektionsschutzgesetz') and data protection guidelines were strictly followed. Informed consent was deemed not necessary as data were fully anonymized before analysis.

Results
Between 2008 and 2017, 48,044 active TB cases that met the RKI reference definition were notified in Germany. 41,932 cases (87.3%) had a culture result available, of which 34,141 were culture positive (81.4%). Of these, DST results were available for all 5 drugs (HRZES) for 26,228 cases or 76.8% (Fig 1). These 26,228 cases were considered to be the final study population.
For those with available information, foreign-born cases comprised 57.1% of all included cases. Similarly, for those with available information, the median age of the entire study population was 45 years and 63.2% were male. Median age among foreign-born cases was 35 years and 62.3% were male (German-born: 58 years and 64.7%, respectively). 12.5% of all cases with information on prior diagnosis were previously diagnosed with TB. Further case characteristics are shown in Table 1.

Trends in drug resistance over time
The pooled national data revealed an overall proportion of any drug resistance of 3,324/26,228 (12.7%) in Germany (Fig 1). No significant overall changes could be observed for H, R, Z, and E monoresistance rates between 2008 and 2017 (p>0.01 for all; Fig 2), while S-monoresistance rates showed a significantly increasing trend over time (p<0.01, respectively) (Fig 2). However, the increasing trend was present until 2016 and the S-monoresistance rates declined in 2017. Overall, MDR-TB was found in 2.1% of cases in Germany but showed a significantly increasing trend over the study period (p<0.01; Figs 2 and 3). MDR-TB accounted for 40/2,891 (1.4%) TB cases in 2008 and for 47/1830 (2.6%) in 2017. Among MDR-TB cases, no clear trend was observed in the rates of resistance to all five drugs, but a remarkable increase from 2012 (10/54 (18.5%) of MDR-TB cases) to 2013 (46/87, 52.9% of MDR-TB cases) (Fig 3).

Resistance patterns overall
Among the 26,228 cases with complete DST results, 3,324 (12.7%) were resistant to at least one of the five drugs (any resistance) (Fig 1). Out of 31 possible resistance patterns, 28 could be identified in the surveillance data, of which 4 patterns accounted for more than ¾ of all drug resistant cases (for S: 24.5%; H: 23.1%, HS: 16.6%; Z: 12.4%; Fig 1). Three possible patterns were not represented (RZE, RZS and ZES) and one pattern was only found once (HZE). Overall, 556 (2.1%) cases had MDR-TB, of which 218/556 (39.2%) showed resistance against all five drugs, 143/556 (25.7%) were resistant to four drugs, 141/556 (25.4%) to three drugs and 54/ 556 (9.7%) were resistant to only H and R (Fig 1). Any resistance to R was present in 643/3,324 (19.3%) cases and 556 (86.5%) of them had MDR-TB, i.e. additional resistance to at least H. 87 R resistant cases had no accompanying H resistance (Table 2), accounting for 0.3% of all included cases and 13.5% of all cases with any R resistance.

Drug resistance according to sex
Within the study population, 9,

Drug resistance according to country of birth
When examining the resistance patterns by country of birth, cases born in Russia, Somalia, Kazakhstan, Turkey, India, Vietnam, Romania, Eritrea, Pakistan and Afghanistan accounted for the highest numbers of drug resistant cases. Cases born in Russia, Kazakhstan and Vietnam presented the highest proportions of any drug resistance with 33.2%, 28.2%, and 26.7%, respectively. For MDR-TB, once again cases born in Russia (14.1%) and Kazakhstan (8.2%) presented the highest proportions, followed by Somalia (3.8%). Because of the recent shift in country of origin of migrants in Germany from predominantly Eastern European countries towards countries from the Horn of Africa and the Middle East, country rankings were also analyzed in two 5-year timespans (208-2012 and 2013-2017).Between 2008 and 2012, Turkey, Russia, and Kazakhstan, had the three highest absolute shares of resistances in Germany. However, between 2013 and 2017, African countries became more dominant and cases born in Somalia and Eritrea, having the top two absolute shares of resistances, with Romania ranked third. Table 3 shows the overall proportions of any drug resistance, any H-resistance, any Rresistance, and MDR-TB for the top 10 countries of birth from 2008 to 2017.

Resistance patterns according to age and country of birth
Overall, German-born cases had a significantly lower proportion of any resistance than foreign-born patients (

Regression analyses
The final multivariable logistic regression results are presented in Table 6

Discussion
Our study is one of the limited number of recent studies worldwide to analyze national surveillance data on drug resistance patterns over a long period of time. Our analysis revealed an Table 3. overall any drug resistance proportion of 12.7% in Germany, among which four resistance patterns were clearly predominant and accounted for more than ¾ of all drug resistant cases inmonoresistance against S, monoresistance against H, combined resistance against HS and monoresistance against Z. Monoresistance against S was the most prevalent drug resistance pattern in cases born in Germany and outside Germany, as well as among all age groups, very likely due to the long historical usage. These findings confirm that the withdrawal of S as a first-line drug in 2012 in Germany [38] has minimal relevance for TB therapy due to high resistance levels against it in the population. As a result, it can be questioned whether systematic surveillance of S resistance is still essential.

Number of cases
Resistance against H presents another highly prominent drug resistance pattern among tuberculosis patients in Germany. Our analysis showed high proportions of H resistance among all subpopulations, but cases originating from Eastern Europe showed remarkably high proportions of resistance. This supports the use of either 3-4 months of combined H and R therapy or 4 months of R monotherapy for LTBI treatment in Germany [19]. We also found unexpectedly high proportions of S and H resistances among children. A potential explanation for this finding might be the presence of selection bias as bacteriological confirmation is challenging in paucibacillary childhood TB and diagnosis may, as a result, be intensified particularly in children with unknown infection sources and suspected drug resistance.

Table 4. The 4 predominant patterns in drug-resistant culture positive TB cases with DST results, by country of birth and age group (for cases with available information), Germany 2008-2017, Isoniazid (H), Rifampicin (R), Pyrazinamide (Z), Ethambutol (E), and Streptomycin (S).
Z monoresistance was another significant finding in our analysis. We saw a remarkably high proportion of Z resistance among the German-born elderly population (60+), being almost 2.5-times higher than the proportion in the foreign-born elderly. One hypothesis might be that elderly cases present higher proportions of prior, agriculturally-acquired infection with Mycobacterium bovis, which is inherently resistant to Z. In fact, of the 145 German-born elderly cases with monoresistance to Z, 95 (65.5%) were reported as having M. bovis as the species of infection. Apart from that, Z-monoresistance was 2.2-times higher in extrapulmonary cases in comparison to pulmonary cases. Some studies have previously indicated that infection with M. bovis more often leads to extrapulmonary TB disease [40,41]. As a result, species specification of TB infection is an important step to support appropriate design of treatment regimens. Between 2008 and 2017, 2.1% of all cases in Germany had MDR-TB. In comparison to the European average of 3.7% in 2016, Germany presents a low proportion of MDR-TB cases [7]. Nevertheless, in contrast to the European-wide trend [7], we observed a slight increase in MDR-TB cases notified over the last 10 years, the reasons for which are likely the increase in migration from high MDR-burden regions, but perhaps also better reporting by local health authorities. Especially of note is that we rarely observed "simple" MDR-TB, i.e. resistance to only H and R and MDR-TB cases mostly presented additional drug resistances. Approximately 40% of MDR-TB cases in our study were resistant to all 5 drugs (HRZES) and only 10% were resistant to H and R only. This supports the WHO recommendation to initiate MDR-TB treatment with four second-line anti-TB drugs and not to rely on any first-line drugs, until detailed DST results are available [18].
Cases with prior TB disease and foreign-born cases have a high risk of MDR-TB. Our analysis showed that cases with prior TB diagnosis had an almost six times higher odds of having MDR-TB and nine times higher odds of having resistance against all five drugs in comparison to new cases. A number of studies have analyzed risk factors associated with MDR-TB [42][43][44][45][46][47][48][49][50][51][52][53] for different populations around the world. All of them found prior treatment to be one of the most significant and consistent predictors for the presence of drug resistance. In our study, although we utilized prior TB diagnosis as a proxy for prior TB treatment, our results are consistent with these previous results. Nevertheless, MDR-TB, and especially resistance against all five drugs, was also prevalent in new cases and in children, confirming that MDR-TB strains are endemic in many settings and that MDR-TB is no longer only a matter of acquired drug resistance.
Cases born outside Germany had a 5.5-fold higher proportion of MDR-TB in comparison to cases born in Germany. Several studies have also found higher proportions of MDR-TB among migrants in comparison to the local populations [44,47,48,50,51]. Cases from Eastern Europe, especially Russia and Kazakhstan, accounted for a substantial proportion of cases with MDR-TB in our study. This is in line with findings from Falzon and colleagues [50] showing that MDR-TB is strongly associated with origin from the former Soviet countries. Surveillance figures further show that Eastern European countries account for 85% of the TB burden and 99% of the MDR-TB cases in the WHO European region [9,50]. However, due to the arrival of a substantial number of refugees and asylum seekers into Germany from 2014-2016, the migration pattern in Germany has undergone a changed. For example, African countries, among them countries with a high burden of drug resistant TB, such as Somalia and Eritrea, have recently become prominent in the drug resistance landscape, accounting for a considerable proportion of drug resistant cases. Between 2013 and 2017, cases from Somalia and Eritrea accounted for 11.2% and 7.2%, respectively, of all foreign-born cases with any resistance (2008-2013: 2.3% and 0.7%) and for 10.9% and 2.9% of all foreign-born MDR-TB cases (2008-2013: 1.0% for both). MDR-TB cases from countries in the Horn of Africa, such as Somalia and Eritrea, were also the subject of a large Europe-wide outbreak investigation in 2016, where the majority of the cases in the cluster were reported in Germany [54]. Nevertheless, the overall proportion of MDR TB remained stable at 2-3%, and we also found MDR-TB in cases born in Germany, especially among younger, German-born cases, which highlights the continued need for rapid DST and strengthened bacteriological confirmation among all age groups. Rapid DST techniques are meaningful, especially in patients with high risk of drug-resistant TB, in order to rapidly commence adequate treatment before phenotypic DST results become available. Recently, R resistance has increasingly been considered as a surrogate marker for MDR-TB [23]. Although R monoresistance is rare worldwide [38], it seems to be prevalent in Germany. 13.5% of R-resistant cases in our study did not have H-resistance and among those without accompanying H-resistance, about 80% of all R-resistant cases were monoresistant.
Together with the high proportion of H resistance in Germany and our finding that MDR-TB is rarely only HR resistance in Germany, our results raise some concerns about the possibility of incorrect diagnosis of drug-resistant TB in Germany if currently available rapid tests were to be used as the only diagnostic tests. Focusing on only R resistance or HR resistance, other important and relevant resistance patterns would be overlooked unless additional testing is conducted [55]. Although more resource-intensive, comprehensive genotypic testing using whole genome sequencing has shown high concordance with conventional phenotypic techniques for first line TB drugs and could, therefore, be another relevant rapid test technology in resource-rich countries in the future [55].
In our multivariable regression analyses, we found that being born outside of Germany and prior TB disease diagnosis were the only consistent risk factors associated with having any resistance, MDR-TB, and complete first-line resistance. In line with our findings, most studies do not report any significant association between sex and MDR-TB [42,43,45,56,57]. Nevertheless, Faustini, Hall and Perucci [44] report male sex to be a risk factor, while Liu et. al. [52] found an association between female sex and MDR-TB. Several other studies have found younger age to be associated with the development of MDR-TB [44,46,49,52,53]. We could not support these findings with the results from our multivariable analysis, but results from our descriptive analysis indicate a similar pattern.
Our study has a number of limitations. Our study is based on reported surveillance data and we did not have insight into the original laboratory results or into detailed DST information on the cases. Because our data includes a considerable number of migrant cases, who can be a mobile population group, our surveillance data is incomplete for a number of variables, especially on DST. Of the 48,044 notified active TB cases between 2008 and 2017 in Germany, 76.8% had DST results available for HRZES. Although this is above the European average, notified DST testing rates in Germany do not meet ECDC's target for TB elimination [7]. According to the ECDC, performing cultures and DST in at least 80% of the cases is a necessary step to achieve the elimination of TB in Europe [58].As a result, our lower DST testing coverage should be interpreted with caution. Moreover, DST for E and Z are less reliable than DST for H and R, especially in MDR-TB cases. Information on resistance against second-line drugs is limited and could therefore not be included in our analysis. Information on social determinants and behavioural risk factors such as HIV status, unemployment, alcohol and drug abuse, smoking, and diabetes is not reported at the national level of the the notification system and could also not be included in the risk factor analysis. Since data on prior treatment was incomplete, we utilized prior TB diagnosis as proxy, but this may mean that we may have missed relevant information on the type of prior treatment and its outcome.

Conclusion
Drug resistance patterns observed in Germany, particularly for MDR-TB cases, are far more complex than expected. Although the overall proportions of drug-resistance and MDR-TB are low, most MDR-TB cases in Germany present with additional resistances against other standard anti-TB drugs and almost 40% of MDR cases showed complete resistance against all five standard TB drugs (HRZES). As drug resistance patterns vary significantly in different subgroups, our findings highlight the importance of considering demographic characteristics of cases and knowing the patients' full drug resistance profile to tailor treatment for optimal outcome. This is especially significant as our study found higher than expected proportions of drug resistances and MDR-TB among children and German-born adults. Finally, our finding of considerably high H monoresistance rates is important towards informing LTBI treatment regimens in Germany.