Multiple Introductions of Multidrug-Resistant Tuberculosis into Households, Lima, Peru

Data on household transmission are needed to develop optimal treatment.

T he discovery and use of discriminating genetic markers such as IS6110 restriction fragment length polymorphisms (RFLPs), spacer oligonucleotides (spoligotyping), and mycobacterial interspersed repetitive unit-variable number tandem repeats (MIRU-VNTRs) (1) have improved our understanding of the transmission dynamics of tuberculosis (TB) (2,3). Genotyping studies, in which strains with matching sets of markers are considered potential members of a single transmission chain, have demonstrated that recent transmission plays a major role, even in low-incidence settings (4,5); that persons with recurrent episodes of TB may be having reinfection rather than relapse (6)(7)(8); that persons may be infected by >1 isolate of Mycobacterium tuberculosis at the same time (9)(10)(11); and that transmission may occur in casual social settings (12).
Molecular epidemiologic studies have also demonstrated that secondary cases among close associates of known case-patients are not always members of the same chain of transmission, i.e., that infection may have been acquired from independent sources (13). Molecular investigations of households of multiple TB patients showed that cohabitating TB patients may be infected with distinct isolates of M. tuberculosis (14)(15)(16). For example, in 2 suburbs of Cape Town, South Africa, which have TB notifi cation rates of ≈320 cases per 100,000 population, researchers found that less than half (46%) of secondary TB cases within households had a TB isolate that matched an isolate from another case within the household by RFLP (16). Overall, <1 (19%) in 5 new TB cases occurring in these communities was the result of within-household transmission.
Although studies have shown that household contacts with TB are likely to have acquired infection independently in high-incidence settings, there are no published estimates of the probability that 2 household members with multidrugresistant TB (MDR TB: resistance to at least isoniazid and rifampin) share a similar genotype and are members of the same transmission chain. Molecular epidemiologic data from households with >1 MDR TB case can help shed   Multiple Introductions of  Multidrug-Resistant Tuberculosis into Households, Lima, Peru light on the transmissibility of highly drug-resistant disease and also help guide public health policy. For example, international guidelines for the management of known contacts of MDR TB patients recommend an empirical drug regimen based either on the drug-resistance profi le of an isolate from the suspected index MDR TB casepatient or on the most common drug-resistance pattern in the community while drug sensitivity tests are pending (17)(18)(19). A better understanding of the relative importance of intrahousehold or community transmission may help to inform the choice of empirical regimen. Despite a decreasing overall incidence of TB in Peru of ≈3.7% per year since 1996, the incidence of MDR TB has increased by ≈4.5% over the same period (20). The increasing incidence of MDR TB in densely occupied urban communities of Lima, Peru, poses obvious challenges for TB control. We report a molecular epidemiologic study within households in Lima in which >1 person received a diagnosis of MDR TB. We used spoligotyping and 24-loci MIRU-VNTR typing (21,22) to identify households that have had >1 introduction of MDR TB, and we explored the association of household factors with these multiple introduction events.

Study Setting, Participants, and Data
The estimated incidence of TB in Lima, Peru, is >130 cases/100,000 persons; this estimate masks substantial heterogeneity in the actual distribution of TB within this large metropolitan area where poor areas often experience several-fold higher local incidence of disease than higherincome areas (23). For example, in 2000 in northern metropolitan Lima (population 3,186,199), the incidence of active TB was 232 cases/100,000 persons (24). A nationwide survey in 2006 reported that 5.3% of all new cases and 23.6% of retreatment cases were MDR TB (25). Since 1996, Partners in Health and Socios en Salud Sucursal Peru have worked with the Peru Ministry of Health to implement a program to treat patients with active MDR TB by using supervised, individualized, antimicrobial drug regimens delivered on an ambulatory basis (26)(27)(28).
We previously reported the TB incidence in a cohort of household contacts of the patients treated for MDR TB (29). A household was eligible for inclusion in the study if >2 members had been treated for MDR TB by this program during 1996-2004, and if >1 MDR M. tuberculosis isolate obtained from each person was available for analysis. All available (pretreatment and ongoing treatment) MDR isolates from patients in eligible households were included in this analysis. Demographic data, drug-susceptibility test results, and information about the physical condition of the household structure were abstracted from the electronic records of the MDR TB program. This study was reviewed and approved by the Committee on Human Studies of the Offi ce of Research Subject Protection of Harvard Medical School.

Laboratory Methods and Drug-Susceptibility Testing
Drug-susceptibility testing and genotyping by using MIRU-VNTR and spoligotyping were performed by the Supranational Reference Laboratory at the University of Massachusetts Medical School. A standard agar plate proportion method was used for drug-susceptibility testing of M. tuberculosis isolates. The fi rst-line and second-line drugs tested were isoniazid (0. . We only included drugs to which resistance had been tested for >70% of isolates in the study.
An ABI Thermal Cycler 2720 (Applied Biosystems, Foster City, CA, USA) was used for PCRs. Initial denaturation at 94°C for 5 min was followed by 35 cycles of denaturation at 94°C for 30 s, annealing at 62°C for 30 s, and elongation at 70°C for 45 s; and a fi nal extension step at 72°C for 10 min. M. tuberculosis H37RV DNA and sterile distilled water were included in each test run as positive and negative controls, respectively.

Spoligotyping
Mycobacterial DNA was prepared by using the same thermolysis protocol as for MIRU-VNTR typing. For DNA amplifi cation, 0.15 μL Tth polymerase (5 U/μL; Roche, Pleasanton, CA, USA) was added to 50 μL of PCR mixture, and the following amplifi cation profi le was used: 3 min at 96°C; 35 cycles for 1 min at 96°C, 1 min at 55°C, and 30 s at 72°C; and 5 min at 72°C.
Spacer oligonucleotide typing was performed by using the Multianalyte Profi ling System (Luminex Inc., Austin, TX, USA). The procedure was conducted according to the protocol reported by Cowan et al. (31) with adaptations for a 96-well format. Fluorescence signals indicating hybridization strength were analyzed by using Bio-Plex Suspension Array System Instrument Luminex 100xMAP Technology (Luminex Molecular Diagnostics Inc., Toronto, Ontario, Canada) and the Bio-Rad BioPlex Manager Program version 4.1.1 (Bio-Rad Laboratories, Hercules, CA, USA). Lineage and the shared type for each isolate were assigned based on matching the spoligotype patterns with those listed in the SpolDB4 database (32).

Identifi cation of Multiple Introductions of M. tuberculosis into a Household
Households were classifi ed as having evidence of repeated introduction of TB from the community if isolates from >2 patients with MDR TB within 1 household had different molecular genotypes. Supply et al. proposed a standard approach for characterizing the relatedness of M. tuberculosis isolates by spoligotyping and 24-loci MIRU-VNTR. They found that the combination of these methods (which requires including >15 of the most diverse loci for MIRU-VNTR analysis) has comparable discriminatory power to IS6110 RFLP typing (22). We present minimum and maximum estimates of the proportion of households judged to have evidence of multiple TB introductions on the basis of spoligotyping and MIRU-VNTR genotyping data.
We also examined a classifi cation approach recently used by Narayanan et al. (7). Nonmatching strains are defi ned as those strains with >1 spoligotype spacer or >1 MIRU-VNTR locus difference. Enabling different degrees of stringency in calling 2 (or more) strains a match refl ects our underlying uncertainty about how rapidly spoligotypes and MIRU-VNTR genotypes change because of mutations at marker loci during the natural history of disease and through chains of transmission that may span decades.

Identifi cation of Reinfection Events
We genotyped all available MDR isolates of patients within study households. Among participants from whom >2 isolates were available, we identifi ed episodes of reinfection on the basis of differences in genotypes.
We used a similar approach for comparing genotypes for identifying episodes of reinfection and repeated household introduction.

Statistical Analysis
SAS version 9.2 (SAS, Cary, NC, USA) was used for statistical analysis. We performed standard nonparametric tests for assessing univariate associations between household-level factors and the probability of repeated introduction.

Results
We identifi ed 105 households in which >1 MDR M. tuberculosis isolate was available from each of >2 different household members. In total, 391 MDR isolates from 236 persons were available for molecular typing. Spoligotyping and MIRU-VNTR analyses were successfully completed on samples from >2 participants from 101 (96%) of these households. These analyses resulted in a set of 384 (98%) isolates from 232 (98%) persons. Characteristics of persons and households included in the study are shown in Table  1. There were an additional 142 households for which we knew of >2 patients with MDR TB, but for whom M. tuberculosis specimens were no longer available for genetic analysis. No statistically signifi cant differences in size, density, or age distribution of members were found between the households that were included and those not included in this study.
Of 384 isolates, 228 (59%) were tested for susceptibility to a suffi cient number of second-line drugs to identify extensively drug-resistant M. tuberculosis strains (MDR plus additional resistance to a fl uoroquinolone and a second-line, injectable antimicrobial drug [either kanamycin, amikacin, or capreomycin]). Thirty-one (14%) of these 228 isolates were confi rmed as extensively drug resistant and were obtained from 15 patients, none of whom were living in the same household.

Multiple Introductions of MDR M. tuberculosis into Households
Using a permissive defi nition of matching in which we included strains that differed by 1 spoligotype spacer to be matched, we estimated that 10 (10%) of households had distinct MDR isolates and showed evidence of repeated introduction. The strictest defi nition of matching, which required exact matches in spoligotype and at all 24-loci of the MIRU-VNTR analysis, showed that 38 (38%) of households had evidence of repeat introduction of MDR TB from the community (Figure). Using the approach of Narayanan et al. (7) for identifying nonmatching strains (pairs with >1 spoligotype spacer or 1 MIRU-VNTR locus difference), we classifi ed 16 (16%) households as settings with multiple introductions of MDR TB. The 16 households in which >2 persons had an MDR M. tuberculosis isolate that was different from that obtained from another person in the household, according to the defi nition of Narayanan et al. (7), are shown in online Appendix Table 1 (www.cdc.gov/EID/content/17/6/969-appT1.htm). Seven of these households also had evidence of within-household transmission of MDR TB. Closer inspection of spoligotypes isolated from these households indicated that 6 of the 16 households, although failing to meet the proposed criterion for matching, had similar isolates (households 112, 192, 557, 960, 263, and 645). If these 6 households are classifi ed as having evidence of within-household transmission, our best estimate of the number of households with evidence of multiple introductions of MDR strains is reduced to 10 (10%). Under these criteria, the percentage of households with only evidence of probable within-household transmission is 90%.
We used the 10 households as our most conservative set of households with evidence of multiple introductions of MDR stains and searched for household factors that were associated with multiple introduction events. We did not fi nd any signifi cant associations; specifi cally, the size and density of households, the quality of the household structure, and time span over which isolates were accrued from households all appeared to be unrelated to multiple introductions ( Table 2). In addition, no signifi cant difference was found in the number of drugs to which the isolate from the fi rst patient was resistant between households that had repeated introduction (mean 5.1 drugs) and households that had evidence of probable within-household transmission (mean 5.3 drugs; p = 0.75).

Evidence of MDR Reinfection
Ninety persons had >1 MDR TB isolate available for analysis. Using the defi nition of matching strains of Narayanan et al. (7), we found that 5 (6%) of these persons had 2 distinct strains of MDR M. tuberculosis during the period of follow-up and the remaining 85 (94%) showed repeated isolation of the same MDR strain (online Appendix Table 2, www.cdc.gov/EID/content/17/6/969-appT2.htm). Closer inspection of the isolates available from these 5 persons showed that 1 person (a 20-year-old man) from household 977 may not have been reinfected. Three isolates were available from this person. The fi rst isolate had a slightly different spoligotype than the 2 isolates subsequently obtained, but the MIRU-VNTR pattern was the same for all 3 isolates.

Discussion
In the absence of molecular epidemiologic data, secondary cases of MDR TB within a household are generally assumed to be the result of within-household transmission. In an area with increasing incidence of MDR TB (20), we found that 90% of household contacts of MDR TB index cases with active disease and drug-susceptibility test results had MDR TB (29). Our present study, in a subset of that cohort, used genotyping on the basis of spoligotyping and 24-loci MIRU-VNTR, which has been shown in other settings to have comparable discriminatory power to IS6110 RFLP (21). Our study shows that there was at least a 10% risk that a subsequent case of MDR TB occurring within the home of a known MDR TB patient was the result of transmission in the community rather than transmission in the household. This estimate represents a lower boundary of the contribution of community transmission to the appearance of secondary MDR cases within a home because matching strains within a household (which we would categorize as within-home transmission) may be caused by transmission from other sources in the community. Because circulating MDR strains were heterogeneous (Table 3), the magnitude of this bias may not be substantial.
We did not fi nd any easily measured household factors associated with risk for repeated introductions compared with within-home transmission. We had hypothesized that a high household density (persons/bedroom) or low quality of household structure may be associated with a higher probability of within-home transmission, conditional upon observing multiple cases within a home, but this hypothesis was not supported by these data. This fi nding may refl ect an absence of this association between household characteristics and risk for within-home transmission or, alternatively, it may refl ect the relatively small number of repeated introduction events that we observed and our limited power to test such associations. Accordingly, although our observations provide convincing evidence that repeated introduction of MDR TB into households occurs in these settings, further studies are needed to determine whether household factors, number of persons within these households, or strains present within these households are associated with an increased risk for withinhome transmission or repeated exposure in the community.
Genetic (33) or acquired susceptibility (34) to infection and disease may play a role in the accumulation of multiple TB cases within households. Because household members are likely to share genetic or environmental risk factors, or both, persons living with TB case-patients may be particularly likely to be infected and acquire disease whether they are infected by their household contact or in the community.
Our fi ndings provide evidence to support international guidelines for management of active TB among contacts of known MDR TB cases (17)(18)(19) because they confi rm that among strains from persons for which genotyping test results are available, <90% of household contacts with MDR TB were infected with the same strain as the index patient. Our fi ndings also highlight limitations associated with such policies. Because subsequent cases of MDR TB in a household may be caused by community transmission, policies that specify that apparent secondary case-patients receive therapy on the basis of the drug-susceptibility profi le of an isolate from the initial MDR TB patient may result either in effective drugs being needlessly withheld or in administration of drugs to which the strain is already resistant. This policy may result in acquisition of additional resistance to second-line drugs and prolonged opportunity for transmission of highly drug-resistant strains within homes and in the community (35,36).
These fi ndings support the use of rapid drugresistance tests to determine drug susceptibility profi les in known contacts of MDR TB patients. Molecular tests for resistance, such as line probe assays and cartridgebased PCRs (i.e., GeneXpert; Cepheid, Sunnyvale, CA, USA), are promising and have been endorsed by the World Health Organization for determining resistance to fi rst-line drugs (37). However, although new diagnostic tests in development also detect resistance to second-line drugs (38,39), these tests have not yet been optimized for use in guiding clinical care. New rapid phenotypic tests for resistance, such as the microscopic-observation drugsusceptibility assay, have also not yet been adequately tested under fi eld conditions for their capacity to be used in selection of tailored regimens for MDR TB (40). Known contacts of MDR TB patients should be a high-priority, ‡Substandard housing was defined as a dwelling with a dirt floor; walls made of straw matting, plastic, or plywood; a roof made of straw matting, plastic, or plywood; or no access to water in the home (data were not available for all households). high-yield study population for assessing the immediate utility of these new tools. A limitation of our study is that we cannot defi nitively distinguish the 2 mechanisms by which distinct MDR isolates may appear within households. First, household members may have been infected by different drugsusceptible strains in the community and acquired drug resistance through defi cient drug treatment. Second, household members may have been directly infected by different MDR strains in the community. Distinguishing between these 2 possibilities is essential because each would cause a distinct public health response. The fi rst mechanism suggests that detailed investigation of individual-level or household-level risk factors for acquisition of MDR TB was needed and would indicate a need for greater treatment support and supervision for patients with drug-susceptible disease. The second mechanism indicates a need to improve infection control in the community or to facilitate diagnosis and effective treatment for persons with MDR TB to reduce the duration of infectiousness. In most circumstances, we expect acquisition and transmission to contribute to the appearance of multiple cases of MDR TB within homes, and efforts to reduce the incidence of drug-resistant disease will need to address these factors.
Although we have insuffi cient data for previous TB episodes and treatment for persons in our study to exclude possible independent acquisition of MDR TB among household members because of inadequate treatment, our fi nding that >4 persons showed evidence of reinfection by a second (i.e., different) MDR TB strain provides evidence that there is a high risk for MDR TB exposure in this community. HIV status was known for only ≈50% of the persons in the study. Among those tested, only 3 (3%) of 102 were HIV infected and none of the 3 HIV-infected persons were among persons in households in which multiple introductions of MDR TB were detected. If coinfection with HIV was common, it would be expected to increase the probability of rapid progression to disease and lead to higher risks of multiple cases of unlinked disease within households. Because HIV co-infection was so rare, it is unlikely that this explains the study results.
Our results extend fi ndings from previous studies showing that a substantial fraction of cohabiting persons have independently acquired TB in the community (13)(14)(15)(16). In contrast to earlier studies that compared relative contributions of within-home and community transmission, all persons in our study had MDR TB. We found that although 90% of households had evidence of intrahousehold transmission, 10% had >2 independent introductions of MDR M. tuberculosis strains from the community. This fi nding suggests that the risk for community or extrahousehold transmission of MDR TB in Lima is high. Furthermore, it indicates that known MDR TB contacts initiating empirical treatment for MDR TB treatment require access to drug susceptibility testing to ensure that they receive the drugs to which their isolate is susceptible. National TB programs should be wary of applying empirical regimens on the basis of populationlevel drug susceptibility data without better understanding of the relative role of intrahousehold and community transmission of MDR TB.