Tuberculosis from Mycobacterium bovis in Binational Communities, United States

The incidence in San Diego is increasing and is concentrated mostly in persons of Mexican origin.

T he pattern of tuberculosis (TB) in the United States is changing as the incidence of TB disease becomes more concentrated in foreign-born persons. Of the annual total US TB cases, >54% are now concentrated in persons born outside of the United States (1); in communities with high immigration, the proportion can exceed 70% (2). TB prevention and treatment strategies, particularly those in communities on the border with Mexico, will need to be adapted to accommodate the changing epidemiology of TB (3).
San Diego, California, together with its sister city Tijuana-Tecate, Mexico, is the largest binational metro-politan region in the United States, accounting for 34% of the southern border population (4). In San Diego County, the Hispanic population has grown from 20% to 29% in the past 15 years (5). Of the total annual TB cases in San Diego, >70% occurred among foreign-born persons, of whom nearly half originated from Mexico (2). A review of culture-positive TB cases in San Diego County in the late 1990s indicated that 6.6% of all adult TB cases and 39% of all pediatric (<15 years of age) TB cases from this region were not caused by Mycobacterium tuberculosis, the most common TB pathogen in the United States, but were instead caused by M. bovis, a pathogen more often associated with TB in cattle (6). This fi nding represented the highest reported proportional incidence of TB from M. bovis among industrialized countries (7).
M. bovis is a pathogen in the complex of bacteria that includes M. tuberculosis, which causes TB in humans and animals. TB from M. bovis has been generally considered rare in the United States after its successful eradication from cattle in the mid-1900s (8), but wider use of laboratory tools for species-level diagnosis of TB pathogens has started to shed light on an unexpected regional presence of M. bovis in communities with large Hispanic populations. While M. bovis TB has been most often documented in Hispanic communities with close proximity to Mexico (6,9), a recent review of M. bovis cases in New York City indicates that the problem is not limited to US regions that border Mexico (10).
The clinical and pathologic characteristics of M. bovis TB is indistinguishable from M. tuberculosis TB in most cases, but there are relevant considerations for prevention and treatment strategies in communities where M. bovis contributes to TB incidence. First, M. bovis is thought to be spread to humans primarily through consumption of raw  (11,12), with only minimal human-to-human transmission (13). Second, M. bovis is almost universally resistant to the key antituberculous drug pyrazinamide (PZA), which necessitates a 9-month treatment duration instead of the standard 6-month, short-course therapy, which is possible with PZA in the treatment regime. Third, higher mortality rates during treatment may be associated with M. bovis (14). Multidrug-resistant (MDR) strains of M. bovis (15)(16)(17), the high proportional incidence of M. bovis (9,18) in pediatric TB cases, and frequent HIV co-infection (19) are important additional considerations in developing effective treatment and prevention strategies for M. bovis. To document the trends and the effect of M. bovis on TB epidemiology, we examined TB case surveillance data from 1994 through 2005 in San Diego County and identifi ed risk factors related to M. bovis disease and deaths during treatment in the last 5 years.

Data Sources
This study used routine TB surveillance data from 1994 through 2005. The study protocol was approved by the Institutional Review Boards of San Diego State University and the University of California, San Diego. Demographic and clinical data were obtained from the Tuberculosis Information Management System (TIMS) database maintained by the San Diego County TB Control Program. Since the early 1990s, a TB isolate has been submitted to the county public health laboratory for every reported TB case. All TB isolates from patient specimens were initially identifi ed as M. tuberculosis complex on the basis of the AccuProbe hybridization protection assay (GenProbe, San Diego, CA, USA). Specimens were further identifi ed as either M. bovis or M. tuberculosis on the basis of culture morphologic fi ndings, the results of the niacin strip test, the nitrate reduction test, and the specimens' susceptibility to PZA (20). Furthermore, all isolates identifi ed as M. bovis from 2004 and 2005 were confi rmed to have spoligotypes consistent with M. bovis (21). Population data for San Diego County were obtained from San Diego Association of Governments' estimates based on census and calculated data.

Study Design
We conducted a retrospective trend analysis of all culture-positive TB cases in the San Diego County TIMS database from 1994 through 2005 that were confi rmed as either M. bovis or M. tuberculosis. We also conducted a detailed retrospective analysis of demographic and clinical variables associated with M. bovis case-patients and deaths during treatment from 2001 through 2005.
Demographic variables from the TIMS database used in the correlates and mortality analyses included sex, age, ethnicity, and country of birth. Clinical variables included: previous history of TB disease, presence or absence of pulmonary disease, presence or absence of multisite disease, presence or absence of acid-fast bacilli (AFB) in sputum smear, presence or absence of pulmonary lesions by chest radiograph, presence or absence of MDR TB, and HIV status.

Analysis Trends
Trends in TB incidence were evaluated by using Poisson regression with time in years as the predictor variable, case number as the dependent variable, and population size as an additional exposure variable. Trend lines for M. bovis and M. tuberculosis were based on incidence predicted by Poisson regression fi tted to the data. Trends in proportional incidence of M. bovis cases (relative to all TB cases) were assessed with a χ 2 test for trend.

Correlates of M. bovis Disease
Demographic and clinical variables shown previously to be associated with TB diagnoses (6) were compared between M. bovis and M. tuberculosis. Variables signifi cant at the 5% level by χ 2 test in univariate analyses were entered into a multiple logistic regression model. The fi nal model was derived by using the likelihood ratio method (22).

Analysis of Mortality Rates during Treatment
All deaths that occurred from the time that a TB case was reported until treatment was completed were documented with death certifi cates and recorded in TB case fi les. For the purposes of this study, causes of death in M. bovis and M. tuberculosis case fi les were transcribed from death certifi cates or California state death records and collated into 7 major causes of death based on the most common causes.
We investigated the apparently higher mortality rates during treatment among M. bovis cases relative to M. tuberculosis cases (14) by using a multiple logistic regression analysis with M. bovis as the exposure variable; death before treatment was completed as the outcome variable; and demographic and clinical variables as potential covariates. Univariate differences between causes of death in M. bovis and M. tuberculosis cases were analyzed with the Fisher exact test.  Table 1). The proportion of TB cases with culture-positive results remained relatively stable at ≈81% of annual reported TB cases.  Table 3).

Analysis of Mortality Rates during Treatment
Of 1,324 culture-positive TB case-patients, 1,119 were evaluated in the analysis of mortality rates during treatment. Fifteen percent (205/1,324) were excluded because of missing data on case survival, including patients who were lost to follow-up or moved during treatment. Of the 1,119 cases, 110 (19 M. bovis and 91 M. tuberculosis) patients died during TB treatment (n = 81) or before   2.55× (p = 0.01) as likely to die before treatment completion than M. tuberculosis patients, after differences in age, race and ethnicity, country of birth, chest radiograph abnormalities, multisite disease, and HIV status were accounted for (Table 4). Univariate analyses of the causes of death in M. bovis and M. tuberculosis cases showed no signifi cant differences (p>0.05) except for the category of "other noninfectious disease," which was overrepresented in the M. tuberculosis group (Table 5).

Discussion
From 1994 through 2005, incidence of M. bovis TB cases in San Diego County increased in absolute number, as a proportion of total TB cases, and relative to the population. In contrast, TB incidence caused by M. tuberculosis declined during the same period. M. bovis cases were concentrated in persons of Hispanic descent, especially those of Mexican origin, and among those <15 years of age, in whom M. bovis accounted for 45% of the culture-positive cases. Deaths during treatment were largely confi ned to adults and were twice as high in M. bovis TB case-patients when compared with M. tuberculosis patients.
Our fi ndings indicate that the incidence of TB caused by M. bovis in southern California is substantially higher than the national rate of 1.5% estimated from TB surveillance data (23) but is similar to the proportional incidence (13%) among Mexican-born case-patients in New York, New York. It was previously hypothesized that TB attributed to M. bovis in San Diego is most likely being driven by recent infections in children and largely reactivated latent infections in adults, secondary to HIV co-infection (6). Our fi ndings confi rm the continued high incidence of M. bovis in children >12 months of age, but the role of HIV co-infection in M. bovis case-patients relative to M. tuberculosis cases is less clear.
Almost half of the culture-positive pediatric TB cases in this binational region of >3 million persons were caused by M. bovis, which has clinical implications. Since M. bovis is intrinsically resistant to PZA, a critical component of the standard 6-month, short-course treatment for M. tuberculosis, M. bovis treatment is usually extended to 9 months of isoniazid and rifampin (14). In southern California, and perhaps other Hispanic communities with close ties to Mexico, empiric extended TB treatment for children without culture-positive disease, particularly those with a history of consuming unpasteurized dairy products, should be considered.
Although one quarter of the M. bovis TB case-patients were co-infected with HIV, HIV was not signifi cantly more associated with M. bovis TB compared with M. tuberculosis in our study or in the previous M. bovis study in this community (6). Confi dence in this fi nding is somewhat limited because 40% of the TB case-patients did not have their HIV status reported. However, because most of the case-patients with an unknown HIV status were <15 or >55 years of age, the age groups at lowest risk for HIV, these missing data likely did not mask an association if one exists.
HIV co-infection plays a role in the epidemiology of adult M. bovis TB, but likely the growing Hispanic population with close ties to Mexico, and not HIV, is the major driving force behind the increasing number of M. bovis cases we observed. Given the long latency of this disease, the unclear role of airborne transmission in M. bovis TB (13)    Our study confi rms earlier preliminary fi ndings (14) that M. bovis case-patients appear to be more than twice as likely to die before TB treatment completion compared with M. tuberculosis patients, despite being treated for the same mean number of days. The association of higher mortality rates during M. bovis treatment persisted after HIV, multisite disease, age, and ethnicity were accounted for. Causes of death related to noninfectious disease, such as malignancy and noninfectious gastrointestinal pathologies, were underrepresented in the M. bovis cases (0 vs. 25%), but, overall, the M. bovis and M. tuberculosis cases were not signifi cantly different with regard to all causes of death.
In mouse models, evidence indicates that certain strains of M. bovis are more virulent than M. tuberculosis strains (24), but those fi ndings are not generally supported in the literature on human M. bovis TB (7). Although our mortality analysis partially controlled for extent of disease, it did not include information on coexisting conditions, stage of HIV disease, diagnostic delays, and prior access to medical care. Therefore, M. bovis deaths might be accounted for by other factors, such as health disparities or treatment differences, which warrant further investigation.
Public health measures to control TB are currently focused on interrupting person-to-person transmission by promptly identifying and treating infectious patients and ensuring that they do not expose new contacts until treatment has rendered them noninfectious. Based on our data, these strategies, which have proven to be effective at reducing M. tuberculosis cases in San Diego and most regions of the United States, appear to be less effective in controlling M. bovis, suggesting that human-to-human transmission of M. bovis is less likely an important mode of transmission in this community. The consumption of contaminated dairy products has been proposed to be the primary source of human TB from M. bovis (25). This hypothesis is supported by the fi ndings of an investigation of M. bovis cases in New York that indicated the likely source of infection was unpasteurized cheese from Mexico (10). Additionally, San Diego pediatric M. bovis cases occur only after the age of weaning, when children are typically fi rst exposed to dairy products (6), and M. bovis was also recently cultured from unpasteurized cheese seized at the San Diego-Mexico border (26,27).
Because of the widespread adoption of pasteurization of all commercially available dairy products in the United States, as well as the aggressive US state agricultural health programs designed to keep dairy cattle free from M. bovis disease, the threat of M. bovis in US dairy products was largely eliminated in the mid-20th century (8). The San Diego-Tijuana binational region, however, shares one of the busiest border crossings in the United States with Baja, Mexico (28), where M. bovis is prevalent in cattle and consumption of unpasteurized dairy products is a common cultural practice (29)(30)(31)(32). Mexican dairy products, including the popular queso fresco (soft, unpasteurized cheese), may be brought into the United States for personal use and are sometimes distributed illegally (27). Given our fi nding that >90% of M. bovis cases in San Diego occurred in Hispanics, most of whom were born in Mexico, consumption of unpasteurized dairy products from Mexico is likely a major risk factor for M. bovis TB in San Diego. Collaboration with Mexico on prevention strategies, from education to regulation of the production of unpasteurized dairy products, and elimination of M. bovis from dairy cattle will be required in the long term to ensure that this mode of transmission is eliminated.

Limitations
A growing awareness of M. bovis as a cause of TB in San Diego since 1980 could have introduced a sampling bias into our trend estimates, but this possible bias is unlikely to have had a considerable effect in the years 1994 of all reported TB cases. Increased efforts to obtain specimens for culture in pediatric TB cases in the years under study did not appear to change the proportion of culturepositive cases during the study period.
The cohort of TB case-patients who were not culturepositive and thus excluded from this analysis was significantly different from the study group. The <15-year age group (36% vs. 4%, respectively) and Hispanic ethnicity (57% vs. 45%) were both overrepresented in the excluded cases. Given that these are the groups most likely to have M. bovis TB, the total incidence of M. bovis, particularly in children, may be underestimated in our study.

Conclusions
San Diego, California, while unique in many respects because of its close proximity to Mexico, is possibly representative of other communities in the United States with large and growing Hispanic populations with ties to Mexico. The considerable and growing incidence of TB from M. bovis, especially in children, and the observed number of deaths during treatment in these cases is of serious concern. It raises the question of the importance of incorporating routine species-level identifi cation into US TB surveillance as the national TB incidence shifts to persons born outside the United States. This surveillance will be greatly facilitated by the national genotyping project implemented by the US Centers for Disease Control and Prevention in 2004 (33), and its use will be particularly important for communities with strong ties to Mexico.