Universal Genotyping in Tuberculosis Control Program, New York City, 2001–2003

Real-time universal genotyping decreased unnecessary treatment.

S ince the early 1990s, selective tuberculosis (TB) genotyping has been used in New York City for outbreak investigations, to identify isolates resistant to at least isoniazid and rifampin (multidrug-resistant TB), and in special studies. TB genotyping was essential to investigate and confirm transmission in a number of settings and to confirm or exclude laboratory contamination (1)(2)(3)(4)(5)(6)(7)(8). A number of programs demonstrated the utility of universal genotyping, which influenced the development of this service in New York City (9)(10)(11)(12)(13)(14)(15)(16). In 2001, the New York City Bureau of Tuberculosis Control began genotyping isolates for every new TB case with spoligotyping and IS6110-based restriction fragment length polymorphism (RFLP) to improve the efficiency of TB control. Two laboratories with extensive genotyping experience were selected through a competitive bidding process. Both were participating laboratories in the National Tuberculosis Genotyping and Surveillance Network and had performed genotyping for selected cases in New York City since the early 1990s (6,17).
The objectives of universal TB genotyping were to more rapidly and efficiently 1) determine the extent and dynamics of ongoing transmission to focus program interventions for specific areas and populations; 2) assess TB transmission in outbreaks to refine contact investigations; 3) identify nosocomial transmission not identified by conventional methods; and 4) identify false-positive cultures so that clinicians could be notified of diagnostic errors quickly and prevent unnecessary TB treatment. We describe the elements and activities required to develop and implement real-time universal genotyping in a large urban TB control program.

Identifying and Obtaining Isolates for Genotyping
Implementation of universal genotyping in New York City consisted, briefly, of 1) requiring submission of the initial positive isolate, reinforced by health code amendment (18,19); 2) advising all relevant laboratories and

Creation of TB Genotyping Databases
Implementing universal genotyping also required developing a comprehensive database to monitor and manage information on specimen collection, shipment, and genotyping, as well as epidemiologic information gathered on each clustered patient. A relational database was created by New York City TB control staff in Microsoft Access (Microsoft Corporation, Redmond, WA, USA) that included 1) genotyping results for isolates identified after January 1, 2001; 2) specimen-tracking information such as date of receipt at the public health laboratory, shipment and reporting dates from each genotyping laboratory, and false-positive culture investigation results; 3) clustered patient information, such as location where each patient spent time during the potential infectious period, locations where TB could have been acquired in the 5 years before diagnosis, cluster characteristics, links between patients, and potential transmission sites; and 4) results of genotyping performed from 1990 to 2000 as part of the selective genotyping activities (3,5,6,8,17). Queries of the database were developed to identify cases with identical RFLP and spoligotype results for "real-time" cluster investigation and investigations of false-positive cultures. Quality assurance exercises to test reliability of results were developed and kept in the database. Queries are performed monthly to identify cases for which an isolate was not submitted to the public health laboratory. In such cases, Bureau of Tuberculosis Control staff sends reminder letters and makes phone calls to ensure that these isolates are received.

Investigation of False-positive Culture Results
A false-positive TB culture is defined as a positive TB culture that is not the result of culture-positive disease in a patient but instead may be due to 1) laboratory cross-contamination during specimen processing; 2) errors in collection or labeling, either on the patient ward or in the laboratory; or 3) contamination of clinical devices, for example, contamination of a bronchoscope during specimen collection. The primary goal of investigations of false-positive cultures is to discontinue unnecessary treatment in patients found to have false-positive TB cultures. Before universal genotyping, suspected false-positive cultures were investigated in 1 of 3 ways: 1) monthly review of patients with a single positive culture; 2) request from Bureau of Tuberculosis Control staff, including case managers, department of health physicians, and epidemiologists; and 3) requests by outside providers and laboratories to investigate cultures not consistent with the patient's clinical picture. With universal genotyping, an investigation can also be initiated when cases have matching genotypes and have been processed within 2 working days of each other at the same facility. For these investigations, genotype information, specimen processing, and other information (e.g., patients hospitalized on the same floor) are reviewed, and suspected false-positive cultures are determined to be confirmed, unlikely, or inconclusive. Treating physicians and clinical staff in the program are notified of the outcome of investigations of false-positive cultures so patient evaluation can be evaluated further and a decision can be made on whether continued treatment is indicated.

Genotype Cluster Investigations
A cluster investigation aims to uncover epidemiologic links between members of a genotype cluster through systematic review of patient records and re-interviews, if needed. We consider real-time investigation of clusters to occur when the cluster investigation components (i.e., record review and re-interview) take place close to the time the most recent case in the cluster is identified. We defined a genotype cluster as >2 cases identified from 2001 to 2003 that had isolates with identical IS6110-based RFLP banding pattern and spoligotype, regardless of the number of IS6110 copies. Patients with a definite epidemiologic link include those who have named each other as contacts, have a contact in common without naming each other as contacts, or have reported a common date range at the same location (e.g., residence, hospital, prison, workplace, single-room-occupancy hotel [any supervised publicly or privately operated facility designed to provide temporary living accommodations], or shelter). The common date range includes the potentially infectious period (i.e., 3 months before start of treatment) for at least 1 patient. Patients with a probable link have spent time at the same location (as above) during the same time frame, exclusive of the infectious period of the patients, without naming each other as contacts. Possible links exist among patients who have a similar social network or have spent time in the same area (no specific location), without naming each other as contacts.
When definite epidemiologic links are found among cluster members through the review of the TB case registry and patient records, information is recorded in the database on the nature of this relationship. If transmission at specific locations is shown, additional contacts are tested at these locations. If no such links exist, an epidemiologist reviews the cases and conducts in-depth patient re-interview to attempt to identify links and previously unidentified locations of transmission. A standard questionnaire is used to re-interview clustered patients. In addition, other registries such as the Department of Homeless Services are searched by cross-matching with the database each quarter to identify other possible exposure locations.

Performance Indicators
Performance indicators are used to evaluate procedures with respect to timely shipping of isolates and reporting of genotyping results and to assess the reliability of genotyping results. Submission time is calculated for clinical laboratories that process samples from New York City TB patients as the time between the date a positive culture is collected and the date the isolate is received at the public health laboratory. Submission time for genotyping is the time between the date the isolate is received at the public health laboratory and the date the isolate is sent for genotyping. Reporting time for the genotyping laboratories is the time between the date the specimen is received at the genotyping laboratory and the date the spoligotype or RFLP is reported to the Bureau of Tuberculosis Control.
The time to completion of investigations of false-positive cultures is defined as the time from specimen collection to investigation completion. The goal is to complete investigations within 90 days of collecting the first positive culture. Time to completion of a cluster investigation is calculated from the date a cluster is identified and an investigation is initiated until a decision is made regarding links between cases in the cluster; the goal is to complete these investigations within 21 days. Because clusters are dynamic, a new investigation is started when an additional case with that particular strain is identified.
Quality assurance exercises to assess the reliability of genotyping results are performed every 6 months. Ten percent of isolates genotyped in the previous 6 months are randomly selected by Bureau of Tuberculosis Control and sent for blinded retyping. Each laboratory repeats genotyping and sends the results to the bureau for comparison with previously reported results. Discrepant results are reviewed and discussed in the working group, and another isolate is requested from the initial processing laboratory to verify results.

Outcomes
The genotyping services process is summarized in the online Appendix Figure (available at http://www.cdc.gov/ ncidid/EID/vol12no05/05-0446_appG.htm). The number of eligible isolates by year is shown in the  In 2003, patients with a false-positive culture were treated unnecessarily for a median of 7 days (range 0-145). This median number of days is considerably lower than that seen before universal genotyping; in 1999, patients identified by retrospective surveillance (i.e., the single-positive culture list) as having false-positive cultures completed a median of 7 months of treatment.
Among 2,378 isolates with a complete genotype (truepositive cultures), 565 spoligotype patterns and 1,600 RFLP patterns were identified; 2,009 (84.5%) of 2,378 patient isolates clustered in 196 spoligotype clusters, and 1,002 (42.1%) of 2,378 patient isolates in 224 RFLP clusters. Eight hundred thirty-seven (35.2%) of the 2,378 isolates had RFLP and spoligotype patterns that matched >1 other isolate pattern; these isolate patterns were grouped into 205 genotype clusters ranging in size from 2 to 81 cases (mean 4 cases/cluster; median 2 cases/cluster). The percentage of clustered patient isolates remained stable during the 3-year period (χ 2 for trend p = 0.3652). While most patient isolates had 9-13 copies of IS6110 (median 11 copies, range 1-23), strains with a lower copy number (<6 bands) were more likely to be clustered. From 2001 to 2003, two large outbreaks occurred that involved strains of 1 and 3 IS6110 copies. After these strains were excluded, the percentage clustered remained higher for patterns with lower numbers of IS6110.
A total of 278 (33.2%) of 837 clustered cases had epidemiologic links identified; of these, 105 (37.8%) had links established through traditional contact investigations. Genotype cluster investigations established links for the remaining 62%: 15% of the links were definite, 11% probable, and 36% possible. For 66.4% of clustered cases (556 cases), no epidemiologic links were identified. Time to completion of cluster investigations decreased from a median of 176 days in 2001 to 37 days in 2003. The delay in completing investigations at the beginning of the project was mostly due to staff vacancies. Cluster investigations uncovered 57 additional links among cases with matching genotypes and 17 additional sites of transmission. Links established through genotype cluster investigations led to 4 expanded contact investigations in congregate settings (2 in homeless shelters, 1 in a single-room-occupancy hotel, and 1 in a local grocery store). These investigations identified additional infected contacts and 4 additional TB patients at a homeless shelter. These sites are now monitored closely for additional patient isolates with these genotypes. Transmission between TB patients was ruled out in >5 site investigations because the genotypes were unrelated, avoiding more extensive case-finding efforts that are needed once transmission is seen.
Four quality assurance exercises were performed from 2001 to 2003 on 216 isolates. The result was 94.4% concordance for spoligotyping and 93.5% for RFLP. Of retyped spoligotype patterns that did not exactly match the original patterns, 50% differed by ±1 spacer, 8% differed in multiple successive spacers, and 42% differed for other reasons. Among retyped RFLP patterns, 57% differed from the original patterns because of the existence or absence of >1 bands, 36% differed because of pattern shifts, and 7% differed for other reasons.

Discussion
We achieved real-time universal genotyping as part of routine TB control with capture and completion comparable to that seen by the National Tuberculosis Genotyping and Surveillance Network sites (30). High participation rates among clinical laboratories were essential to the completeness of genotyping. Timely submission of isolates from clinical laboratories and continuing decrease in submission time from the public health laboratory to the genotyping laboratories also facilitated efforts to achieve real-time investigation of false-positive cultures and clusters.
Implementation of TB genotyping in a large TB control program is complex. It requires TB control, epidemiology, and laboratory resources, and the costs are substantial. New York City contracts with genotyping laboratories carry an annual cost of nearly US $150,000 ($20,000 for spoligotyping and $125,000 for RFLP). In addition, 2 to 3 epidemiologists are allocated for database management and cluster investigation in New York City. Nonetheless, we have seen added value from universal genotyping. Additional sites of transmission were found on the basis of results of cluster investigations. Expanded investigations conducted at these sites identified additional patients and infected contacts who were subsequently treated for TB and latent TB infection. Genotyping information has also been useful by showing that TB cases clustered in place and time can have unrelated genotypes. For example, unrelated genotypes of >2 TB cases diagnosed in a setting with a high prevalence of TB infection may provide evidence that the cases did not occur as a result of transmission within that setting. Thus, more limited contact investigations of persons exposed to each of the patients can be performed instead of the more aggressive expanded contact investigation or case-finding activities that would be required if the isolates had matching genotypes. In addition, the efficiency of investigations of false-positive cultures increased as a result of universal genotyping, since a greater proportion of investigations initiated through genotyping matches yielded true false-positive culture results than investigations initiated through other methods. The amount of unnecessary treatment for these patients also decreased.
The higher rates of clustering seen in low copy-number isolates by RFLP alone support our decision to use 2 genotyping assays; this phenomenon has been reported previously (11). In addition, the rapid availability of spoligotype results allowed earlier initiation of investigations of both clusters and false-positive cultures than would have been possible with RFLP results alone. Particularly useful was close communication with the Wadsworth Center on interpretation of spoligotype matches for "rare" spoligotypes (seen less often than average for most spoligotypes in our database) and on prioritization of these isolates for investigation as clusters or false-positive cultures.
Since January 2004, mycobacterial interspersed repetitive unit and spoligotyping analyses are performed on all isolates as part of the Centers for Disease Control and Prevention's National Tuberculosis Genotyping Program. The availability of this additional assay will allow us to examine the extent to which MIRU further differentiates genotype clusters on the basis of RFLP and spoligotyping. MIRU may also reduce the time to obtain the genotype result and initiate a cluster investigation since it, like spoligotyping, requires few organisms and does not require live culture. Implementing the national genotyping service will also greatly reduce the financial costs for TB control jurisdictions interested in using genotyping to enhance their current program activities (31).