Mycobacterium tuberculosis Cluster with Developing Drug Resistance, New York, New York, USA, 2003–2009

In 2004, identification of patients infected with the same Mycobacterium tuberculosis strain in New York, New York, USA, resulted in an outbreak investigation. The investigation involved data collection and analysis, establishing links between patients, and forming transmission hypotheses. Fifty-four geographically clustered cases were identified during 2003–2009. Initially, the M. tuberculosis strain was drug susceptible. However, in 2006, isoniazid resistance emerged, resulting in isoniazid-resistant M. tuberculosis among 17 (31%) patients. Compared with patients with drug-susceptible M. tuberculosis, a greater proportion of patients with isoniazid-resistant M. tuberculosis were US born and had a history of illegal drug use. No patients named one another as contacts. We used patient photographs to identify links between patients. Three links were associated with drug use among patients infected with isoniazid-resistant M. tuberculosis. The photographic method would have been more successful if used earlier in the investigation. Name-based contact investigation might not identify all contacts, particularly when illegal drug use is involved.

patterns. As the investigation continued, this defi nition was expanded and included patients whose isolates had identical spoligotype, 12-loci MIRU-VNTR results, and IS6110 RFLP patterns with ± 1 band.

Drug Susceptibility Testing
TB drug susceptibility testing (DST) was performed at the NYC Public Health Laboratory and the Wadsworth Center on initial M. tuberculosis isolates by using either BACTEC 460 or Mycobacterial Growth Indicator Tube 960 (Becton Dickinson, Sparks, MD, USA). A standard agar-proportion method with Middlebrook 7H10 media was used to confi rm resistance (16)(17)(18). If DST indicated isoniazid resistance, DNA sequencing of the catalaseperoxidase G (katG) and enoyl reductase A (inhA) genes (19) was performed at the Public Health Research Institute Tuberculosis Center.

Contact Investigation
Contact investigations were conducted per BTBC guidelines (20). For contacts of infectious index patients, staff assessed hours of TB exposure during infectious periods of patients, defi ned as the 12-week period before the patient began appropriate TB treatment (20,21). Infectious periods were extended to date of symptom onset if TB symptoms started >12 weeks before treatment began. Contacts having documented latent TB infection or TB symptoms were referred for medical evaluation and treatment.

Cluster Investigation
An investigation of patients with the same M. tuberculosis strain was initiated to identify chains of transmission within the cluster and uncover epidemiologic links between TB patients. An epidemiologic link between 2 patients indicated that patients were linked by person, place, or time. Defi nite epidemiologic links between patients required 1 of the following criteria: named another patient as a contact, had a common contact, reported being in the same location during a patient's infectious period, or recognized each other's names or photographs. Probable epidemiologic links indicated that patients were in the same location during the same date range regardless of the infectious period of either patient or that 1 patient recognized another's name or photograph. Possible epidemiologic links occurred when patients lived or visited an area within 0.8 km (0.5 miles) of another or had a similar social environment. If >1 link was established between 2 patients, the strongest link was counted.
Routine demographic and clinical data were obtained from the NYC TB registry and patient interviews. Additional data on homelessness and correctional history were obtained from NYC and New York State databases.
Information regarding contacts and places of association (e.g., residences, worksites, and schools) of patients was analyzed to establish links between patients and to derive transmission hypotheses. To substantiate these hypotheses, we reinterviewed patients and their contacts by using a structured questionnaire. The questionnaire was updated with information obtained during patient interviews to ensure that hypotheses were reassessed throughout the investigation.
In October 2007, the NYC Department of Health and Mental Hygiene (DOHMH) Offi ce of General Counsel approved use of names and photographs of patients and their contacts during interviews by BTBC. Cluster investigators administered informed consent forms. Consent forms indicated that names or photographs would be obtained and shown to persons being interviewed as part of the cluster investigation. If the patient denied voluntary permission but had an incarceration history, a public record booking photograph was used. To avoid disclosing confi dential medical information, fi ctitious names and unrelated photographs were included in the compilation of names and photographs. Investigators did not confi rm or deny a TB diagnosis of any person or how persons were related. During interviews, investigators asked if patients or contacts recognized any names or photographs. If recognition was indicated, the interviewer probed to understand how persons were linked.

Statistical Analysis
We compared categorical data by using Pearson χ 2 or Fisher exact tests, as appropriate. For continuous data, the Mann-Whitney test was used to compare medians. Statistical analyses were conducted by using SAS version 9.1 (SAS Institute, Inc., Cary, NC, USA).
Places of association were geocoded through the NYC Department of City Planning's Geosupport Desktop Edition Software 9.6.9. Geocoded locations were imported into ArcGIS 9.2 (ESRI, Redlands, CA, USA) and mapped. Locations not geocoded by street address were geocoded by street intersection or other features. The ArcGIS point distance geoprocessing tool was used to calculate Euclidean distances between places of association of patients. Data were obtained as part of an outbreak investigation. Therefore, NYC DOHMH and CDC deemed this activity nonhuman subjects research.

Results
During 2003-2009, we identifi ed 54 cases of TB as part of this cluster ( Figure 1). Patient residence at TB diagnosis by NYC neighborhood is shown in Figure 2. Among 35 (65%) patients who lived in Upper Manhattan at diagnosis, median distance between the residence of any 2 patients was 1.4 km (range 0.01 km-6.6 km). Median distance between any 2 patients residing in the South Bronx (n = 10) at diagnosis was 2.9 km (range 0 km-5.8 km). Initially, the strain was susceptible to fi rst-line anti-TB drugs. However, in 2006, isoniazid resistance emerged in a patient isolate at TB diagnosis. By 2009, 17 (31%) patients had isoniazid-resistant M. tuberculosis at diagnosis. All isoniazid-resistant isolates had the Ser315Thr mutation in the katG gene and no mutations in the inhA gene region sequenced.

TB Genotyping
Forty-seven (87%) of the 54 patients had isolates with a matching spoligotype, IS6110 RFLP pattern, and 12-loci MIRU-VNTR result ( Figure 3). Forty-eight (89%) isolates met the original cluster case defi nition. Six (11%) were identifi ed as cluster-associated patients on the basis of the expanded cluster case defi nition. As of December 31, 2008, within the CDC National Tuberculosis Genotyping Service database of 32,581 patient isolates, 6 with this cluster's spoligotype and 12-loci MIRU-VNTR result were reported outside NYC (New York [n = 3], Delaware [n = 1], Georgia [n = 1], and Pennsylvania [n = 1]) (22). Among the 3 patients who resided in New York State, 1 was diagnosed in NYC and is therefore counted in the cluster ( Figure 1); no link to NYC was identifi ed for the other 2 patients.

Patient Characteristics
Patient median age was 41 years (range 10-77 years); 74% were non-Hispanic black and 69% were male (Table). Among 37 patients with drug-susceptible M. tuberculosis, 73% were male and 38% were foreign born. The 17 patients with isoniazid-resistant M. tuberculosis were predominately US born (82%) and had a history of illegal drug use (59%) or incarceration (47%).
The shift of patient characteristics with time is shown in   Among the 22 persons who disclosed a history of illegal drug use, 19 (86%) indicated noninjection drug use. The mother of a child with isoniazid-resistant M. tuberculosis also reported using illegal drugs. Drugs mentioned by patients connected to illegal drug use included smoking or snorting cocaine (n = 15), smoking marijuana (n = 6), and using heroin (n = 3).

Contact Investigation
Among 48 patients eligible for contact investigation, 1,226 contacts were identifi ed (median 9, range 0-153 contacts/patient). Twelve investigations of TB exposures in congregate settings were conducted. None of the clustered patients named one another as contacts. Contact investigation identifi ed 1 clinically diagnosed TB case linked to a cluster-associated patient with drug-susceptible M. tuberculosis.

Cluster Investigation
All 2-patient combinations (n = 1,431) were analyzed for epidemiologic links. Routine cluster investigation identifi ed 3 defi nite epidemiologic links; only 1 of these links involved a patient with isoniazid-resistant M. tuberculosis. One defi nite epidemiologic link was based on a common contact between a patient with drugsusceptible M. tuberculosis and a patient with isoniazidresistant M. tuberculosis. The other 2 links were based on patients living in the same apartment building during the infectious period of 1 of the patients. Cluster investigation methods identifi ed 3 probable epidemiologic links; all involved patients attending the same mosque during an overlapping date range. All 54 patients had a possible epidemiologic link to at least 1 other cluster patient; 98% of patients had multiple possible epidemiologic links. Of the possible epidemiologic links identifi ed, 81% were geographic and 29% involved illegal drug use. Other possible epidemiologic links were identifi ed on the basis of shared patient characteristics such as having a country of origin in West Africa, being infected with HIV, and history of mosque attendance, taxi driver occupation, incarceration, or homelessness.
During 2007-2009, patients were asked for permission to use names and photographs. Ten (59%) of the 17 patients with isoniazid-resistant M. tuberculosis, 1 (17%) of 6 patients with drug-susceptible M. tuberculosis, and 7 (64%) of 11 contacts granted permission. Public-record booking photographs were used for 2 patients. Four additional probable epidemiologic links were established through name and photograph use; all were associated with illegal drug use. Patients did not indicate familiarity with fi ctitious names and unrelated photographs that were presented.

Discussion
Despite using substantial resources within BTBC and beyond, we did not clearly identify chains of transmission in this outbreak. Only 3 defi nite epidemiologic links were identifi ed between patients, and only 1 was associated with the rapidly emerging or spreading isoniazid-resistant strain. The strongest link of this cluster is geographic; patients primarily spent time in the same neighborhoods. Although matching genotype does not always signify recent transmission, geospatial concentration and epidemiologic data indicate ongoing and recent transmission of this rare genotype in NYC. Contact investigation results showed evidence of possible transmission. However, no confi rmed This outbreak was only identifi ed through genotyping. PCR-based methods (spoligotyping and 12-loci MIRU-VNTR analysis) better defi ned this TB cluster. Supplementing contact investigation with laboratory tools to examine strain relatedness (e.g., real-time genotyping and DST) can help TB control program staff identify and investigate outbreaks. Although all patient specimens had a matching genotype, DST results showed 2 phenotypes, and therefore >2 distinct transmission chains within the cluster. Identifying separate transmission chains enabled cluster investigators to develop and test hypotheses specifi c to each chain of transmission. Common characteristics within each transmission chain implied discrete social networks, but these networks could not be confi rmed by using routine cluster investigation methods.
Emergence of isoniazid resistance in this cluster cannot be clearly explained. None of the patients with drug-susceptible M. tuberculosis showed failure of treatment. Presumably, 1 person, identifi ed by investigators as a shared contact between a patient with drug-susceptible M. tuberculosis and a patient with isoniazid-resistant M. tuberculosis, had a history of taking medications for TB and showed development of isoniazid-resistant M. tuberculosis that had not been reported to BTBC. This person died; therefore, cluster investigators were unable to confi rm this hypothesis despite medical record review and pharmacy surveillance. This investigation was limited by patients' unwillingness to report their contacts, possibly because of fear of disclosing immigration status (not asked by BTBC staff), illegal drug use, or involvement in other illicit activities. Other possible explanations include forgetting or not knowing their contacts by name (2,23). Certain patients used aliases (not tracked in the NYC TB registry) and claimed to only know their contacts by fi rst names or aliases. Pervasiveness of aliases within patient social networks stymied contact investigation efforts and made establishing epidemiologic links between patients diffi cult.
High prevalence of illegal drug use within the cluster led investigators to explore how specifi c drug-use practices contribute to TB transmission. Studies reported that such specifi c drug-use practices as shotgunning (inhaling smoke from rock cocaine or marijuana and blowing the smoke directly into the mouth of another) and hotboxing (smoking drugs in a small, enclosed space to maximize narcotic effect through fi rst-hand and second-hand smoke) were associated with TB transmission (24,25). Although these practices were not specifi cally mentioned by patients or their contacts, specifi c questions were not asked until later in the investigation. After consulting with substanceuse experts, BTBC revised their cluster-investigation questionnaire and provided investigators with additional training on patient-interview procedures and drug-use  subculture. Understanding drug-use behavior helps TB control personnel elicit sensitive transmission information. BTBC also modifi ed how substance-use information is collected and recorded in the TB registry. Transmission through casual contact and increased virulence are possible explanations for extensive transmission of this strain and lack of recognition among patients. Although TB transmission from casual contact is considered rare, it has been documented (26)(27)(28)(29)(30). If this strain, like other outbreak strains (29), was highly virulent, extensive transmission among patients who did not recognize each other would have been possible. Moreover, geographic proximity of patients to one another might have increased opportunities for TB exposure and supported transmission through casual contact. In addition, positive results for acid-fast bacilli in smears of respiratory specimens among cluster-associated patients were substantial (70% overall, 93% among cocaine users) and considerably greater than recent past NYC TB patients (range 42%-46% during 2003-2008) (NYC DOHMH, unpub. data), thus increasing likelihood of transmission. Investigation fi ndings were consistent with those of a London study that reported that pulmonary TB patients who used cocaine were more likely to be sputum smear positive at diagnosis (31), perhaps related to delays in seeking medical care.
Photograph and name use yielded the strongest epidemiologic links between patients with isoniazidresistant M. tuberculosis. It was the only method that confi rmed patient recognition within the cluster. All epidemiologic links established through photograph recognition were related to illegal drug activity. Other outbreak investigations have highlighted unwillingness of patients to share social contacts when these contacts are connected to illegal activities (4,5,13).
Insights gained from using name and photograph data in an ongoing investigation will benefi t TB control programs. This method would have been more successful if used earlier in the investigation. TB control personnel contemplating adopting this strategy should obtain legal guidance before an outbreak occurs because privacy laws vary from one locality to another.
This outbreak investigation highlights an array of challenges for US-based TB control programs. Understanding and preventing TB transmission among hard-to-reach populations requires considerable resources. Conventional contact investigation can be inadequate for identifying and curtailing TB transmission among diffi cultto-reach-populations. New methods, including using name and photograph data, are needed for TB elimination. Ms Perri is the genotyping coordinator at the Bureau of Tuberculosis Control at the New York City Department of Health and Mental Hygiene. Her research interests include the molecular epidemiology of tuberculosis and infectious disease control in resource-limited settings.