INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The incidence of tuberculosis is influenced by several factors acting individually or together (1). In the United States, an increasing proportion of tuberculosis cases arises from the foreign-born population; thus, birth in a country with a high incidence of tuberculosis is an important risk factor for tuberculosis (2, 3). Other identified "high-risk" factors for tuberculosis that are more prevalent in the U.S.-born population include human immunodeficiency virus (HIV) infection, homelessness, and drug abuse (4). These factors, along with a deterioration in the public health infrastructure, are believed to have fueled the resurgence of tuberculosis in the late 1980s. During the past 4 yr, however, the rate of tuberculosis in the United States has decreased. This decrease has been attributed to increased funding for and improvement in tuberculosis control programs (2).

As the United States moves back on track toward the elimination of tuberculosis, continuing gaps in our tuberculosis control efforts must be identified so strategies can be developed to further reduce the incidence of tuberculosis. On the basis of available information, such strategies may have to be different for the U.S.-born and the foreign-born populations. For instance, in the foreign-born population, an expanded use of preventive therapy may be needed (7). But among U.S.-born persons at risk for tuberculosis, many of whom have associated characteristics such as HIV infection or homelessness, studies have suggested the need to control nosocomial transmission of tuberculosis or conduct active case-finding in homeless shelters (8, 9). However, few studies have directly determined which factors are still fueling tuberculosis incidence in a given population, and none has addressed the differences that might exist in the U.S.- and foreign-born populations.

We set out to determine those factors that are contributing to tuberculosis incidence in the U.S.- and foreign-born populations in San Francisco. We used 5-yr of epidemiologic and tuberculosis genotyping data to identify a cohort of patients with strains of Mycobacterium tuberculosis not previously identified in the city and to elucidate their secondary cases. The amount of disease from transmission of M. tuberculosis within and between the U.S.- and foreign-born populations were determined and the reasons for development of secondary cases were ascertained. Our findings suggest tuberculosis control strategies most applicable for further control and perhaps the eventual elimination of tuberculosis in San Francisco.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population and Routine Data Collection

Since 1991, we have been conducting a population-based study of the molecular epidemiology of tuberculosis in San Francisco. Data were collected on all patients in whom bacteriologically confirmed tuberculosis was diagnosed and reported to the San Francisco Department of Public Health between January 1, 1991 and December 31, 1995, and from whom M. tuberculosis isolates were available for genotyping. The basic components of this study, including the collection of patient data and the genotyping of M. tuberculosis isolates, have been previously described (10).

The availability of 5 yr of molecular epidemiologic data from San Francisco allowed us to identify a cohort of patients infected with strains of M. tuberculosis not previously identified in the city and elucidate their secondary cases. After matching the restriction fragment length polymorphism (RFLP) pattern of all isolates to each other, we excluded all cases in 1991 and 1992 because we could not be certain if their strains of M. tuberculosis were recently introduced (within the previous 2 yr) into the population. For instance, a strain identified for the first time in 1992 may or may not have been introduced into the city within the previous 2 yr since genotyping data from 1990 were not available for comparison. We also excluded cases reported after 1992 if the RFLP pattern of their M. tuberculosis isolate matched the pattern of an isolate from a case diagnosed in 1991 or 1992. We did this because we were not interested in the transmission consequences of these earlier cases. Furthermore, because the transmission consequences of a case cannot be ascertained without adequate follow-up, cases from 1995 were excluded unless their strain was initially identified during 1993 and 1994. Finally, we excluded cases meeting predetermined criteria for having resulted from laboratory cross-contamination (11). The cohort, therefore, included only cases from 1993 and 1994 caused by strains not identified in the city previously and cases from 1995 that appeared to be the result of transmission from a 1993 or 1994 case.

The data collected routinely for all cases, including all items on the Report of a Verified Case of Tuberculosis (RVCT), have been previously described (10, 12). In addition to cross-matching the tuberculosis and AIDS registries, patient records were reviewed to collect information on a diagnosis of HIV infection or AIDS. The Department of Public Health routinely conducts investigation of contacts to a case of pulmonary tuberculosis.

Collection of M. tuberculosis Isolates and RFLP Analysis

The collection of M. tuberculosis isolates, the performance of IS6110-based RFLP analysis, and the data management have been previously described (10, 13, 14). Isolates with identical IS6110 patterns that had more than five bands were considered to be clonal. Isolates with identical IS6110 patterns that contained five or fewer bands were analyzed with a second genotyping technique utilizing a probe for the polymorphic guanine-cytosine rich sequence (PGRS) (15). The DNA was digested with Sma1 and probed using a 34 bp probe based on the consensus sequence of the PGRS (ATCGGCAACGGCGGCAACGGCGGCAACGGCGG). On the basis of visual comparison, isolates with identical number, relative intensity, and molecular weights of the bands were considered clonal.

Epidemiologic Investigation of Clusters

Because clustering by RFLP analysis may not necessarily indicate transmission, we collected epidemiologic data to determine if transmission of M. tuberculosis may have occurred among clustered cases. After determining the likely duration of infectiousness, we classified the source and secondary cases for each cluster. We then assigned a likelihood that transmission linkages existed between the putative source and secondary cases.

To collect the epidemiologic data, we reviewed clinical records, information on contact investigations conducted by the Department of Public Health, and hospital records of patients if any person in the cluster was hospitalized during his or her infectious period (defined below). In addition, one third of the clustered patients were part of a prospective study in which they were interviewed concerning their whereabouts for the 2 yr prior to the diagnosis of tuberculosis.

For each clustered patient, we determined the likely duration of infectiousness. For patients with smear-positive pulmonary disease, the duration of infectiousness was defined from the onset of symptoms as indicated in the medical record or 4 wk before the collection of the initial smear-positive respiratory specimen, whichever occurred earlier, to the date sputum smears became consistently negative. For patients with smear-negative pulmonary disease, the duration of infectiousness was defined from the onset of symptoms or 2 wk before the collection of the initial smear-negative respiratory specimen, whichever occurred earlier, to 2 wk after the start of antituberculosis chemotherapy that was effective in rendering sputum cultures consistently negative. Patients with only extrapulmonary tuberculosis were considered to be noninfectious.

We used predetermined criteria to identify the possible source and secondary cases in each of the clusters. The putative source case must have an infectious period that started more than 3 mo prior to the start of the infectious period of a secondary case unless the secondary case was infected with HIV; in that instance, the difference in the beginning of the infectious periods was shortened to 4 wk.

For each clustered patient, we assigned a likelihood that transmission linkages existed with other members in the cluster (16). A transmission link was "definite" when there was documentation that patients were in contact with each other when one or more of them was likely to have been infectious. A "possible" transmission link was assigned when the documentation was insufficient to meet the criteria for a definite transmission link but patients were in the same or similar setting in San Francisco (homeless shelter, substance-abuse facility, jail) or had the same type of sociobehavioral characteristic such as being homeless or using the same type of illicit drug. "Unable to determine" was assigned when there was no evidence to either suggest or definitively exclude transmission between clustered patients. Transmission was considered "improbable" when available evidence suggested that transmission could not have occurred between clustered patients. All persons in clusters with "definite" or "possible" transmission linkages were considered to have been involved in the development of disease from recent transmission (within the previous 2 yr) in San Francisco.

Using the results of the epidemiologic assessment, all secondary cases were attributed to either a U.S.- or foreign-born source case. All nonsecondary cases were also classified as "initial" cases. "Initial" cases were those that did not result from recent infection in San Francisco according to both epidemiologic and genotyping data, and included cases with unique RFLP patterns, clustered cases determined by epidemiologic analysis to be a source case, and clustered cases with transmission linkages that were deemed "unable to determine" or "improbable." We determined the percentages of initial, source, and secondary cases, the ratio of secondary to initial cases, and the ratio of source to other initial cases. Finally, we determined the major reasons why secondary cases developed. The two-tailed chi-square test was used to compare proportions.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population and RFLP Patterns Obtained

From 1991 to 1995, 1,567 cases of tuberculosis were diagnosed and reported to the San Francisco Department of Public Health, 1,365 (87.1%) of which were confirmed by the isolation of M. tuberculosis. Viable isolates of M. tuberculosis from 1,177 patients (86.2%) were available for IS6110-based RFLP analysis. Of the 218 isolates with less than six copies of IS6110, 189 had PGRS genotyping performed. Overall, 1,148 of the 1,365 patients (84.1%) with culture-positive tuberculosis had their isolates typed by IS6110 and, if necessary, PGRS typing. Patients without genotyping results were more likely to be foreign-born or older than 65 yr of age, and less likely to have a diagnosis of AIDS, a recent history of homelessness, or illicit drug use (all with p < 0.05).

From the 1,148 cases with genotyping results, we excluded all the 511 cases diagnosed in 1991 and 1992 because we could not be certain if their strains of M. tuberculosis were recently introduced (within the previous 2 yr) into the population. Among the remaining 637 cases, which were all diagnosed after 1992, we further excluded 117 cases because the RFLP pattern of their M. tuberculosis isolate matched the pattern of an isolate from a case diagnosed in 1991 or 1992. We did this because we were not interested in the transmission consequences of these earlier cases. We also excluded 145 cases diagnosed in 1995 because the RFLP pattern of their M. tuberculosis isolate did not match the pattern of an isolate from a case diagnosed in 1993 or 1994. We did this because the transmission consequences of a case cannot be ascertained without adequate follow-up. Finally, we excluded eight cases because they met the criteria for laboratory cross-contamination. The final study cohort of 367 cases included only cases from 1993 and 1994 caused by strains not identified in the city previously and cases from 1995 that appeared to be the result of transmission from a 1993 or 1994 case (Table 1).

Clustering of Cases by Genotyping

Of the 367 cases, 289 (78.7%) had unique RFLP patterns. The other 78 cases (21.3%) with clonal organisms were grouped into 29 clusters. There were 17 two-person clusters, six three-person clusters, four four-person clusters, and two five-person clusters. Ten clusters (23 persons) contained only foreign-born persons, 12 (37 persons) contained only U.S.-born persons, and seven (18 persons) contained both foreign and U.S.-born persons. Among the 252 foreign-born cases in our study cohort, 221 (87.7%) had unique RFLP patterns. Among the 115 U.S.-born cases, 68 (59.1%) had unique RFLP patterns.

Likelihood of Transmission between Clustered Cases

After epidemiologic investigation of the clusters, we assigned a likelihood of transmission to each case (Table 2). Of the 78 clustered cases, only 30 (39%) had "definite" or "possible" transmission linkages with another clustered case. Of the 30 cases with some evidence of transmission linkages, 28 (93%) were born in the United States. U.S.-born clustered cases were much more likely to have transmission links than were foreign-born clustered cases (28 of 47 and two of 31, respectively, p < 0.0001). Of the 11 clusters with "definite" or "possible" transmission linkages, nine contained only U.S.-born persons; the other two clusters had a mixture of U.S.- and foreign-born persons (Table 3).

Thirteen clustered cases were classified as having "improbable" transmission linkages in San Francisco. Three cases were not in San Francisco at a time when other persons in the cluster were infectious; two of them arrived in San Francisco as immigrants under the "B-waiver" program and, upon their arrival, had abnormal chest radiographs and positive sputum cultures for M. tuberculosis (17). The isolates from two other cases were resistant to two or more first-line antituberculosis drugs, whereas isolates from other cases in the same cluster were drug-susceptible. Because none of those with drug-susceptible disease developed drug-resistant disease during their treatment, it is unlikely that they transmitted their M. tuberculosis organisms to patients with drug-resistant disease. Another three cases were the first case diagnosed in a cluster but had only extrapulmonary tuberculosis. Although our analysis of the sequence of disease development put these cases as the source case, it is highly improbable they were source cases. Finally, five of the above eight cases were in two-person clusters. As a result, the other five cases in these clusters were not considered to have been involved with transmission of M. tuberculosis in San Francisco.

Of the 13 cases with "improbable" transmission links, 11 were foreign-born and 10 were in clusters with only foreign-born persons. In fact, foreign-born persons were in all the "improbable" clusters. Overall, 35% of the foreign-born persons in clusters (11 of 31 cases) were not considered to have been involved with transmission in San Francisco; this percentage is much higher than the 4% (2 of 47 cases) observed among U.S.-born cases (p < 0.001). Transmission could neither be established nor excluded in 17 (36%) of the U.S.-born cases and 18 (58%) of the foreign-born cases in clusters (p = 0.20). The prospective interview of one third of the clustered cases did not reveal additional information on transmission between these cases beyond that obtained from the retrospective review of public health and medical records.

Mechanisms for Development of Disease

On the basis of the epidemiologic investigation of clusters, we determined which were the source and which were the secondary cases for 10 of the 11 clusters with "definite" or "possible" transmission linkages (Table 3). In one two-person cluster (Cluster 7), we could not determine who the index case was because the infectious periods of the two cases overlapped. However, since both patients were born in the United States, we were able to attribute all secondary cases in the 11 clusters to either a U.S. or foreign-born index case (Figure 1).

View larger version (31K): [in this window] [in a new window]   Figure 1.   Transmission of Mycobacterium tuberculosis within or between the U.S.- and foreign-born populations. Arrows indicate the direction of transmission from a source to a secondary case. Persons unrelated to secondary cases had unique tuberculosis genotyping results or absence of epidemiologic linkages to any other patients during the study period. Number of patients in each catagory is in parentheses.

During 1993 and 1994, there were 348 initial cases of tuberculosis. By the end of 1995, an average follow-up of 2 yr (median, 756 d), 19 secondary cases could be attributed to these initial cases (Figure 1). Only 2% (two of 115) of the tuberculosis cases in U.S.-born persons could be attributed to foreign-born persons, and none of the cases in foreign-born persons could be attributed to U.S.-born persons (Figure 1).

All 252 foreign-born cases were initial cases, suggesting that tuberculosis in these patients resulted from reactivation of latent infection acquired prior to their arrival in San Francisco. Of the 115 U.S.-born cases, 96 (83.5%) were initial cases (p < 0.0001 compared with foreign-born persons). Among the 19 U.S.-born secondary cases, 17 (89%) could be attributed to source cases who were born in the United States.

Cases attributable to U.S.-born persons had a much higher ratio of secondary to initial cases than cases attributable to foreign-born persons (1:5.6 versus 1:126, respectively, p < 0.0001). Among the 96 U.S.-born initial cases, nine (9.4%) were source cases, whereas only two (0.8%) of 252 initial foreign-born cases were source cases (p < 0.001).

Among the 30 clustered cases, 23 had AIDS, 21 used illicit drugs, 13 were homeless, and 28 had one or more of the aforementioned "high-risk" characteristics. The two clustered cases without any of these conditions were foreign-born, leaving all 28 U.S.-born clustered cases with at least one of these "high-risk" characteristics. Among the 19 secondary cases, 18 had AIDS, 17 abused illicit drugs, and nine were homeless.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this molecular epidemiologic study, we have determined how various factors contribute to the community epidemiology of tuberculosis in San Francisco. We have shown that there are currently two parallel epidemiologic patterns of tuberculosis occurring simultaneously in the cityone in the foreign-born population and one in the U.S.-born population. Although these parallel epidemiologic patterns can be described by differences in the demographic and sociobehavioral characteristics, more important are differences in the mechanism by which tuberculosis is developing in the two populations. Understanding such mechanisms has permitted us to establish priorities for further control and hopefully the eventual elimination of tuberculosis in San Francisco.

The migration of persons already infected with M. tuberculosis into San Francisco appears to be responsible for most, if not all, foreign-born tuberculosis cases in the city. Because most patients came from countries with a high prevalence of tuberculosis, we conclude they were likely to have been infected prior their arrival into the United States. This conclusion is consistent with results of other studies that have shown an association between exposure to M. tuberculosis prior to immigration and subsequent risk of tuberculosis after immigration (7, 18, 19). Our study, however, has provided the most direct evidence to date supporting the contention that most foreign-born persons develop tuberculosis from reactivation of infection acquired prior to their arrival into the United States.

In contrast, nearly 20% of U.S.-born cases developed disease after recent infection in San Francisco, and an undetermined additional percentage may have resulted from recent infection outside of San Francisco. The greater amount of disease from recent transmission in the U.S.-born population is also reflected in the percentage of U.S.-born cases that produced secondary cases, more than 11 times that observed in foreign-born cases. This suggests a greater percentage of U.S.-born cases are transmitting M. tuberculosis or the U.S.-born population is more likely to acquire tuberculous infection and rapidly develop disease, or both are occurring.

Within San Francisco, factors such as HIV infection, homelessness, or drug abuse have a substantial impact on the development of tuberculosis. During our study period, we found that, in every instance, disease resulting from the transmission of M. tuberculosis within San Francisco involved either a source or a secondary case with these conditions. Although we and others have reported that persons with these conditions are involved in the transmission of M. tuberculosis (8-10, 20- 22), this is the first study to document that the propagation of tuberculosis in an urban community can be largely contained to subpopulations with these distinct characteristics, thereby identifying a focus for future tuberculosis control efforts.

When we evaluated which specific control measure failed, the most important problem identified was the failure of contact investigation to identify infected contacts. Contact investigations associated with the source cases failed to identify the eventual secondary case in nearly 70% of our clusters, even though some source and secondary cases clearly knew each other. But even after being identified by the health department as contacts to an infectious case, several persons still developed tuberculosis. We found instances where a delay in the diagnosis of the source case meant there was insufficient time for contacts to be evaluated and given preventive therapy before they developed disease. We also found that failure to evaluate a contact and failure to ensure adherence to preventive therapy resulted in secondary cases.

Interestingly, we found very little disease resulted from transmission of M. tuberculosis between the U.S.- and foreign-born populations. Several studies have described the epidemiology of tuberculosis in foreign-born persons, but few studies have explored the epidemiologic interaction with the native population (3, 23, 24). On the basis of state-specific tuberculosis rates for U.S.- and foreign-born persons, McKenna and colleagues (3) concluded that transmission from foreign-born to U.S.-born persons is probably not extensive. Yang and colleagues (25), using RFLP analysis, identified transmission of M. tuberculosis between persons from Greenland and Denmark, but not from other foreign countries. Our study presents a more comprehensive analysis of how persons from different birthplaces and with varying prevalence of high-risk conditions influence the epidemiology of tuberculosis than has heretofore been published.

The epidemiologic separation between the U.S.- and foreign-born populations is likely the result of several factors. From a sociocultural perspective, new arrivals into the United States who have the highest rates of tuberculosis are more likely to live with other foreign-born persons in their own ethnic communities; this reduces the likelihood that tuberculosis would be transmitted from these patients to other U.S.-born persons. Biologically, persons born in countries where the incidence of tuberculosis is high are likely to have been infected with M. tuberculosis prior to their arrival in the United States and thus are relatively resistant to new infection, including those from U.S.-born patients (26, 27). In addition, conditions such as HIV infection, homelessness, and drug abuse, which could increase the risk of tuberculous infection or the likelihood of developing disease, or both, are infrequent in the foreign-born population in San Francisco.

Before we discuss the implications of our findings, several limitations of our study should be mentioned. We may have underestimated the amount of disease from transmission of M. tuberculosis within the city for the following reasons: the retrospective collection of data may have failed to identify epidemiologic relationships that actually existed; cases without culture-confirmation may have resulted from transmission in the city but would not have been included in this analysis; cases developing more than 2 yr after infection would not have been identified; and we used very conservative criteria for matching RFLP patterns and therefore could have failed to identify persons with the same strain of M. tuberculosis. We also could have overestimated the amount of disease from transmission in the city since clustered cases linked solely on the basis of sociobehavioral characteristics or being in similar locations may not have actual transmission linkages. However, there is growing evidence that such cases are involved in recent transmission of M. tuberculosis (9, 16, 21).

There may also be limitations to the generalizability of our results. Our findings were derived from an area with a well-functioning tuberculosis control program, and the study was conducted during a period when the community transmission of tuberculosis was decreasing. Compared with results from the 1991-1992 period (10), both the number and size of RFLP-linked clusters have decreased. Therefore, our results may not apply to communities with a poorly functioning tuberculosis control program. In addition, our results may not apply to areas where the foreign-born population has a higher prevalence of HIV infection, homelessness, or drug abuse. In such areas, disease from recent transmission may be responsible for a higher proportion of secondary foreign-born cases. In a recent report that described transmission in San Francisco primarily during 1991 and 1992, more than 20% of the tuberculosis cases in Mexico-born persons resulted from recent infection, and most had the high-risk characteristics listed above (28).

By understanding the effectiveness and limitations of current tuberculosis control efforts in San Francisco we can establish priorities for further control and eventual elimination of tuberculosis in the city. First, in the foreign-born population, disease from recent transmission of M. tuberculosis has virtually been eliminated. No doubt current control efforts have contributed to this. But these efforts are unlikely to be able to further control and eliminate tuberculosis in this population. Within the city, there is a large reservoir of foreign-born persons infected with M. tuberculosis and more infected immigrants will arrive in the future. Disease from reactivation of latent infection acquired outside of the city will continue to occur unless foreign-born persons infected with M. tuberculosis are identified and given preventive therapy. In our study, most foreign-born persons with tuberculosis did not meet current criteria for the isoniazid prophylaxis because of their age and the absence of other risk factors (29). But if isoniazid prophylaxis was given to older (> 35 yr of age) foreign-born tuberculin reactors as recommended by a recent cost-effectiveness analysis (30), some of the cases of tuberculosis might have been prevented. Our study complements this economic analysis by providing an epidemiologic and programmatic argument for increasing tuberculosis screening and isoniazid prophylaxis for the foreign-born population older than 35 yr of age.

Second, to further control tuberculosis in the U.S.-born population, there must be an intensification of contact investigation. In San Francisco, this intensification should focus on improving the identification of contacts and accelerating the pace of contact investigation, especially among persons with HIV infection, drug abuse, or homelessness. Persons with these risk characteristics are frequently involved in the transmission of tuberculosis in urban settings (9, 10, 22). However, it is difficult to identify and follow-up on contacts to drug users or homeless persons. These people are frequently suspicious of health care workers and may not know or divulge the names of their contacts; and their contacts may not have close ties with the index case. Because a high percentage of our secondary cases were infected with HIV, the pace of contact investigation must be accelerated. The current pace of contact investigation appears adequate for contacts who are not HIV-infected, but the pace is clearly insufficient for those who are infected with HIV given a shorter period of latency from infection to disease (20).

Third, to further control and eventually eliminate tuberculosis in the U.S.-born population in San Francisco, there should be an expansion of programs to screen for tuberculosis and tuberculous infection among HIV-infected persons, drug users, and homeless persons. Such screening programs are important because it appears that many infected contacts will not be identified by contact investigation. Because of this difficulty, some have suggested a shift from screening only known contacts to screening of social networks or in specific locations (9, 31). If applied to San Francisco, screening should be expanded in drug and alcohol treatment centers and shelters or hotels where unstably housed persons reside. But any screening program must also ensure completion of preventive therapy if it is to be effective in preventing tuberculosis. In the population that may benefit from screening, the use of directly observed preventive therapy or incentives may be the only way to ensure completion of preventive therapy (32, 33).

Fourth, our findings suggest that the transmission of M. tuberculosis from the foreign-born to the U.S.-born population can be minimized when services are available and administered effectively to treat and prevent tuberculosis in foreign-born persons. With the provision of such free services in San Francisco, foreign-born persons rarely infect U.S.-born persons.

Finally, because it may be difficult to prevent tuberculosis in foreign-born persons who were infected before their arrival in the United States, efforts to control tuberculosis in countries with a high prevalence of tuberculosis should be supported. While the impact of this approach may not be immediately apparent, the elimination of tuberculosis in the United States would likely be impossible without a reduction in the incidence of tuberculosis in other countries (34). In this regard, one can learn from the successful effort to eliminate polio from the Americas, which has required a multinational effort and partnership between resource-rich and resource-poor countries (35).