In 2005, all 53 countries of the World Health Organization’s European region were invited to participate in the MDR TB surveillance project by Euro TB and the National Institute for Public Health and the Environment (RIVM). Countries were encouraged to provide both epidemiologic and genotyping data on MDR TB cases reported since January 1, 2003. However, countries that could provide only epidemiologic data were also included in the surveillance project. National surveillance institutions sent quarterly updates of individual and anonymous data on MDR TB cases reported since 2003 to EuroTB, according to a standardized data file specification (www.eurotb.org/mdr_tb_surveillance/pdf/DFS_NSI_short.pdf). Each patient had a unique record identifier by country. Common definitions of variables were used by participating countries, including demographic, clinical, and genotyping information. In 18 countries, the country of origin was defined as the country of birth. Because of confidentiality constraints, country of birth was not reported for some countries. Country of citizenship was used to qualify the origin of patients for 2 countries. Either country of birth or country of citizenship was used in 3 countries, and information on geographic origin was unavailable in 1 country. A patient had to be reported again (with a new identifier) when a new MDR M. tuberculosis isolate was obtained >24 months after the previously reported MDR TB isolate. However, the possibility of a case being reported twice in the same country or in 2 countries was low.

National laboratories sent insertion sequence (IS) 6110 restriction fragment length polymorphism (RFLP) patterns from available MDR TB strains collected since 2003 to the RIVM every 3 months, either as a Bionumerics bundle (Applied Maths, Sint-Maartens-Latem, Belgium) or as a scanned image (14). To be included in the project, laboratories had to perform drug susceptibility testing and participate in an international quality assurance program. For quality assurance, either participating laboratories represented supranational reference laboratories of the World Health Organization and the International Union against Tuberculosis and Lung Disease and participated in the yearly proficiency testing for isoniazid, rifampin, streptomycin, and ethambutol (15,16), or laboratories had their external quality assurance conducted by such a reference laboratory. For 1 country, drug susceptibility testing was performed in a laboratory abroad. More details (e.g., on drug susceptibility test methods) can be found in the 2006 EuroTB report (17).

European cluster information was communicated by the RIVM (Bilthoven, the Netherlands) to EuroTB on a quarterly basis. The final database was maintained by EuroTB and included 3 sections: A) demographic and clinical variables, B) cluster information on the basis of molecular patterns at the country level, and C) European cluster information. Data from sections A and B were matched according to a patient code; data from sections B and C were matched using the strain code attributed at the national level.

IS6110 RFLP was the recommended genotyping method (18) to report IS6110 RFLP patterns to RIVM. For most countries, genotyping was performed locally; for 2 countries, genotyping was performed at RIVM. A European cluster was defined as >2 MDR TB cases with M. tuberculosis isolates that shared identical IS6110 RFLP patterns in >2 countries (18). A national cluster included cases diagnosed in a single country. Determination of a national cluster (a cluster including MDR TB cases diagnosed in 1 participating country) could be based on 3 typing methods: IS6110 RFLP typing (18), spoligotyping (19), and/or typing that used mycobacterial interspersed repetitive units with variable numbers of tandem repeats (20).

In each laboratory, quality control of the molecular typing practices and computer-assisted analysis was included as described in the standardized methods (14). The DNA fingerprint patterns of MDR TB strains received at the RIVM were submitted to the molecular database managed by the RIVM (MDRTBase) by using the Bionumerics software (Applied Maths), and the database manager performed a quality check. The IS6110 RFLP patterns newly added to the MDRTBase were then compared with all other patterns stored in this database. The MDR TB strains were classified as clustered (included in a European cluster) or “unique” (not included in a European cluster). In addition to specifying European cluster reports, the database manager also specified strains belonging to the Beijing genotype family of M. tuberculosis by compariing them with the 19 reference RFLP patterns of Beijing strains described by Kremer et al. (21).

Factors associated with clustering or infection with a Beijing genotype strain were determined by the unadjusted and adjusted logistic regression model using SAS software (SAS Institute, Cary, NC, USA). The explanatory variables were demographic (sex, age, origin) and clinical (pulmonary vs. extrapulmonary TB, TB history) characteristics of MDR TB cases. XDR TB was studied as an explanatory variable for patients with drug susceptibility results meeting the definition of XDR TB (22) in both cluster and strain analyses. The proportion of clustering among Beijing strains was compared with the proportion of clustering among strains of other genotypes.

Because of the limited numbers per category, “unknown” categories were removed from all variables included in the model, as was the age category “<15 years” and the category “other” for the variable country of origin. Analyses, including clustering and the XDR variable, were performed in the univariate model only because of the variables’ strong association with origin in the Baltic states (χ² 97.1, p<0.001) and other countries of the FSU (χ² 20.5, p<0.001).

Twenty-four countries from the European Union and western Europe plus Croatia and Macedonia participated in the MDR TB project (Table 1). Epidemiologic data on 2,494 MDR TB cases reported from January 2003 through July 2007 were sent to EuroTB. The Baltic states (Estonia, Latvia, and Lithuania) reported 1,616 cases, representing 65% of total cases. In some countries, epidemiologic data were reported for <3 years (Germany, Lithuania, Poland, Romania, and Spain).

Laboratories from 19 countries sent 672 IS6110 RFLP patterns to the RIVM, 593 of which also had epidemiologic information and were linked to the epidemiologic database maintained by EuroTB (Table 1; Figure 1). Linkage of molecular and epidemiologic data was not possible for data from the United Kingdom and for data on a few cases from 6 other countries. The average proportion of MDR TB cases documented with both epidemiologic and genotyping data in 18 countries was 593/1,523 (39%), varying from 9% to 100%. After we removed data from 2 countries with low data completeness for genotypes (Germany and Lithuania), we found that the average proportion of MDR TB cases with both epidemiologic and genotyping data was 68%. Individual data on second-line drug tests for 2 to 5 second-line drugs included in the XDR TB definition (22) were reported for 1,302 cases by 16 countries from January 2003 through July 2007 (Figure 1).

Of the 672 MDR TB strains with IS6110 RFLP patterns reported, 288 (43%) were identified as belonging to European clusters (Table 2; Figure 1). The proportion of MDR TB strains included in European clusters was <20% in 7 countries (Croatia, France, Italy, Slovenia, Spain, Switzerland, and the United Kingdom). A large proportion (38%) of the RFLP patterns submitted to the molecular database originated from Estonia, where 80% (183/228) of the strains clustered. The number of MDR TB strains identified in national clusters was 170 among 330 strains (51%) reported in 13 of the 19 countries with genotyping data. In these 13 countries, the proportion of intercountry clustering was 28%.

Eighteen distinct European clusters (range 2–175 cases) were identified in 16 countries (Table 3). Clusters E0051, E0054, E0055, E0057, E0063, E0066, and E0069 were characterized by Beijing genotype strains, comprising 242/288 (84%) clustered MDR TB cases. The characteristics of the 4 largest clusters are described below.

Risk factors for clustering were analyzed for 521 cases documented with both molecular and epidemiologic data reported by 16 countries. Lithuania and Germany were excluded from this analysis due to incomplete genotyping data. Among these 521 cases, 250 (48%) belonged to European clusters and 271 (52%) had isolates with unique DNA fingerprint patterns (Figure 1). The results of univariate and multiple logistic regression analyses to determine factors associated with clustering are presented in Table 4.

Univariate analysis showed that origin from the Baltic states or other FSU countries, age category 45–64 years, and pulmonary tuberculosis were significantly associated with clustering. Among these variables, only origin from the Baltic states (adjusted odds ratio [OR] 25.1, confidence interval [CI] 9.9–64.0) or other FSU countries (adjusted OR 8.7, 95% CI 3.2–23.2) remained strongly associated with clustering in the multivariate model. The proportion of XDR TB cases was significantly higher among clustered strains than among unique strains (OR 4.1, 95% CI 1.9–8.7).

Analysis to identify Beijing genotype strains was possible for 500 of the 521 MDR TB strains with both epidemiologic and genotyping data (data from Germany and Lithuania were excluded). Of these 500 strains, 284 (57%) were identified as the Beijing genotype and 216 (43%) were non-Beijing genotypes (Figure 1). Univariate analysis demonstrated that origin from the Baltic states or other FSU countries, age categories 45–64 years and >65 years, and pulmonary TB were significantly associated with strains of the Beijing genotype (Table 5). Among these variables, origin from Baltic states and other FSU countries remained strongly associated with Beijing genotype strains in the multivariate analysis. Proportions of clustering and XDR TB were significantly higher among Beijing strains than among other genotype strains, (OR 19.6, 95% CI 12.3–31.2, and OR 5.5, 95% CI 2.1–14.3, respectively). After exclusion of the largest cluster, E0051 (n = 166 cases with epidemiologic data), characterized by a Beijing strain, the association between clustering and the Beijing genotype was 5× less but still quite strong for 334 MDR TB cases with genotyping data (OR 4.5, 95% CI 2.7–7.6).