Tuberculosis · TransmissionTuberculosis Strain Discordance Among Household Contacts: What Does It Tell Us?

Tuberculosis clusters within households, and household contact investigation is a cornerstone of case finding. Yates and colleagues point out, however, that the proportion of disease among household contacts that results from recent transmission within the household, rather than from shared susceptibility or exposure elsewhere, remains unclear.¹

Molecular strain typing allows isolates from different patients to be compared, supporting or excluding a putative transmission link. As the paper notes, several studies have made a striking observation: index patients and their household contacts frequently carry discordant strains of the Mycobacterium tuberculosis complex. To quantify this, the research team updated an earlier systematic review, adding a further ten years of data, and used the results to estimate the proportion of all tuberculosis disease attributable to recent household transmission.¹

The authors searched MEDLINE, Embase, Global Health, Web of Science and POPLINE for English-language studies published between 1 January 1985 and 26 January 2024. Eligible studies described at least ten pairs of individuals with a history of household contact in which both had active tuberculosis and both isolates had been strain-typed by RFLP, MIRU-VNTR or whole-genome sequencing. Studies that collected residence data only for isolates of the same strain, or that preselected both index and secondary patients, were excluded. Two reviewers independently screened records and extracted data, and corresponding authors of included studies were contacted, several of whom reanalysed their data for inclusion.¹

Rather than use a generic tool, the investigators classified risk of bias according to the two factors they considered most likely to distort strain discordance. Studies using lower-resolution 12- or 15-locus MIRU-VNTR, or conducted during an outbreak, were rated high risk and excluded from the main analysis; studies using 24-locus MIRU-VNTR or RFLP were rated moderate risk, and those using whole-genome sequencing low risk. Of the 30 included studies, six were at low risk, 20 at moderate risk and four at high risk.¹

Pooled estimates were obtained with an inverse variance heterogeneity model after a double-arcsine transformation, with random-effects and generalised linear mixed models as sensitivity analyses. Results were stratified by WHO-estimated national tuberculosis incidence at each study’s midpoint: low (<20 per 100,000 per year), moderate (20–100) and high (>100). To estimate the proportion of all disease attributable to recent household transmission, the review multiplied strain concordance by the proportion of all people with tuberculosis who had a history of household exposure to a putative index patient, a parameter derived from external data on the assumption of five-person households.¹

The 30 included studies came from 18 countries: nine from low-incidence, eight from moderate-incidence and 13 from high-incidence settings. Three used 12- or 15-locus MIRU-VNTR, 20 used 24-locus MIRU-VNTR or RFLP, and seven used whole-genome sequencing. The median sampling duration was 48 months, and the median number of strain-typed pairs per study was 29.¹

Using the inverse variance heterogeneity model, the authors report an overall pooled prevalence of strain discordance of 0.47 (95% CI 0.32–0.62). Most of the variability was due to between-study heterogeneity (I² 90.3%, 95% CI 69.0–95.3%). The funnel plot showed outlying estimates but, in the investigators’ assessment, did not suggest publication bias.¹

Incidence matters

Discordance differed significantly by setting (p=0.003), with pooled estimates of 0.30 (95% CI 0.25–0.34) in low-incidence settings, 0.25 (0.15–0.36) in moderate-incidence settings and 0.59 (0.41–0.76) in high-incidence settings. Heterogeneity was also evident according to the criteria used to preselect index patients (p=0.0010), although some strata contained few studies. There was no evidence of heterogeneity by strain-typing resolution (p=0.34), duration of follow-up (p=0.48) or risk-of-bias category (p=0.34), and conclusions were unchanged in sensitivity analyses using random-effects or generalised linear mixed models, or excluding studies from settings where households were large compounds.¹

The reasons for discordance, the paper explains, differ by context. In low-transmission settings, reactivation of remotely acquired infection, particularly among migrants, often explains discordant strains. In high-transmission settings, discordance mainly reflects independent acquisition of M. tuberculosis outside the household.¹

Heterogeneous infectiousness is proposed as one contributor: a small subset of people with tuberculosis produce disproportionately large numbers of infectious aerosols, and people generally have more contacts outside the home than within it, so the most infectious individuals are likely to transmit more often in the community. Differences in social contact networks and in time to diagnosis and effective treatment amplify this effect.¹

Assuming five-person households, Yates and colleagues estimate that the median proportion of all tuberculosis disease attributable to recent household transmission is 10.8%, with study-specific estimates ranging from 1.4% to 21.1%. This proportion remained low across a range of assumptions about household size; in sensitivity analyses, it exceeded 20% only when ten-person households were assumed. Average household size is approximately five in low-income countries and 2.5 in high-income countries. Importantly, the review emphasises that this conclusion is driven largely by the low proportion of people with tuberculosis who have been exposed to a putative household index case, and is therefore robust to misclassification of strain discordance.¹

When set alongside other evidence cited in the paper, the findings are consistent: specifically, a modelling study from a South African township estimating that 84% of infections occur outside the home, tuberculin skin test surveys suggesting that fewer than 20% of infections are attributable to household exposure, and a large molecular epidemiological study from England estimating that 3.9% of tuberculosis disease resulted from recent household transmission.¹

Clustering without transmission

This does not mean that tuberculosis fails to cluster within households; it clearly does. At a population prevalence of 1%, the investigators calculate that the chance of at least one of four contacts in a five-person household also having tuberculosis would be 3.9%, falling to 0.04% at a prevalence of 0.01%. The observed proportions are far higher: 11.8% in low- and middle-income countries and 11.5% in high-income countries. Because only part of this excess can be explained by transmission within the household, the remainder probably reflects shared determinants of susceptibility, such as overlapping contact networks, poor nutrition and HIV-associated immunosuppression. Unless these vulnerabilities are addressed, the paper argues, excess risk will persist after the index patient is no longer infectious.¹

The study’s strengths, as set out by its authors, are the number of household patient pairs, their geographical diversity and the stringent inclusion criteria. Missed cases, undiagnosed mixed infection or misattribution of the index case should not, in their view, systematically bias estimates of discordance, particularly as most households contained only one patient pair.¹

Several limitations are also acknowledged. Substantial heterogeneity remained after stratification, probably reflecting unmeasured differences in household composition and use of public space, and restriction to English-language publications may have introduced language bias. Because M. tuberculosis has a slow molecular clock, strain typing alone cannot define transmission networks or distinguish household transmission from two members independently acquiring the same strain from a common source outside the home; both would tend to overestimate household transmission. The attributable fraction estimates relied on external data, and treating pairs from the same household as independent may have marginally overstated precision.¹

Two groups need particular care. Tuberculosis in young children is usually diagnosed clinically, leaving no isolate for strain typing, and the review suggests that because preschool children have fewer contacts outside the home, a higher proportion of their disease could be attributable to household transmission. The review also could not account for the intensity of exposure: household members who sleep closer to the index patient are more likely to be infected, so discordance is likely to be less common when index and secondary cases shared a room during the infectious period.¹

1. Yates, T. A., Chipinduro, M., Aldridge, R. W., Khozoee, B., Winter, J. R., Tanser, F., Abubakar, I., & Kranzer, K. (2026). Mycobacterium tuberculosis complex strain discordance among people with tuberculosis and a history of household contact: A systematic review and meta-analysis. The Lancet Microbe. Advance online publication. https://doi.org/10.1016/j.lanmic.2026.101511