Published ahead of print on December 18, 2003, doi:10.1164/rccm.200306-856OC
© 2004 American Thoracic Society
Clustered Tuberculosis CasesDo They Represent Recent Transmission and Can They Be Detected Earlier?Departments of Tuberculosis Control and HIV & STI Research, Municipal Health Service; Department of Human Retrovirology, Academic Medical Centre, Amsterdam; Diagnostic Laboratory for Infectious Diseases and Perinatal Screening, National Institute of Public Health and the Environment (RIVM), Bilthoven, The Netherlands Correspondence and requests for reprints should be addressed to Henk van Deutekom, M.D., Department of Tuberculosis Control, Municipal Health Service, P.O. Box 2200, 1000 CE Amsterdam, The Netherlands. E-mail: hvdeutekom{at}gggd.amsterdam.nl
Clustered tuberculosis cases with Mycobacterium tuberculosis isolates showing identical restriction fragment length polymorphism patterns are assumed to be the result of disease transmission. In a prospective, population-based study in the province of North Holland, The Netherlands, we combined molecular methods with highly detailed epidemiologic information to determine why many clustered cases are not detected at an early stage. Of 481 patients, 138 (29%) fell into 43 clusters, suggesting recent transmission in 20%. Of 155 patients in clusters occurring within 2 years before or after the diagnosis of the disease, 21 (14%) had no epidemiologic links with other patients. Independent predictors of the absence of such links were female sex and Turkish, Moroccan, or other African ethnicity. Of 47 patients with a clear epidemiologic link, 37 (24% of 155) were identified early, e.g., by contact tracing, and 10 (6%) were missed. In 85 (55%) patients, an epidemiologic link was likely but undetected when using conventional contact tracing. Compared with clearly linked patients, only male sex was independently associated with presence in this last group. Our results indicate that 86% of clustered study patients had epidemiologic links and that opportunities for earlier identification using conventional tuberculosis control strategies are limited.
Key Words: tuberculosis • transmission • DNA fingerprinting • epidemiology • contact tracing Genetic fingerprinting of Mycobacterium tuberculosis isolates using restriction fragment length polymorphism (RFLP) has been used extensively in studies of tuberculosis transmission (1–4). These studies have demonstrated that M. tuberculosis isolates from epidemiologically linked patients showed identical RFLP patterns. Such "clustered" cases of tuberculosis appeared to be the result of recently transmitted infection with rapid progression to clinical disease. RFLP technology has also been applied to large cohorts of patients with tuberculosis, mostly in urban areas in industrialized countries, finding that a substantial proportion (28–72%) of urban cases occur in clusters (5–15). These results have been interpreted to demonstrate that up to 40% of tuberculosis cases are the result of recent transmission. In one study, only 10% of such cases had been identified by conventional contact tracing (5). Recently, a 2.5-year study in Amsterdam, The Netherlands, found that 47% of 459 patients with culture-proven tuberculosis fell into 53 clusters (16). Assuming that one patient in each cluster is the index patient, the proportion of tuberculosis cases transmitted during the study period was calculated to be 35%. Only 5.6% of these cases had been identified by conventional contact tracing. The results from such studies suggest that the conventional tuberculosis control strategies are insufficient to prevent ongoing transmission of tuberculosis in urban areas in industrialized countries. However, clustering of tuberculosis cases might not always be the result of recently transmitted infection. Well conserved M. tuberculosis strains with stable and identical DNA fingerprints can be found among patients who have never been in contact (17–19). In this prospective, population-based study of patients with tuberculosis in Amsterdam and the surrounding province of North Holland, we combined DNA fingerprinting of M. tuberculosis isolates with highly detailed epidemiologic information. We investigated in which patients clustering was an indication of epidemiologic linkage and therefore of recent transmission of infection. Furthermore, we determined which of these patients had been identified by conventional contact tracing, which were missed, and why they had been missed. Our main objective was to investigate why so many clustered cases are not detected by conventional contact tracing and to find opportunities for earlier identification. Some of the results of this study have been reported previously in the form of an abstract (20).
Study Population The city of Amsterdam is a part of the province of North Holland, which consists of both urban and rural areas. In 1999, the tuberculosis case rate in Amsterdam (population 727,053) was 32.5 per 100,000 people; in the rest of the province (population 1,776,147), it was 8.6 (21, 22). Our study population included all 664 patients with tuberculosis residing in North Holland who were diagnosed between July 1, 1998 and July 1, 2000 and reported to the Municipal Health Services, as is mandatory in The Netherlands. Patients were routinely interviewed about their social and medical histories by a public health nurse, using an extensive standardized questionnaire. If a patient had died or was too young to be questioned, a close family member was interviewed (n = 110 or 17%). At this interview, as part of our study, informed consent was requested for a second, nonroutine interview in case the patient was found to be part of a cluster or was nonclustered against the expectations of our research team. We combined the results of the first interviews with sociodemographic and clinical data. On the basis of this information, our multidisciplinary research team assessed each case as to how, where, and when the patient could have been infected with tuberculosis and to whom further transmission could have occurred. The M. tuberculosis isolates from patients with culture-positive tuberculosis (n = 483; 73% of the total number of patients) were then subjected to RFLP analysis, and results were compared with our previous assessments regarding epidemiologic links to other cases. For each cluster, a profile was prepared on the basis of the characteristics of patients in that cluster. Patients who had given informed consent were interviewed again if they were part of a cluster with one or more patients in North Holland who had been diagnosed during our study period and/or within the 2-year period before the study patient was diagnosed. Furthermore, patients had a second interview if our initial assessment had placed them in one cluster but RFLP analysis put them in another cluster or if their M. tuberculosis isolate had a unique DNA fingerprint pattern. During this second interview, we discussed with the patient the profiles of other cluster members without mentioning their names. Full attention was paid to circumstances under which he/she could have been infected by other cluster members or could have induced further transmission. Thereafter, final conclusions were drawn by the research team with regard to the most likely transmission pattern(s).
RFLP Analysis Identical strains recovered from different patients comprised a cluster of cases. Strains found in only one person were considered unique.
Statistical Analysis
These assignments were on the basis of the team's initial assessments as to how the patients could have been infected and whether they could have given rise to further transmission; the results of RFLP typing, which were compared with the initial assessments; and our final conclusions concerning the most likely transmission patterns, drawn after the second interview. Table 1 shows the sizes and characteristics of the five groups. Logistic regression analyses were performed to explore differences between the groups. The following variables were considered as potential determinants of being part of one of the groups: sex and age; ethnicity, defined as country of origin of patients' mother—Dutch, Turkish/Moroccan, Surinamese/Netherlands Antillean, Asian, African other than Moroccan, and other countries; nationality, defined by same categories as ethnicity; duration of stay in The Netherlands; place of residence (Amsterdam, outside Amsterdam); illicit drug use (injecting and noninjecting); homelessness; human immunodeficiency virus infection; alcohol abuse (
General Characteristics Between July 1, 1998 and July 1, 2000, 664 persons were diagnosed with tuberculosis in the province of North Holland. Of these, 386 (58%) were males. The mean age of the patients was 38 years (SD: 19 years); 480 (72%) were of foreign ethnicity, originating largely from countries with a high prevalence of tuberculosis. Sites of tuberculosis were: pulmonary in 340 (51%), extrapulmonary in 268 (40%), and both sites in 56 (9%) patients. The majority of the patients (414 or 62%) were residents of Amsterdam.
Clusters The 138 patients who were part of a cluster in North Holland fell into 43 clusters ranging from 2 to 12 patients, with 54 (39%) being in 27 small clusters (2 patients), 61 (44%) in 14 medium-sized clusters (3–9 patients), and 23 (17%) in 2 large clusters (11 and 12 patients). Assuming, in accordance with other studies, that in each cluster one patient acted as an index patient for the other(s), the number of tuberculosis cases probably acquired by transmission during the 2-year study period was 95 (138–43), accounting for 20% of 481 culture-proven cases in North Holland. Of these 95 recently infected patients, 19 (20%) had been identified by conventional contact tracing.
Clustering on the Basis of DNA Fingerprinting versus Expectations Only two patients were assigned to Group 5 (i.e., not clustered with the expected other patients). One was diagnosed with smear-negative, culture-positive pulmonary tuberculosis 3 months after smear-positive pulmonary tuberculosis was diagnosed in his brother, with whom he was in frequent contact. The other patient was a worker in a homeless shelter, who was diagnosed with tuberculous pleurisy during a contact investigation. Both patients had M. tuberculosis strains unrelated to those of their putative index patients; each was part of another cluster. Table 2 shows the determinants of belonging to a cluster if no epidemiologic links with other cluster members could be found (Group 4). By univariate analysis, female sex, ethnicity from Turkey or Morocco and from other African countries, a shorter duration of residency in The Netherlands, residency outside Amsterdam, and being an asylum-seeker were significantly associated with being part of Group 4. Multivariate analysis revealed that female sex and Turkish or Moroccan and other African ethnicity were independent risk factors for being part of such a cluster.
Among clustered patients in whom an epidemiologic link was evident at diagnosis (Groups 1 and 2) or became likely after RFLP typing (Group 3), we assessed the determinants of being undetected in an earlier stage by comparing Group 3 with Groups 1 and 2. As shown in Table 3, univariate analysis found that male sex, age over 34 years, and illicit drug use made patients more likely to lack opportunities for early identification. In multivariate analysis, only male sex was an independent risk factor for being undetected in an early stage.
The number of patients in Group 2 (n = 10) was too small to allow analyses using logistic regression. These patients, who could have been identified by contact tracing, were missed for the following reasons: three were not mentioned by the index patient despite being close contacts; three did not participate in a contact investigation despite repeated notices, after having been exposed to an infectious index patient; and four were not called by the Department of Tuberculosis Control for contact investigation due to administrative error (2) or misjudgment as to the frequency and intimacy of contact (2).
We studied the transmission dynamics of tuberculosis and evaluated our conventional contact-tracing strategies in the province of North Holland, composed of both urban and rural areas. Prospectively collected, highly detailed, epidemiologic data were combined with RFLP analysis of M. tuberculosis isolates. On the basis of the results of other molecular epidemiologic studies in industrialized countries, concern has been expressed about the effectiveness of conventional tuberculosis control measures, such as contact tracing after the ring principle (27). The relatively high number of tuberculosis cases found in clusters has generally been interpreted as a result of recent transmission. This interpretation suggests shortcomings in conventional contact tracing, which detected only a minority of these cases (5, 6, 8, 11, 13, 16). Our findings suggest that in the vast majority (86%) of patients, clustering indeed represents recent transmission. However, they also suggest that opportunities for early identification of clustered patients by using conventional tuberculosis control strategies are limited. In only 21, or 14%, of clustered patients (Group 4), even meticulous evaluation of all available data revealed no epidemiologic link with other patients in the same cluster. Relatively many patients in this group share characteristics such as origin in Turkey, Morocco, or other African countries, a relatively short stay in The Netherlands, and being an asylum-seeker. However, only five (3%) of all our clustered patients were asylum-seekers. The Group 4 clustering probably reflects reactivation of latent tuberculosis infections shortly after arrival in The Netherlands. These infections were independently acquired in their country of origin, where the population of M. tuberculosis is genetically more homogeneous. With the exclusion of such patients from the determination of the clustering rate in our study area, the rate decreased slightly from 29 to 25%; the proportion of the patients who acquired tuberculous infection during our study period decreased from 20 to 17%. For the remaining 134 patients (155 minus the 21 of Group 4) in whom clustering was likely to represent recent transmission, we determined those in whom conventional tuberculosis control strategies could have provided opportunities for early detection. The group of patients not offering such opportunities, Group 3, was rather large: 85 or nearly 64%. The composition of this group was very heterogeneous but included a majority of patients with an increased risk of tuberculosis, such as illicit drug users and human immunodeficiency virus–infected persons. Factors such as male sex, age over 34 years, and illicit drug use were associated with the disease not being detected in an early stage; male sex was the only independent risk factor. Much tuberculosis transmission in our study area appears to occur casually, in local shops and public transport, making targeted contact tracing barely feasible. Furthermore, contact tracing is difficult to perform among groups with an increased risk of tuberculosis and therefore of transmitting the disease, such as illicit drug users, homeless persons, and human immunodeficiency virus–infected persons. They often share social settings and/or living facilities, where tuberculosis may spread rapidly. Index patients from these groups often do not know their contacts by (full) name; these contacts, after being exposed to the index patient during the infectious period may have moved to untraceable places by the time the index diagnosis is made. Our extensive second interviews with patients from Group 3, after the results of the RFLP analysis became available, revealed in no case a well defined point of departure for targeted tracing, which could have identified them earlier, although transmission from the source case to these patients was very likely. It was disappointing to find no clear leads for earlier detection of disease among the groups most vulnerable to tuberculosis. The number of clustered patients offering opportunities for early identification (Groups 1 and 2) was rather small: 47, or 35%, of the epidemiologically linked patients. Of these, 10, or 21% (Group 2), had been missed by contact tracing due to inadequate response of the index patient, the contact, or the Department of Tuberculosis Control. Because of the small number of patients in Group 2, we found no significant patient characteristics that could induce more alertness to an increased risk of having been missed by contact tracing. It must be stressed, however, that Group 1, with 37 patients who were identified and linked to another patient as early as possible, mostly by contact tracing, is actually larger. The benefits of conventional control strategies are probably greater than this number suggests, for two main reasons. First, patients with active tuberculosis found by contact tracing often have limited disease, with negative cultures, and thus were not included in our analyses. In our study population, patients who were detected by contact tracing were less likely to have a positive culture than those who were detected otherwise (odds ratio, 0.18; confidence interval, 0.10–0.32; data not shown). Second, contact tracing and identification of latently infected contacts, followed by preventive chemotherapy, usually precludes active disease; thus, our study population did not include persons with recent infection, who were successfully traced before disease developed (28). Only two of the patients identified by contact tracing had M. tuberculosis strains, which were unrelated to their index case. In San Francisco, however, 30% of contacts with active disease had such strains (29). This may be explained by the fact that patients with tuberculosis in San Francisco more frequently belong to risk groups for tuberculosis, as do their contacts. Contact investigation among such groups may more often yield coincidental tuberculosis cases, unrelated to the index case, than in our study area. In The Netherlands, further reduction of the already small number of patients in whom diagnosis should have been made earlier but was missed by contact tracing (Group 2) would provide little benefit for the tuberculosis situation in general, although this number should be kept as low as possible to benefit individuals. The broadest advantage would be gained from the early identification, preferably before active disease develops, of patients in Group 3, who offered no opportunities for early detection using conventional contact tracing. This would require large-scale screening of population groups because only male sex was found to be an independent risk factor for being in Group 3. Expansion of screening programs beyond the well-known risk groups, such as illicit drug users and homeless persons, offers little promise when the epidemiology of tuberculosis in countries like The Netherlands is affected more by immigration from countries with a high prevalence of tuberculosis than by a home-grown chain of transmission. The energy and resources for such interventions could probably better be spent on tuberculosis control in countries with a high prevalence of the disease.
The authors thank the public health nurses and doctors of the Department of Tuberculosis Control of the Municipal Health Services in North Holland for their help and collaboration, as well as Miriam Dessens and Mimount Enaimi of the National Institute of Public Health and the Environment (RIVM) for their important contribution regarding the restriction fragment length polymorphism typing of the Mycobacterium tuberculosis isolates and the DNA fingerprint analysis. The authors also thank Lucy D. Phillips for editorial review.
Supported by the Netherlands Organization for Health Research and Development (ZonMw). Conflict of Interest Statement: H.V. has no declared conflict of interest; S.P.H. has no declared conflict of interest; P.E.W.D. has no declared conflict of interest; M.W.L. has no declared conflict of interest; A.H. has no declared conflict of interest; D.V. has no declared conflict of interest; R.A.C. has no declared conflict of interest. Received in original form June 27, 2003; accepted in final form December 16, 2003
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
5 U/day); being an asylum-seeker or other recent immigrant; site of tuberculosis (pulmonary, extrapulmonary, both sites), and results of sputum smear (only in case of pulmonary tuberculosis). Furthermore, we assessed how the disease was diagnosed in the patient: by symptoms, source/contact tracing, screening of risk groups, others. After building univariate models, variables with a p value less than 0.10 were considered as candidate variables for the multivariate models. These were built in a stepwise manner, and all first-order interaction terms were checked to see whether they improved the final multivariate model. A p value less than 0.05 was considered statistically significant. 
