The study population comprised patients recruited through HTI, a population-based active surveillance and molecular epidemiologic study that enrolls persons with TB reported to the City of Houston Department of Health and Human Services and Harris County Public Health and Environmental Services. In this study, we examined data from TB patients enrolled by HTI who had been reported during October 1995–September 2004.

Study participants were interviewed by trained interviewers using a standardized questionnaire designed to gather details of participant demographics, living situation, transportation, travel and social contacts, tobacco and alcohol use, illicit drug use, sexual behavior, history of incarceration, and personal medical history. Clinical information was supplemented through record review of all available inpatient, outpatient, and public health medical records.

We analyzed M. tuberculosis isolates with 3 molecular typing methods. First, isolates were characterized by an internationally standardized protocol of insertion sequence [IS] 6110 restriction fragment length polymorphism (RFLP) profiling (11) and analyzed with BioImage Whole Band Analysis version 3.2 software (BioImage, Ann Arbor, MI, USA). Second, for isolates with <5 IS6110 copies, the spacer oligonucleotide type (spoligotype) was determined by using a commercially available kit (Isogen Bioscience BV, Maarsen, the Netherlands) according to the manufacturer’s instructions. A member of the Beijing family was defined as having an octal spoligotype pattern of 000000000003771 (previously designated as S1) (12). Third, isolates were assigned to a principal genetic group on the basis of polymorphisms at codon 463 of the katG gene, which encodes catalase peroxidase, and at codon 95 of the gyrA gene, which encodes the A subunit of DNA gyrase (13). Susceptibility testing was performed by using BACTEC 460 radiometric culture system (Becton Dickinson, Sparks, MD, USA), at the hospital or reference laboratories supplying isolates to the HTI (14).

We assigned race/ethnicity on the basis of self-report. M. tuberculosis isolates were defined as clustering if they satisfied 1 of the following criteria: 1) >2 isolates from the same principal genetic group with identical IS6110 banding profiles containing >5 copies of IS6110 or 2) isolates with identical IS6110 banding profiles containing <4 copies of IS6110 and sharing the same spoligotype and principal genetic group. We evaluated clusters H03 and H33 as a single cluster because they differ in only 1 band by IS6110 RFLP.

We defined drug-resistant TB as an isolate resistant to at least 1 drug, including isoniazid, rifampin, ethambutol, pyrazinamide, or streptomycin. Multidrug-resistant TB was defined by resistance to isoniazid and rifampin, with or without resistance to other agents.

Questionnaire data were entered into a longitudinal database (initially Epi Info version 6.02b, Centers for Disease Control and Prevention, Atlanta, GA, USA; and subsequently Microsoft Access, Microsoft, Redmond, WA, USA) for storage and analysis. Statistical analyses were conducted with STATA version 8.0 SE (StataCorp., College Station, TX, USA). Descriptive analysis was initially conducted, and selected demographic, social, and clinical factors were studied by univariate logistic regression to test for differences between US-born blacks and US-born whites. We used US-born whites, as opposed to all others, as the comparison group because of the heterogeneity in the latter group, particularly in non–US-born Hispanics and Asians. Variables with p<0.2 in the univariate analysis were considered in the elaboration of a multivariate logistic regression model to identify with factors independently associated with black race in TB patients. Covariates were not included in the model if they substantially decreased the sample size because of missing data. In our final model, we used the variable injection drug use instead of drug use because of colinearity and a strong association with the former variable. Covariates that were no longer significant after adjustment for other significant variables in the full model were dropped from the final parsimonious model.

Individual participants gave written informed consent before enrollment in the study. We obtained parental consent and patient assent for participants 13 to <18 years of age, and a parent served as a proxy for the interview of participants <13 years of age. Proxy permission and interview also were used for participants who had died or were not mentally capable of giving informed consent. The study was approved by the Institutional Review Board of Baylor College of Medicine, Houston, Texas, and affiliated hospitals, and the University of Texas Health Science Center–Houston, Committee for the Protection of Human Subjects.

A total of 4,312 persons with TB were reported in Harris County during October 1995–September 2004. Of those, 3,662 (85%) agreed to participate in the study and were interviewed by trained HTI personnel. The study population comprised 1,318 (36.0%) US-born blacks, 1,220 (33.3%) Hispanics, 545 (14.9%) US-born non-Hispanic whites, 463 (12.6%) Asians, 85 (2.3%) foreign-born blacks, 20 (0.5%) foreign-born non-Hispanic whites and 11 (0.3%) others. Of participants interviewed, 3,064 (84%) had a positive culture for M. tuberculosis; 2,806 (92%) of those isolates underwent molecular characterization. Interviewed participants were more likely to be younger, black, HIV-seropositive, and M. tuberculosis culture–positive than were persons who refused to participate or could not be located (p<0.01 for first 3 comparisons and p = 0.03 for the last comparison).

Although the overall US incidence of TB decreased from 8.0 to 4.9 cases per 100,000 persons during 1996–2004, TB incidence in Harris County has remained fairly stable since 2000 after an initial decrease in 1996–1999 (Figure 1). Blacks consistently had the highest TB incidence in Harris County. In 2004, blacks accounted for approximately 56% of all TB cases among US-born participants despite representing only 18% of the Harris County population (Texas State Data Center and Office of the State Demographer, http://txsdc.utsa.edu/tpepp/txpopest.php).

US-born blacks were more likely than US-born whites to be younger, unmarried, and unemployed; have less education; reside in the inner city, earn less income; use public transportation; have renal and extrapulmonary disease; be HIV seropositive, M. tuberculosis culture–positive, and part of a cluster; and have drug-resistant M. tuberculosis (Table 1).

The 614 HIV-infected TB patients in our study reflected a 28% rate of co-infection for US-born blacks and 20% for US-born whites (p<0.001). Rates of HIV co-infection were markedly lower for Hispanics and Asians (9.1% and 1.7%, respectively). Pulmonary disease was the predominant clinical presentation in the study population; however, extrapulmonary disease was reported more frequently for US-born blacks than for US-born whites (p<0.01).

Resistance to at least 1 anti-TB agent was recorded for 80 (7.3%) isolates from US-born blacks, compared with 22 (4.5%) isolates from US-born whites (p = 0.04). Multidrug-resistant TB represented only 1% of all cases. Mortality rate (i.e., death within 180 days after TB diagnosis) was ≈10% and was similar for US-born blacks and US-born whites.

Among all 4,312 cases reported during the study period, 3,578 (including enrolled and nonenrolled participants) had at least 1 positive M. tuberculosis culture. Isolates from 3,227 (90%) were genotyped, including 1,984 (61.5%) that matched at least 1 other isolate. A total of 242 clusters were identified, and cluster size varied from 2 to 172 patients (mean 57.2, median 27).

Isolates of 1,765 (63%) of the 2,807 enrolled persons were clustered, including 822 (82%) of 1,007 isolates from US-born blacks and 349 (77%) of 448 isolates from US-born whites (p = 0.05). Rates of clustering were significantly lower for Hispanics and Asians (52% and 28%, respectively; p<0.01), probably underscoring a higher proportion of reactivation of latent infection in these patients. Cluster size for US-born blacks was larger (mean 69.3, median 45) than for US-born whites (mean 46.7, median 14; p<0.001). TB strains belonging to the Beijing family represented 26% and 29% of isolates from US-born blacks and US-born whites, respectively (p = 0.25). Two (H03/H33 and L16) of the 9 largest clusters containing at least 30 patients had more US-born blacks than US-born whites (Table 2, Figure 2).

In our final model (Table 3), factors independently associated with black race among TB patients included younger age, fewer years of education, inner city residence, use of public transportation, prison history, renal disease, and HIV seropositivity. US-born blacks were significantly less likely than U.S.-born whites to report current homelessness, smoking, alcohol abuse, injection drug use, and same-sex behavior. Although male sex, not being married, M. tuberculosis culture positivity, employment, and extrapulmonary disease were significant in the univariate analysis, they were no longer significant in the full model and therefore were dropped from the final parsimonious model. Because including certain covariates significantly decreased the sample size (and the representativeness of the dataset), we developed a second model (data not shown) by adding drug resistance and clustering to the final model. In this model, drug resistance remained significantly higher for US-born blacks than for US-born whites (odds ratio [OR] 1.76; 95% confidence interval [CI] 1.02–3.04, p = 0.04); however, clustering did not remain significant (OR 1.13, 95% CI 0.82–1.54, p = 0.46).