We prospectively collected data on all patients who had TB sputum smear–positive samples and for whom TB had been diagnosed during January 1, 2003–December 31, 2006, by the National TB Program by microscopic examination of >1 Ziehl-Neelsen–stained sputum smears (7). M. tuberculosis isolates were typed by spoligotyping, 15-loci variable number of tandem repeats (VNTR) typing, and partly by IS6110-based restriction fragment-length polymorphism (RFLP) typing (8). We collected sociodemographic data through structured interviews with individual patients about their households (defined as all persons who share the same floor and the same food).

Index case-patients were defined as all persons for whom TB had been diagnosed through December 31, 2004. From our database, we identified as household case-patients their household members for whom TB was diagnosed within 24 months after enrollment of the index case-patient. We compared the genotypes and DNA fingerprints of strains isolated from household case-patients with strains isolated from index case-patients in the same household. TB was diagnosed passively, i.e., at the time persons sought care at the study clinics because of symptoms.

Isolated strains of M. tuberculosis were defined as identical if spoligotype and RFLP and/or VNTR patterns, as applicable, were the same (differing by <1 locus) in the household and index case-patients. Mixed infections were defined as RFLP types with discordant spoligotypes or VNTR patterns with multiple alleles on >1 locus in 1 patient isolate. Genotypes were defined by spoligotyping as described by Brudey et al.; the Beijing genotype was defined as any isolate without direct repeat spacers 1–34 and >3 of the spacers 35–43 (9,10).

Data were entered into Epi Info version 6.04 (Centers for Disease Control and Prevention, Atlanta, GA, USA). Analyses were performed in Stata version 8 (StataCorp, College Station, TX, USA). Patients with negative cultures or cultures that grew nontuberculous mycobacteria were excluded.

Through December 31, 2004, a total of 1,589 patients were registered. We excluded 36 patients because of mismatching and 20 because of laboratory/technical errors; therefore, sputum specimens of 1,533 patients were cultured. After excluding 57 negative cultures and 34 cultures that grew nontuberculous mycobacteria, we included 1,442 (90.8%) index case-patients in the analyses. These patients had 4,141 household contacts >15 years of age. During 24 months of follow-up, >1 household case-patients were identified for each of 12 (0.8%) index case-patients, including 1 with 2 household case-patients. Household case-patients were not significantly associated with sex, age, family size, M. tuberculosis genotype, or smear grade of the index case-patient (data not shown). None of the 12 index and 13 household case-patients had mixed infections.

Of the 13 household case-patients, 1 (a 72-year-old man) had recurrent infection; 3 had infections were caused by Beijing genotype (23%). For 2 (17%) household case-patients (95% confidence interval [CI] 1.9%–45.4%), the infecting strain was similar to that from the index case-patient. The interval between diagnoses for index and household case-patients was (mean ± SD) 13.0 ± 2.8 weeks for identical strains, and 47.1 ± 33.1 weeks for different strains (p = 0.114).

Of TB case-patients who had had exposure to a TB patient in their household during the preceding 2 years, 17% harbored the same strain as the index case-patient. Even if we classified as similar 1 case pair with only 2 loci differences according to VNTR (no RFLP type available: patient 4 in the Table), i.e., if we assumed that this difference was caused by evolution of the VNTR pattern or a processing error, the proportion of infections acquired outside the household still would be 77% (10/13). Although we cannot exclude the possibility of transmission from another person in the same household for whom TB had not been diagnosed by the TB Program and included in our database, this finding suggests that in our study population, most TB cases resulted from transmission outside the household. These results differ from observations in low-incidence settings but are similar to those in high-incidence settings in South Africa, Gambia, and Malawi (11–14), even though the follow-up periods of these studies differed. The high proportion of cases resulting from transmission outside the household in those studies and ours may be explained by high exposure to different M. tuberculosis strains circulating in the community. Having common factors that determine the risk for breakdown of TB infection to disease also may play a role but is less likely in our study because HIV prevalence is lower in rural Vietnam than in South Africa. Alternatively, the large proportion of nonmatching genotypes within households could reflect specimen mislabeling. Although we excluded specimens that had been mislabeled at collection, additional mislabeling may have occurred in the laboratory. However, because mislabeling is expected to occur at random, for most of the nonmatches to be caused by errors, nearly half of all specimens must have been mislabeled, which seems unlikely.

We identified household TB cases for ≈1% of the index cases, which is less than the 6%–7% reported in studies of active case finding (1,14). This finding could reflect studies that included household case-patients of all ages with all forms of TB, whereas we included only persons >15 years of age who had smear-positive TB. In addition, we relied on self-reporting rather than on active TB screening. Although self-reporting may have limited the number of household cases we identified, it was unlikely to have affected the proportion resulting from household transmission. An exception should probably be made for young children, who because of less social mixing, may have higher probability of being infected within than outside the household. The shorter interval between diagnoses of index and household case-patients with identical strains, as well as the observation that the average interval for case-patients with different strains was approximately half the 2-year study period (i.e., consistent with random occurrence over time) supports our interpretation that the strains that differed between household and index cases were from another source.

In rural Vietnam, where HIV prevalence is low, most persons with secondary TB are infected by a source case-patient outside of their households. This finding argues against active TB case finding among household members as an effective method for improving case finding.