The Institutional Review Board of Harvard School of Public Health and the Research Ethics Committee of the National Institute of Health of Peru approved the study. All study participants or their guardians provided written informed consent, and children <18 years of age provided assent.

The study design and methods were previously described in detail (22). In brief, we conducted a prospective cohort study of household contacts of pulmonary TB patients in Lima, Peru. The study area comprised 20 districts inhabited by ≈3.3 million residents living in urban areas and in periurban, informal shantytown settlements.

During September 2009–August 2012, we identified patients >15 years of age who had received a diagnosis of clinically presumptive pulmonary TB at any of 106 participating health centers. We confirmed the microbiological status of their disease with either a positive sputum smear or mycobacterial culture. We also recorded the index patient’s sociodemographic data; baseline smear status; presence or absence of cavitary disease; tobacco and alcohol use; HIV status; time from symptom onset until initiation of treatment; and, for patients, with drug-resistant TB, the time from TB diagnosis until start of an effective treatment (i.e., a drug regimen deemed appropriate for the index patient’s drug-resistant profile).

Within 2 weeks after index patient diagnosis, we enrolled all consenting household contacts. We assessed baseline M. tuberculosis infection status with the tuberculin skin test (TST) in household contacts with no history of a positive TST or TB. Household contacts who had signs or symptoms of TB underwent clinical evaluation and, if indicated, initiated treatment under Peru's National Tuberculosis Program guidelines (23). We offered HIV testing to all study participants. In accordance with Peru’s National Tuberculosis Program guidelines, isoniazid preventive therapy (IPT) was offered to household contacts <19 years of age and persons with specified concurrent conditions. At the time of household contact enrollment, we collected age, sex, sociodemographic, and occupational information; height and weight; alcohol and tobacco use information; HIV status; self-reported diabetes mellitus; history and presence of M. bovis BCG vaccination scars; use of IPT; history of TB; and housing features. We repeated the TST 6 and 12 months after initial evaluation among household contacts who had previously tested negative; 2, 6, and 12 months after initial evaluation, we reevaluated them for pulmonary and extrapulmonary TB. We also accessed the medical records of health centers in the catchment areas to identify household contacts whose TB was diagnosed at a health center during the follow-up period.

Among 3,027 index patients with culture-positive isolates at baseline, 2,143 had isolates that also underwent whole-genome sequencing (WGS) using the Illumina HiSeq 4000 platform (Illumina, https://www.illumina.com) with a read length of 100–150 bp and >50-fold coverage. The raw sequence data were trimmed using the sickle package (24) and mapped to the H37Rv reference genome using the BWA-MEM algorithm (25). We used a coverage-based approach, specifically SAMtools (default settings) (26) and pilon (27), to identify single-nucleotide polymorphisms across the whole genome. We assigned a call as missing if the valid depth of coverage at a specific site was <8 reads, if the mean read-mapping quality at the site did not reach 9, or if none of the alternative alleles accounted for >85% of the valid coverage.

Among index patients for whom WGS data were available, we identified M. tuberculosis lineages and sublineages on the basis of a previously published barcode that differentiates the groups on the basis of single-nucleotide polymorphisms (28). For index patients without available WGS data, we genotyped their M. tuberculosis isolates using 24-locus mycobacterial interspersed repetitive units–variable-number tandem repeats (MIRU-VNTR) (29). We used the best-match algorithm from the MIRU-VNTRplus online database (http://www.miru-vntrplus.org) to identify the lineages of the typed samples by cross-referencing them with reference lineages stored on the database (30).

We considered 4 distinct outcomes: M. tuberculosis infection at baseline, time to TST conversion during the 12 months of follow-up, M. tuberculosis infection by 12 months of follow-up, and incident disease during the 12-month follow-up time. We classified household contacts as infected at baseline if they reported a history of TB, reported a previous positive TST result within 6 months after study enrollment, or had a positive TST (induration size >10 mm [>5 mm for HIV-positive household contacts]) at baseline. We considered initially TST-negative household contacts to have undergone TST conversion if their TST status converted from negative to positive or if TB developed during follow-up. We considered household contacts to be infected with M. tuberculosis by 12 months of follow-up if contacts were infected at baseline or had a TST conversion during the 12-month follow-up period (22). We considered household contacts to have co-prevalent TB if it was diagnosed within 2 weeks after the diagnosis in the index patient. If household contacts received diagnoses 2 weeks–15 months (1-year follow-up plus a 3-month buffer period) after index patient diagnosis, we considered incident TB to have developed in those contacts during follow-up. We based TB diagnosis among contacts >18 years of age on the same criteria we used for index patients; the diagnosis among household contacts <18 years was based on published consensus guidelines from an expert panel on classifying TB in children (31).

We made an a priori decision to analyze child household contacts (<15 years) and adult household contacts (>15 years) separately for several reasons. First, we considered child contacts to be a more representative age group for household-based TB transmission than their more mobile, socially active adult counterparts. Second, several previous studies suggest that Beijing strains more commonly affect children than adults (18–20). We used complete data analyses in all multivariate adjustment models and conducted all statistical analyses in R version 3.5.1 (https://www.r-project.org).

Child contacts exposed to Beijing strains were more likely than those exposed to non-Beijing strains to be infected at baseline (risk ratio [RR] 1.27 [95% CI 1.06–1.53]; adjusted RR [aRR] 1.23 [95% CI 1.02–1.50]) (Table 3) and by 12 months of follow-up (RR 1.20 [95% CI 1.08–1.35]; aRR 1.22 [95% CI 1.09-1.38]) (Table 4). The prevalence of M. tuberculosis infection at baseline (RR 0.99 [95% CI 0.91–1.08]; aRR 1.00 [95% CI 0.91–1.09]) and by 12 months (RR 1.00 [95% CI 0.96–1.05]; aRR 1.01 [95% CI 0.97–1.06]) did not vary by index patient M. tuberculosis strain in adult household contacts (Tables 3, 4). In the sensitivity analyses, we obtained almost identical results when we excluded household contacts who reported a history of TB, a previous positive TST result within 6 months after study enrollment, or a history of TB treatment (Appendix Tables 5, 6).

The TST status of children exposed to Beijing strains were more likely than that of children exposed to other strains to convert from negative to positive during 12 months of follow-up in both the univariate and multivariate analyses (hazard ratio [HR] 1.53 [95% CI 1.09–2.14]; adjusted HR [aHR] 1.65 [95% CI 1.17–2.33]). Exposure to Beijing strains had no differential effect in adults (HR 1.02 [95% CI 0.80–1.30]; aHR 1.03 [95% CI 0.80–1.33]) (Table 5; Figures 1, 2). In the sensitivity analyses, we defined TST conversion by an increment of >6 mm in TST induration size during follow-up, and the results changed by <5% (Appendix Table 7).

After we adjusted for potential confounders, we found the HR for TB among child contacts exposed to an index patient with Beijing strains compared with non-Beijing strains to be 1.45 (95% CI 0.77–2.7). We found no evidence that exposure to Beijing strains affected the risk for incident TB in adult contacts (aHR 1.06 [95% CI 0.64–1.77]) (Table 6; Figures 3, 4). The results persisted when we restricted the analysis to household contacts in whom TB was diagnosed within >30 days after index patient diagnosis (Appendix Table 8).

Although they lacked statistical significance, the aRRs for TB outcomes were higher for BCG-vaccinated children exposed to Beijing strains than for non–BCG-vaccinated children exposed to Beijing strains (aRR 1.01 [95% CI 0.69–1.49] for M. tuberculosis infection at baseline; aRR 1.21 [95% CI 0.90–1.63] for infection by 12 months; aRR 1.84 [95% CI 0.77–4.39] for TST conversion; aRR 1.96 [95% CI 0.41–9.39] for incident TB). This finding raises the possibility that BCG vaccination might be less efficacious in children exposed to Beijing strains than in those exposed to other strains (Table 7). Among BCG-vaccinated adult household contacts exposed to Beijing strains, the aRRs were 1.00 (95% CI 0.78–1.29) for M. tuberculosis infection at baseline, 1.02 (95% CI 0.89–1.16) for infection by 12 months, 1.00 (95% CI 0.48–2.06) for TST conversion, and 0.53 (95% CI 0.17–1.68) for incident TB (Table 7).