|Home | About | Journals | Submit | Contact Us | Français|
Background.Extensively drug-resistant tuberculosis (XDR-tuberculosis) is a global public health threat, but few data exist elucidating factors driving this epidemic. The initial XDR-tuberculosis report from South Africa suggested transmission is an important factor, but detailed epidemiologic and molecular analyses were not available for further characterization.
Methods.We performed a retrospective, observational study among XDR-tuberculosis patients to identify hospital-associated epidemiologic links. We used spoligotyping, IS6110-based restriction fragment–length polymorphism analysis, and sequencing of resistance-determining regions to identify clusters. Social network analysis was used to construct transmission networks among genotypically clustered patients.
Results.Among 148 XDR-tuberculosis patients, 98% were infected with human immunodeficiency virus (HIV), and 59% had smear-positive tuberculosis. Nearly all (93%) were hospitalized while infectious with XDR-tuberculosis (median duration, 15 days; interquartile range: 10–25 days). Genotyping identified a predominant cluster comprising 96% of isolates. Epidemiologic links were identified for 82% of patients; social network analysis demonstrated multiple generations of transmission across a highly interconnected network.
Conclusions.The XDR-tuberculosis epidemic in Tugela Ferry, South Africa, has been highly clonal. However, the epidemic is not the result of a point-source outbreak; rather, a high degree of interconnectedness allowed multiple generations of nosocomial transmission. Similar to the outbreaks of multidrug-resistant tuberculosis in the 1990s, poor infection control, delayed diagnosis, and a high HIV prevalence facilitated transmission. Important lessons from those outbreaks must be applied to stem further expansion of this epidemic.
(See the editorial commentary by Dharmadhikari, on pages 1–3.)
Drug-resistant tuberculosis was recognized as a significant public health threat in industrialized countries in the early 1990s, following several outbreaks of multidrug-resistant (MDR) tuberculosis [1–5]. These outbreaks were characterized by transmission of drug-resistant tuberculosis in congregate, institutional settings with poor infection control. Secondary tuberculosis cases occurred largely among human immunodeficiency virus (HIV)–infected patients, with a short incubation period between tuberculosis exposure and disease onset and a high mortality rate involving a rapid progression to death 4–16 weeks after symptom onset . In response to these outbreaks, clinical, public health, and health systems interventions were implemented to contain and later reverse the rising tide of drug-resistant tuberculosis in these countries . A key component of these interventions was to create infection control facilities and procedures to prevent tuberculosis transmission [7–9].
Unfortunately, the lessons learned from these MDR-tuberculosis outbreaks were not adopted by tuberculosis control programs worldwide. Consequently, the global prevalence of drug-resistant tuberculosis continued to rise in the decades following these outbreaks. The most recent global report of drug-resistant tuberculosis indicates the highest rates of MDR-tuberculosis and extensively drug-resistant (XDR) tuberculosis ever reported . Outbreaks of drug-resistant tuberculosis have continued to occur worldwide, including a highly publicized outbreak of XDR-tuberculosis from our study site in Tugela Ferry, South Africa [11–15]. Similar to the outbreaks in the 1990s, we identified XDR-tuberculosis cases among HIV-infected patients. These cases likely occurred as a result of transmission because 55% had no history of tuberculosis treatment.
From 2005 through 2009, 516 cases of XDR-tuberculosis were diagnosed in Tugela Ferry , equating to an average annual incidence of 52 XDR-tuberculosis cases per 100 000 population. Because conditions in the Tugela Ferry hospital are similar to those implicated in the 1990s outbreaks—congregate wards with poor ventilation, high census of HIV-infected patients, and absence of infection control policies or isolation facilities—we sought to further investigate the role of nosocomial transmission in the ongoing XDR-tuberculosis epidemic. In this study, we combined epidemiologic investigation, molecular genotyping, and social network analysis to characterize XDR-tuberculosis transmission patterns in Tugela Ferry.
We performed a retrospective, observational study among patients who received a diagnosis of XDR-tuberculosis in Tugela Ferry from January 2005 through December 2006. XDR-tuberculosis patients were identified by reviewing the district hospital's tuberculosis register. Consecutive XDR-tuberculosis patients with available medical records were included in this study.
Tugela Ferry is a rural area in KwaZulu-Natal province, with a population of 160 000 spread across an area of 2500 km2. In 2006, the incidence of drug-susceptible tuberculosis was 1100 cases per 100 000 population, and the MDR-tuberculosis incidence was 119 cases per 100 000 . More than 80% of tuberculosis cases were co-infected with HIV, and the prevalence of antenatal HIV infection was 37% (prevalence among all KwaZulu-Natal residents in 2005, 16.5%) . A single, 355-bed government district hospital provides inpatient care for this area. Congregate inpatient general medicine and tuberculosis wards are separated for men and women; patients with suspected but not yet confirmed tuberculosis are admitted to the general medical wards. The wards hold 30–40 patients each, with beds side by side, approximately 1 m apart. At the time of this study, there were no mechanical air extraction fans, isolation rooms, or airborne infection control policies.
Since June 2005, clinicians in Tugela Ferry have been encouraged to evaluate all patients with suspected tuberculosis with culture and drug-susceptibility testing (DST), in addition to smear microscopy, given the concern for a high prevalence of drug-resistant tuberculosis. This practice differed from South African national policy, which recommended culture and DST only for cases of treatment failure or re-treatment . Detailed methods regarding sputum collection, microscopy, culture, and DST has been described previously . DST results were typically available 6–12 weeks following sputum collection; at the time of this study, rapid diagnostic tests for drug resistance were not available.
All new tuberculosis patients initiated empirical first-line therapy (2 months of isoniazid, rifampin, ethambutol, and pyrazinamide treatment, followed by 4 months of isoniazid and rifampin treatment), while patients requiring re-treatment initiated a standard category II regimen . Second-line therapy for drug-resistant tuberculosis was not available at the Tugela Ferry hospital. Patients with confirmed MDR-tuberculosis or XDR-tuberculosis were transferred to the drug-resistant tuberculosis referral hospital in Durban for treatment. The average time from sputum collection to transfer was 16 weeks for XDR-tuberculosis patients , reflecting delays in both laboratory processing and health systems, during which patients continued to receive first-line tuberculosis therapy. Patients could remain on the inpatient wards or could be discharged home, if clinically stable, while awaiting their diagnosis and transfer to the tuberculosis referral hospital.
Upon transfer to the tuberculosis referral hospital, XDR-tuberculosis patients received a standardized treatment regimen of kanamycin, ofloxacin, ethionamide, ethambutol, pyrazinamide, and terizidone for at least 4 months, followed by the same regimen, without kanamycin, for an additional 18 months. When capreomycin and para-aminosalicylic acid became available in South Africa in 2007, they replaced kanamycin and ofloxacin. Third-line tuberculosis drugs and surgical treatment were not routinely used.
We reviewed patient medical records to collect data on demographic characteristics, HIV history (HIV status, baseline CD4 T-cell count and viral load, and antiretroviral therapy [ART] use), tuberculosis history, laboratory results, and hospitalization history (any admissions from 1 January 2000 through 31 December 2007). For any patients whose medical records could not be located, we collected information from the drug-resistant tuberculosis register about age, sex, diagnosis date, DST pattern, and survival. We compared patients with available medical records to those whose records could not be located for any differences on these characteristics.
To determine hospital exposure of patients who developed XDR-tuberculosis, we identified epidemiologic links, using dates of admission and discharge from the district hospital in Tugela Ferry. The XDR-tuberculosis diagnosis date was defined as the date of collection of the specimen that grew Mycobacterium tuberculosis that was XDR. Patients were conservatively considered infectious beginning 2 weeks before their XDR-tuberculosis diagnosis date (Supplementary Figure 1) and were assumed to remain infectious until death or the end of the study period, since no patients were documented to have culture conversion. A patient's vulnerable period was defined as any time >6 weeks before their XDR-tuberculosis diagnosis date (Supplementary Figure 1), to allow for an “incubation period” between exposure and XDR-tuberculosis onset [21, 22].
Patients were considered epidemiologically linked if all of the following criteria were met: (1) they were hospitalized concurrently for at least 1 day, (2) they had the same sex, (3) one patient was infectious, and (4) the other patient was in the vulnerable period.
Patient-days of exposure were calculated as the cumulative number of a days that a vulnerable patient was exposed to an infectious XDR-tuberculosis patient in the hospital; thus, if a vulnerable patient overlapped with 1 infectious XDR-tuberculosis patient for 5 days, the vulnerable patient would have 5 patient-days of exposure. We also stratified exposure by acid-fact bacilli (AFB) smear status (ie, smear-positive vs smear-negative), to account for potential differences in transmission risk.
To determine genotype clusters, mycobacterial culture isolates recovered from XDR-tuberculosis patients underwent spoligotyping  and were classified using Spoligotype International Type (ST) numbers . Isolates with matching spoligotypes underwent IS6110-based restriction fragment–length polymorphism (RFLP) analysis . Genomic DNA from a subset of isolates with matching spoligotype and RFLP patterns was further analyzed for genetic polymorphisms by direct sequencing of 10 drug resistance–determining regions: rpoB, katG, mabA-inhA (including upstream regions), pncA, embB, rpsL, rrs, tlyA, gidB, and gyrA [26–28].
We identified genotype clusters among patients with available XDR-tuberculosis isolates. A cluster was defined as ≥2 patients with isolates of identical spoligotype and RFLP pattern. DNA sequencing of drug resistance–determining regions was used to further differentiate any large clusters. We then evaluated epidemiologic links among patients within genotype clusters, as described above.
We constructed transmission networks using the following rules: patients were included in a network if they had an epidemiologic link to a member of the network, and patients with earlier XDR-tuberculosis diagnosis dates were considered to have infected epidemiologically linked patients who had later diagnosis dates. Transmission networks were drawn with earlier cases on the left and later cases on the right. For each network, we calculated the component size, network density, and centrality. Transmission networks were initially constructed for all patients in this study. Then, to focus on documentable transmission, we constructed individual transmission networks among male and female patients in the major genotypic cluster (KZN).
We analyzed patients' demographic and clinical characteristics and hospitalization history as percentages, means and SDs, and medians and interquartile ranges (IQRs). Statistical and social network analyses were conducted using SAS, UCINET, and PAJEK software.
The study protocol was approved by the ethics committees of the Albert Einstein College of Medicine, Yale University, and University of KwaZulu-Natal and by the KwaZulu-Natal Department of Health.
In 2005–2006, XDR-tuberculosis was diagnosed in 243 patients, ranging from 0 to 21 cases per month (Figure 1). Medical records were available for 148 XDR-tuberculosis patients (61%), who did not differ significantly in demographic characteristics, diagnosis date, DST patterns, or survival from those whose charts were not available (data not shown).
Of these 148 patients, 83 (56%) were female, and the median age was 34 years (IQR, 29–41 years; Table 1). Among 126 patients (85%) tested for HIV, 98% were HIV infected, of whom 31% were receiving ART. The median CD4 T-cell count was 64 cells/mm3 (IQR, 22–169 cells/mm3) at XDR-tuberculosis diagnosis. Eighty-seven XDR-tuberculosis patients (59%) had a positive AFB sputum smear. Although the majority of patients (125 [84%]) had been previously treated for tuberculosis with first-line tuberculosis medications, only 2 (1%) had ever received second-line drugs for treatment of MDR-tuberculosis.
A total of 138 patients (93%) were admitted to the district hospital while infectious with XDR-tuberculosis (Figure 2). The median duration of hospitalization was 15 days (IQR, 10–25 days). Our examination of the daily hospital census revealed that there was at least 1 infectious XDR-tuberculosis patient (range, 1–10 patients) in the hospital during 661 (91%) of 730 days in 2005–2006 (Figure 2).
Before diagnosis with XDR-tuberculosis, 113 patients (76%) were hospitalized at least once and spent a median of 22 days (IQR, 13–34.5 days) admitted (Table 1). Of these, 80 (71%) were exposed to at least 1 infectious XDR-tuberculosis patient in the hospital. The median exposure was to 5 infectious XDR-tuberculosis patients (IQR, 3–8 patients), resulting in 36.5 patient-days of exposure (IQR, 23–77.5 patient-days).
When analysis of hospital exposure was restricted to AFB smear–positive XDR-tuberculosis patients, the results were similar: 79 hospitalized patients (70%) were exposed to at least 1 AFB smear–positive XDR-tuberculosis patient. The median exposure was to 3 AFB smear–positive XDR-tuberculosis patients (IQR, 2–6 patients), for a median of 27 patient-days (IQR, 16–53 patient-days) of exposure.
Of 148 XDR-tuberculosis patients, spoligotyping could be performed for 86 (77%); spoligotyping was not performed for the remaining patients because a specimen was not stored (37 patients) or was nonviable (25 patients). Spoligotyping indicated a predominant cluster (ST60) comprising 92% of isolates (79 patients; Figure 3A); isolates from the remaining 7 patients were distributed over 1 small cluster (ST 53; 2 patients) and 5 unique patterns (STs 33, 42, 90, 136, and 1166; 1 patient each). To further differentiate the large ST60 cluster, IS6110-based RFLP was performed on 53 ST60 isolates that were viable. Three RFLP patterns were observed (KZN, F28, and A), with the KZN strain accounting for 96% (51 patients) of the ST60 cluster isolates (Figure 3B).
To further assess clonality, 27 of the ST60/KZN isolates underwent DNA sequencing of 10 resistance-determining regions. All 27 isolates demonstrated identical mutations across all genes tested (katG: S315T; mabA-inhA: –8T>A; rpoB: D516G, L533P, I1106T; pncA: ins C after 456; embB: M306V; rpsL: wildtype; rrs: A1401G; tlyA: wild-type; gidB: L16R, Δ152–281; and gyrA: A90V).
We characterized the transmission networks of the 51 XDR-tuberculosis patients (22 men and 29 women) in the large genotype cluster (ST60/KZN). An epidemiologic link could be established for 19 men (86%) and 23 women (79%; Figure 4). The median number of epidemiologic links between genotypically identical XDR-tuberculosis patients was 3 (IQR, 2–5 links). Patients demonstrated a high degree of interconnectedness: each of the transmission networks was a complete connected component (ie, all of the XDR-tuberculosis patients in the network were connected to one another through epidemiologic links). Periods of exposure and time to diagnosis demonstrated sequential positivity for as many as 3 generations of possible transmission among men and 5 generations of transmission among women. Patients in the highest quartile of epidemiologic links (1 man with 11 links and 6 women with 6, 8, or 10 links) were central to these generations of transmission; however, these numbers were too small to draw inferences about patient characteristics for central persons. Transmission networks were constructed for the entire group of XDR-tuberculosis patients, and the distribution of linkages within the KZN cluster was similar to that for the entire group, as were the distributions of days of exposure and of time to diagnosis.
In this study, we examined the role of nosocomial transmission among a large cohort of patients who developed XDR-tuberculosis in rural South Africa. We found that the majority of patients had been admitted to a single hospital and had experienced frequent, prolonged exposure to XDR-tuberculosis patients. Our use of multiple genotyping methods, including DNA sequencing of resistance-determining regions, demonstrated that this epidemic was highly clonal. However, social network analysis suggests that, rather than a point-source outbreak from a single patient, there was a high-degree of interconnectedness that allowed for multiple generations of XDR-tuberculosis transmission over time. On the basis of these data, we conclude that several factors associated with this hospital—long length-of-stay, crowded congregate wards, poor infection control practices, and high prevalence of HIV infection—created dangerous conditions for extensive nosocomial transmission and rapid dissemination of the XDR-tuberculosis epidemic. These results have important implications for tuberculosis infection control, especially in settings with high HIV infection and drug-resistant tuberculosis prevalence.
We used several genotyping methods to ensure adequate discriminatory power to identify transmission links between study patients. Our results show that the majority of patients were infected with genetically identical M. tuberculosis isolates, confirming that transmission was the principal mechanism by which patients contracted XDR-tuberculosis. The major role of transmission is further supported by the fact that LAM4/KZN, the predominant XDR-tuberculosis strain, was found in only 27% of the MDR-tuberculosis cases and in 4% of the drug-susceptible tuberculosis cases from the same hospital . If the majority of cases had arisen from stepwise acquisition of resistance, the genotypic diversity seen among MDR and drug-susceptible tuberculosis cases would have been preserved among XDR-tuberculosis cases, as has been observed in studies from other South African provinces [30, 31]. Moreover, the fact that XDR M. tuberculosis isolates showed an identical array of mutations among the 10 sequenced resistance-determining regions further supports the hypothesis that resistance to first- and second-line tuberculosis drugs had developed before this XDR-tuberculosis strain was widely transmitted.
XDR-tuberculosis transmission is most likely to have occurred in the hospital, since the majority of patients could be linked epidemiologically, through multiple, concurrent hospital admissions. These hospital admissions created large, interconnected networks that allowed several generations of nosocomial transmission. The high rate of hospitalization and extended length of stay, combined with the congregate design of wards in this hospital—containing 30–40 beds separated from each other by <3 feet—created a substantial risk for exposure to infectious XDR-tuberculosis patients [4, 32]. Moreover, the long delay inherent in diagnosing XDR-tuberculosis likely contributed to transmission by allowing infectious XDR-tuberculosis patients to be admitted unrecognized and untreated to the congregate wards. While infectiousness of tuberculosis patients can be highly variable, studies have consistently shown that patients receiving inadequate treatment regimens, including those for drug-resistant tuberculosis, account for the majority of transmission events . In this study, at least 1 infectious XDR-tuberculosis patient was unknowingly present in the hospital wards nearly every day of the study period. With these underlying conditions, multiple generations of XDR-tuberculosis transmission were likely to have occurred, rather than a single transmission event as typically seen in outbreaks.
Several recent studies place the current findings in the context of the broader XDR-tuberculosis epidemic occurring in KwaZulu-Natal province. The first cases of XDR-tuberculosis in KwaZulu-Natal are now known to have occurred as early as 2001 [20, 34]. Laboratory data show that XDR-tuberculosis prevalence has grown exponentially from 2001 through 2007; by 2007, a total 792 XDR-tuberculosis cases had been diagnosed throughout the province, from 58 different hospitals and every district [20, 35]. This rapid geographic dissemination suggests a major role for primary transmission in this epidemic. Furthermore, a study of XDR-tuberculosis isolates from various towns across KwaZulu-Natal province supports the likelihood of clonal expansion of this epidemic, since all patients had nearly identical genetic sequences on whole genome sequencing . The spoligotype, RFLP, and resistance mutation patterns of those isolates were identical to those of the predominant cluster found in our study . Taken together, these data suggest that the XDR-tuberculosis cases in our study may be part of a multiyear, multi-institutional epidemic of XDR-tuberculosis, similar to the W-Beijing outbreaks [2, 37, 38], that began years before it was detected in Tugela Ferry. Although further studies are needed to investigate this hypothesis, XDR-tuberculosis dissemination could have occurred through province-wide transmission networks that traversed multiple hospitals, given their commonality of congregate wards with poor infection control and their association with a single hospital to which all cases of drug-resistant tuberculosis were referred.
The findings from this study provide important insights for stemming the growth of this epidemic. Because the majority of XDR-tuberculosis cases are incurable [17, 31, 39, 40], a comprehensive strategy, similar to those used in the United States and Europe in the 1990s, must be used to prevent additional cases from developing . Control strategies must include not only efforts to improve cure rates for susceptible and MDR-tuberculosis cases, but also infection control programs that can prevent transmission of drug-resistant M. tuberculosis strains. The success of infection control programs hinges on their ability to promptly identify and separate XDR-tuberculosis patients from other patients [41, 42]. Thus, in addition to well-recognized administrative, environmental and personal protection measures, implementation of rapid diagnostic testing for tuberculosis drug resistance [43, 44] and redesign of healthcare facilities to minimize congregate spaces (eg, wards and outpatient waiting areas) are critical elements of infection control programs. Indeed, if a rapid drug resistance diagnostic test, isolation rooms, and smaller wards were available in our setting, the large daily census of XDR-tuberculosis patients (Figure 2) and their subsequent exposure to vulnerable patients may have been prevented. Last, earlier and greater use of ART can reduce HIV-infected patients' vulnerability to progression of tuberculosis and prevent new cases from contributing to the ongoing cycle of transmission. Although financial and human resources are limited in developing countries, implementation of these strategies may prove cost-effective when compared to the large public health and societal costs of further expansion of this epidemic [45, 46].
These study findings are subject to a few limitations. First, our findings are likely minimal estimates of nosocomial transmission, because we only studied incident XDR-tuberculosis cases from 2 years of a multiyear epidemic and were able to locate medical records of only two-thirds of the cases. With greater capture of cases within the study period and expansion to the surrounding years, it is likely we would have found more epidemiologic links and additional generations of XDR-tuberculosis transmission. Additionally, since culture and DST are not routinely obtained on all tuberculosis suspects in other districts in this province, many XDR-tuberculosis cases are never diagnosed and may be additional unidentified links in the transmission networks. Second, our genotyping data were incomplete because of nonviable or unavailable isolates, so the size of the clusters is likely an underestimate. Again, this may have resulted in unidentified links in the transmission network. Last, this study was limited to evaluating only hospital links among patients. Thus, transmission that may have occurred in the outpatient clinics, households, or the community was not measured.
XDR-tuberculosis is now recognized to be present in all of South Africa's provinces  and in all of its neighboring countries . It is clear that poor chemotherapy leads to selection of drug-resistant strains. However, our study adds to the growing evidence that transmission of drug-resistant strains is a critical factor driving the XDR-tuberculosis epidemic [47–49]. Because similar conditions exist in hospitals throughout low- and middle income countries, efforts to control the epidemic of drug-resistant tuberculosis will require not only strengthening of tuberculosis treatment programs, but also implementation of infection control programs, which include redesign of healthcare facilities and roll-out of rapid tests for diagnosing drug-resistant tuberculosis . A substantial financial and human resources investment in combating the epidemic of drug-resistant tuberculosis in the United States was successful in reversing the epidemic [7, 51]. Similar interventions are needed worldwide to reverse the concerning trend in the global epidemiology of drug-resistant tuberculosis.
Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.
Acknowledgments.We are grateful to the courageous staff of the Church of Scotland Hospital, the Umzinyathi district Department of Health, and the KwaZulu-Natal province tuberculosis control program. We thank the Inkosi Albert Luthuli Central Hospital/National Health and Laboratory Services for performing all tuberculosis culture and DST for patients from Tugela Ferry. We thank Jeremy Connors for his technical assistance in carrying out gene sequencing. This work also would not have been possible without the many medical students, residents, and research assistants who contributed to data collection and data cleaning.
Financial support.This work was supported by a Doris Duke Charitable Foundation Clinical Scientist Development Award (2007070; principal investigator [PI], N. R. G.) and a pilot grant from the Einstein/Montefiore Center for AIDS Research (NIH AI-51519; PI, N. S. S.). Dr Shah is also the recipient of a Doris Duke Clinical Scientist Development Award (2007071; PI, N. S. S.). Additional support for this study was provided by the Howard Hughes Medical Institute KwaZulu-Natal Research Institute for Tuberculosis and HIV/AIDS (55006543; PI, P. M.) and the National Institutes of Health Fogarty International Center (1D43TW008264-01). No funding source played a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Potential conflicts of interest.All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.