Of the 1,004 HIV-negative participants in TBTC Study 22, there were 8 instances of treatment failure and 61 cases of relapse. Of these 69 cases of adverse TB treatment outcomes, isolates from 64 patients were successfully genotyped (Figure). Of the 930 study participants who did not experience treatment failure or relapse, 76 did not complete therapy and 172 did not complete follow-up. Of the remaining 687 patients, 296 (43%) were randomly selected as controls for this analysis; their 296 baseline isolates were successfully genotyped (Figure).

Demographic and clinical characteristics of cases and controls are provided in Table 1. As in the entire study cohort, white race, being underweight, pulmonary cavitation, bilateral pulmonary involvement, and 2-month sputum smear and culture positivity were associated with an increased risk for treatment failure and relapse among the participants of this case–control analysis (Table 1). Patients with active disease caused by a Beijing strain had an increased risk for treatment failure or relapse (15 [23%] of 64 cases vs. 42 [14%] of 296 controls; odds ratio [OR] 1.9, 95% confidence interval [CI] 0.9–3.6, p = 0.07). When analyzed as separate endpoints (Table 2), treatment failure was not associated with active disease caused by a Beijing strain (OR 0.9, 95% CI 0.1–7.2, p = 1.00), but relapse was significantly associated (OR 2.0, 95% CI 1.0–4.0, p = 0.04). Active disease caused by the Indo-Oceanic lineage was associated with a lower risk for treatment failure or relapse, although the small number of isolates from that lineage (n = 25) and the lack of any cases of treatment failure or relapse among patients with active disease caused by this lineage precluded further evaluation of this association.

Adjustment for the 5 clinical, radiographic, and microbiologic risk factors for treatment failure and relapse in the parent study had little effect on the OR of active disease being caused by a Beijing strain and relapse (adjusted OR 2.2, 95% CI 1.0–5.0, p = 0.05) (Table 3). Similarly, treatment assignment (rifapentine vs. rifampin) had little effect in a multivariate model into which this factor was forced (adjusted OR for active disease caused by Beijing strain 2.3; 95% CI 1.0–5.3, p = 0.04).

We next evaluated the effect of race/ethnicity on the association between relapse and active disease caused by a Beijing strain (Table 4). Asian–Pacific Islanders were not at increased risk for relapse in the entire TBTC Study 22 cohort or in this nested case–control analysis (OR 0.6, 95% CI 0.2–1.4, p = 0.23). However, Asian–Pacific Islanders who had active disease caused by a Beijing strain were at increased risk for relapse (OR 11, 95% CI 1.0–108, p = 0.04) (Table 4).

We further evaluated the association between relapse and active disease caused by a Beijing strain among race/ethnicity groups by using stratified analysis. In an analysis limited to Asian–Pacific Islanders, adjustment for other risk factors for relapse had little effect on the association between relapse and active disease caused by a Beijing strain (adjusted OR 15.8, 95% CI 1.3–192, p = 0.03) (Table 5). Among other race/ethnicity groups, there was no association between relapse and active disease caused by a Beijing strain (Table 5).

Using isolates from participants in a large prospective trial of supervised intermittent treatment, we found that active disease caused by a Beijing strain of M. tuberculosis was associated with a 2-fold increased risk for adverse TB treatment outcomes. The relationship between Beijing strains and treatment outcomes was driven by relapse because of their greater frequency and an apparent lack of any association between Beijing strains and treatment failure. Adjustment for the clinical, radiographic, and microbiologic risk factors for treatment failure and relapse and for treatment assignment in TBTC Study 22 had little effect on the association between active disease caused by a Beijing strain and the risk for relapse. In an exploratory analysis of the relationships of the effect of race/ethnicity on the association between Beijing strains and treatment outcomes, Asian–Pacific Islander patients were at increased risk for relapse if they had active disease caused by a Beijing strain, albeit with wide CIs around the risk estimate (adjusted OR 13.9, 95% CI 1.3–164).

Two other studies have evaluated the association between the Beijing genotype and relapse. A case–control analysis from Vietnam reported an adjusted OR of 3.2 for treatment failure or relapse among patients with active disease caused by a Beijing strain (16), and a cohort study from Singapore found an adjusted OR of 2.64 for relapse associated with Beijing strains (17). Our study provides additional support for the hypothesis that active disease caused by a Beijing strain is associated with increased risk for relapse (adjusted OR 2.2, 95% CI 1.0–4.9). The use of a nested case–control analysis of carefully characterized patients in a clinical trial enabled us to evaluate the relationship between Beijing strains and treatment outcomes in analyses adjusted for host factors previously associated with an increased risk for relapse. We were also able to eliminate effects of other possible confounding factors, such as treatment duration and adherence to treatment. Additionally, because reinfection could obscure the relationship between infection with a Beijing strain and the risk for relapse, we were able to remove cases of reinfection from this analysis. That adjustment for other risk factors for relapse, such as pulmonary cavitation and being underweight, had little effect on the association is additional evidence that active disease caused by a Beijing strain increases the risk for relapse.

After evolving in eastern Asia, the Beijing family of M. tuberculosis has spread around the world (26,29). Why might active disease caused by a Beijing strain confer an increased risk for relapse? Laboratory studies have suggested that Beijing strains may be better adapted for intracellular growth (30,31) and are more virulent in animal models of TB, perhaps by evading immune responses (32,33). Furthermore, an altered DNA repair enzyme in Beijing isolates that confers a mutator phenotype may confer greater flexibility to respond to adverse conditions (34), such as those posed by multidrug therapy.

The possibility that Beijing strains evade immune responses and are more virulent is generally borne out in studies with humans. The Beijing family is associated with extrapulmonary TB (35,36). That Beijing strains have been implicated in many outbreaks of TB suggests that they may be more efficiently transmitted or have an enhanced ability to progress to active disease than do other strains. In a study in Russia, active disease caused by a Beijing strain was associated with more severe radiographic manifestations of pulmonary TB (11). However, 2 smaller studies did not find an association between Beijing strains and radiographic severity of pulmonary TB (37,38). Because of its case–control design, our study cannot directly address the unresolved question of whether active disease caused by a Beijing strain is associated with radiographic severity of disease (such a comparison would require a cohort study design). However, adjustment for cavitation and bilateral pulmonary involvement did not affect the association between Beijing strains and relapse in our study. This finding suggests that the mechanism of the association between Beijing strains and relapse is not mediated by radiographic severity of disease.

Our study suggests that the increased risk for relapse associated with active disease caused by a Beijing strain may be related to the race/ethnicity of the patient; risk for relapse was higher for persons of Asian–Pacific Islander descent. There was also a trend toward increased risk for relapse among black patients infected with a Beijing strain, according to univariate analysis (OR 2.9, 95% CI 1.0–8.5; p = 0.07), although this trend was not retained in adjusted analysis (Table 5). It is notable that the population in which the Beijing genotype conferred the greatest risk for relapse was of Asian–Pacific Islander race/ethnicity and that the Beijing genotype evolved in eastern Asia. Coevolution of the Beijing genotype among persons of East Asian descent may have selected factors that contribute to transmissibility and a decreased response to therapy.

Our study has at least 4 limitations. First, the cohort from which this nested case–control analysis was drawn was composed of patients who enrolled in a randomized trial, who differed from the broader patient population at study sites. One clear bias in the study population is that TBTC Study 22 was limited to patients with drug-susceptible TB. Therefore, we cannot evaluate the association between Beijing strains and drug resistance. Second, despite the size of the study cohort, our case–control analysis had limited statistical power to detect associations, particularly in exploratory analyses of the relationships between M. tuberculosis genotype, race/ethnicity of the host, and the risk for relapse. Third, race/ethnicity was defined by using broad categories developed for the United States census and used in the TB surveillance system. However, these categories are crude approximations of the genetic background of patients. For example, the category Asian–Pacific Islander includes several distinct ethnic groups that have substantial differences in genetic backgrounds. We did not have access to human genetic material that would enable precise delineation of the genetic backgrounds of the trial’s participants. Fourth, we did not adjust for multiple comparisons, and our study should be viewed as an exploratory, hypothesis-generating analysis.

In summary, our study offers additional evidence that a common genotype of M. tuberculosis, the Beijing family, is associated with increased risk for relapse after completion of supervised short-course TB treatment. The finding that the population at greatest risk for relapse if they had active disease caused by a Beijing genotype was persons of Asian–Pacific Islander race/ethnicity suggests that the coevolution of this bacterial strain and the human population may have selected factors that confer a poor response to therapy.