The study population included all patients with culture-confirmed TB reported to the California TB registry during 2003–2011 and who had initial drug susceptibility testing results for isoniazid, rifampin, and pyrazinamide. Patients were classified as having M. bovis disease if the initial drug susceptibility results indicated resistance to pyrazinamide and susceptibility to isoniazid and rifampin. Sociodemographic, clinical, and treatment outcome information for all case-patients was abstracted from the TB registry. TB case registry data were matched to the California HIV/AIDS registry to identify HIV co-infection status. This analysis was conducted as part of the California Department of Public Health’s mandate to routinely collect and analyze surveillance data for public health purposes. The CDC (Atlanta, GA, USA) determined that the project was not human subject research and did not require approval by an institutional review board.

To compare the differences between child and adult populations, we stratified M. bovis case-patients by their age at the time TB was reported. The child population consisted of patients <15 years of age, and the adult population consisted of patients >15 years of age. To evaluate characteristics associated with TB disease caused by M. bovis compared with TB disease caused by M. tuberculosis, we conducted bivariate analysis with variables that were added to the national TB surveillance system in 2010. These variables include birth country of parents/guardians (for child case-patients), primary reason evaluated for TB disease, diabetes mellitus, and other immunosuppressive conditions. Immunosuppressive conditions, excluding HIV co-infection and diabetes mellitus, comprised end-stage renal disease, anti–tumor necrosis factor-α therapy–associated immunosuppression, solid organ transplant–associated immunosuppression, or other immunosuppressive condition as indicated by medical records or a health care provider. In addition, patient variables previously shown to be associated with TB disease caused by M. bovis (i.e., country of birth, race/ethnicity, age, HIV co-infection, site of disease, and death before treatment completion) were included in the bivariate analysis (7,19). Sociodemographic and clinical variables shown to be significantly associated with M. bovis disease at the bivariate level were put into a logistic regression model. The final logistic regression model was constructed by using the backward stepwise elimination procedure, removing predictors with p>0.05.

For TB cases during 2004–2011, the sensitivity and positive predictive value of the pyrazinamide monoresistance case definition were calculated by using spoligotyping data as the reference standard. The signature spoligotypes differentiating M. bovis and M. tuberculosis have been described previously in detail (20). Spoligotyping was conducted at the California Microbial Diseases Laboratory (Richmond, CA, USA), and the resulting data were entered into the TB Genotyping Information Management System (http://www.cdc.gov/tb/programs/genotyping/tbgims/default.htm). The TB registry and TB Genotyping Information Management System databases were merged by patients’ unique case numbers after spoligotyping data were deduplicated. Spoligotyping results indicating a genotype other than M. bovis or M. tuberculosis (including M. bovis bacillus Calmette-Guérin (BCG) strains) and isolates missing spoligotyping data or initial drug susceptibility testing results for pyrazinamide, isoniazid, or rifampin were excluded from analysis

Analyses were conducted by using SAS version 9.3 (SAS Institute, Cary, NC, USA). Separate Poisson regression models were used to identify temporal trends in the incidence of TB caused by M. tuberculosis and M. bovis. Year of TB case report was used as the explanatory variable, TB case number as the dependent variable, and state population size as the offset variable. Population denominators were obtained from the US Census Bureau’s current population survey (21). Trends of the annual percentage of TB cases attributable to M. bovis were examined by using the Cochran-Armitage trend test. Epidemiologic differences between M. bovis and M. tuberculosis were compared by using the χ² test or Fisher exact test. Differences in median time to completion of therapy from start of therapy were analyzed by using the Wilcoxon rank-sum test. Trends and differences in disease characteristics were considered statistically significant if p<0.05 (2-sided).

In this large population-based epidemiologic study of M. bovis disease in California, we observed an increase in the annual percentage of TB cases attributable to M. bovis from 2003 through 2011. M. bovis disease accounted for nearly 25% of culture-positive TB cases in children. Patients with M. bovis disease were more likely than those with M. tuberculosis disease to die during treatment, and most deaths were among adults with concurrent immunosuppressive conditions. Hispanic ethnicity (compared with non-Hispanic white), extrapulmonary disease, and diabetes and other concurrent immunosuppressive conditions (excluding HIV co-infection) were independently associated with M. bovis disease compared with M. tuberculosis disease.

During 2010–2011, all (6/6) child case-patients with M. bovis disease had >1 parent/guardian who had been born in Mexico, and 4 of the 6 children had lived in Mexico for >2 months. These results, in conjunction with the a high percentage of extrapulmonary disease among the case-patients and the lack of M. bovis disease among children <12 months of age, are consistent with findings in previous epidemiologic studies that suggested that human M. bovis disease in the United States often results from consumption of unpasteurized dairy products originating from foreign countries, including Mexico (6,12,22,23). National pasteurization requirements, strict regulations on the importation of dairy cattle from Mexico into the United States, and an effective US bovine TB eradication program have substantially reduced M. bovis infection in US-born cattle (22,24). In the past century, the prevalence of M. bovis disease in US dairy herds has decreased from ≈5.0% to <0.001% (25). Furthermore, M. bovis isolates from humans in San Diego have been found to be genetically related to M. bovis strains from cattle in Mexico (22). Expansion of M. bovis disease surveillance and genotyping to include whole-genome sequencing may add discriminatory power beyond traditional genotyping and help to identify the source and route of transmission of M. bovis (26). Sharing of whole-genome sequencing data between countries and different health agencies may enhance future national and international prevention interventions.

Despite the relatively high percentage of extrapulmonary disease, 57% of M. bovis case-patients in our study had pulmonary involvement; this percentage is consistent with findings in other studies in the United States (7,8). Person-to-person transmission of M. bovis is considered infrequent, but the magnitude of such transmission has not been precisely quantified (27). Our findings are consistent with those from past research, which has shown that case-patients with M. bovis and M. tuberculosis pulmonary disease do not differ significantly in 2 of the key indicators of infectivity: presence of lung cavitation on chest radiographs and presence of acid-fast bacilli in sputum smears (6,7,28). In addition, previous research from pulmonary TB contact investigations showed that TB infection conversion rates among contacts did not differ significantly by mycobacterial species of the source case, suggesting that M. bovis is equally as transmissible as M. tuberculosis (28). However, current TB contact investigation guidelines do not include risk factors for M. bovis transmission. Although it is recommended to prioritize immunocompromised contacts of M. tuberculosis patients for evaluation, our data suggest that it may be even more important to prioritize immunocompromised contacts of M. bovis patients during contact investigations.

In our evaluation of the definition of pyrazinamide monoresistance, we found that isolates genotyped as M. tuberculosis may have been misclassified as M. bovis because of the pyrazinamide monoresistance definition. Misclassifications occurred most notably among the non-Hispanic Asian patients and might be explained by host, environmental (e.g., regional programmatic differences in TB treatment), and bacterial characteristics. From a recent national multivariate analysis of pyrazinamide resistance, Kurbatova et al. (15) suggested that bacterial lineage, not host characteristics, was the primary association between pyrazinamide monoresistance and M. tuberculosis disease.

Despite the possible definition-associated overestimation of M. bovis disease, the results from our study may still underestimate the true burden of M. bovis disease in California. Cases in children may be underestimated in the study population because sputum or gastric aspirate specimens are not consistently obtained from young children (29). In addition, M. bovis disease may be present in case-patients in the culture-negative subpopulation. We found case-patients with culture-negative TB to be similar to case-patients with culture-confirmed M. bovis disease with respect to age, extrapulmonary disease, and Hispanic ethnicity (data not shown).

Given the limitations of traditional genotyping and surveillance, we could not assess whether the increase in the number and percentage of adult M. bovis case-patients was to the result of recently acquired TB or reactivation of a previous infection. In addition, analysis of the new variables that were added to the national TB surveillance system in 2010 was hindered by the small number of child case-patients with M. bovis disease during 2010–2011. Also, because we did not conduct a medical chart review, we could not determine the cause of death among TB case-patients who died before the completion of TB treatment. Because of limitations in the national TB surveillance report form, we were unable to assess several possible relevant risk factors, including consumption of unpasteurized dairy products and the potential protective effect of the BCG vaccine against M. bovis disease. Although there is little information in the literature on the efficacy of the BCG vaccine in protecting against human M. bovis disease, the vaccine is notably protective against extrapulmonary disease and childhood TB disease, both of which are characteristic of M. bovis disease (30–33).

In summary, human M. bovis disease incidence has not declined in California, and the percentage of TB cases attributable to M. bovis has increased, exceeding the overall average for the United States. In California, there are ongoing interventions designed to limit the demand for and distribution of unpasteurized and contaminated dairy products, which are associated with M. bovis disease and other foodborne diseases (23). Elimination of human M. bovis disease in California likely requires further implementation of programs to reduce M. bovis contamination of dairy products in countries that have bovine TB, including pasteurization and test and cull interventions. In the interim, actions in the United States can help facilitate this effort. For example, M. bovis genotyping surveillance could be expanded and key risk elements (e.g., consumption of unpasteurized dairy products) could be captured in the national TB surveillance system to help determine M. bovis transmission routes. In addition, current TB contact investigation guidelines should be changed to include risk factors for M. bovis transmission. Last, future education efforts to prevent acquisition of M. bovis should focus on Hispanic binational families and adults with concurrent immunosuppressive conditions, including diabetes.