Epidemiology of Mycobacterium bovis Disease in Humans in England, Wales, and Northern Ireland, 2002–2014

Despite slightly increased cases in these areas, human infection with this cattle pathogen remains rare.

357 cases and observed increased annual case numbers (from 17 to 35) and rates. Most patients were >65 years of age and born in the United Kingdom. The median age of UK-born patients decreased over time. For 74% of patients, exposure to risk factors accounting for M. bovis acquisition, most frequently consumption of unpasteurized milk, was known. Despite the small increase in case numbers and reduction in patient age, M. bovis infection of humans in England, Wales, and Northern Ireland remains rare.
A fter the 1960s, the number of human cases of tuberculosis (TB) caused by Mycobacterium bovis decreased significantly in England, Wales, and Northern Ireland, coinciding with widespread implementation of milk product pasteurization and national bovine TB control programs (1)(2)(3). During the past 2 decades in these 3 countries, an average of 30 cases of M. bovis in humans occurred annually; numbers decreased in the early 2000s before again increasing (4)(5)(6). During the same period, incidence of M. bovis in cattle herds in parts of England, Wales, and Northern Ireland increased substantially but has now plateaued (4,(7)(8)(9).
M. bovis control (2,7,10,11) attracts political, public health, and media interest because of potential spread from animals to humans, effects on animal health and trade (1), and the role of wildlife in the transmission cycle (12). Highly visible interventions, including wildlife management to prevent transmission to livestock, are used to attempt to control M. bovis spread (6,9,10), thereby protecting human health.
Compared with other countries in western Europe, the rate of TB among humans in the United Kingdom is high: 9.6 cases/100,000 population (6,240 cases) in 2015 (13). Most TB cases occurred in those born abroad, who probably acquired infection before entering the United Kingdom. Although only 1.1% (42 cases) of culture-confirmed TB cases were caused by M. bovis (6), it remains a public health priority.
The drivers of the epidemiology of M. tuberculosis are well described (13)(14)(15). However, there is comparatively less information on the sources of M. bovis in humans, other than the recognized risks of unpasteurized milk consumption and close contact with infected cattle (1,3). We provide an update on the demographic characteristics of humans with M. bovis disease in England, Wales, and Northern Ireland (16). To address the gap in knowledge regarding lesser known sources of acquisition, we describe the demographic and clinical characteristics of humans with TB caused by M. bovis compared with M. tuberculosis. In addition, we describe potential human exposures that may indicate M. bovis acquisition and include a genotyping comparison of the causative organisms.

Data Collection
Results from culture-positive laboratory isolates were sent from Mycobacterium reference laboratories in England, Wales, and Northern Ireland to Public Health England. These results were matched with notified TB cases from ETS, used for statutory notification of TB, by use of a probabilistic matching method (17).
Data on demographics (age, sex, ethnicity, country of birth, time since UK entry, address, and occupation); clinical factors (site of disease and previous diagnosis); and social risk factors (current or past imprisonment, homelessness, drug and alcohol misuse) were obtained from ETS notifications. For nonnotified M. bovis cases, the only patient demographic information available was age, sex, and address; the disease site was inferred from specimen site. For analysis, we used the age groups 0-14, 15-44, 45-64, and >65 years and the ethnic groups white, black African, Indian subcontinent (Indian, Pakistani, and Bangladeshi grouped together), and other. After assignment to a geographic area of residence based on address, the place of residence was classified as rural or urban by using 2011 census classifications (18).
After identification of an M. bovis case (based on phenotypic, PCR, and genotypic methods [19,20]), a questionnaire (online Technical Appendix, https://wwwnc.cdc.gov/ EID/article/23/3/16-1408-Techapp1.pdf) (21) was issued to collect information on potential recognized current or past M. bovis exposures. These exposures were contact with a human TB patient, travel (for >2 weeks) to or residence in a country with high TB incidence (defined as having an estimated rate of >40 cases/100,000 population during 2002-2014), consumption of unpasteurized milk product, occupational contact with animals, physical contact with wild (nondomestic) animals, and physical contact with any animal with TB (including pets).

M. bovis Trend Analysis
We calculated incidence rates per 100,000 population by using mid-year population estimates produced by the UK Office for National Statistics (22). We used Poisson regression to calculate the incidence rate ratio to assess the trend in M. bovis incidence over time. We used a nonparametric test for trend across ordered groups to assess the age trend of M. bovis patients and the χ 2 test for trend to assess the proportion of M. bovis among culture-confirmed TB cases.

Factors Associated with M. bovis Disease and M. tuberculosis Disease
Demographic and clinical characteristics for M. bovis notified patients were compared with those of M. tuberculosis notified patients by using univariable and multivariable logistic regression to calculate odds ratios to identify factors associated with M. bovis disease. A forward stepwise multivariable logistic regression model was used, including sex and all variables with a p value <0.2 in univariable analysis; likelihood ratios were assessed after each stepwise addition to the model. In addition, we conducted a stratified analysis based on place of birth (UK-born/non-UK-born). A p value of <0.05 was considered statistically significant. We tested interactions between biologically and statistically plausible variables in the model by using likelihood ratios. All analyses were conducted by using Stata 13.1 (StataCorp LLC, College Station, TX, USA).

Exposures to Risk Factors Associated with M. bovis Disease
To identify frequent exposure to risk factors among the cohort, we used case exposure history, as collected through the questionnaire (online Technical Appendix), for descriptive analysis. In addition to obtaining questionnaire information about contact with another human TB patient, for culture-positive isolates identified during 2010-2014, we also obtained 24-loci mycobacterial interspersed repetitive unit-variable tandem repeat (MIRU-VNTR) strain typing results (20) from Mycobacterium reference laboratories. This information enabled us to identify strain type clusters, defined as >2 human TB cases with indistinguishable MI-RU-VNTR profiles (or with an indistinguishable profiles but with 1 case only typed to 23 loci), Clustered cases were further investigated to identify possible epidemiologic links, the identification of which suggest recent human-tohuman transmission (23). Among 297 M. bovis patients for whom place of birth was recorded, 214 (72.1%) were born in the United Kingdom. The most frequent countries of birth for the others were Nigeria (18 patients), Morocco (9 patients), and India (8 patients). The age distribution differed significantly between those born and not born in the United Kingdom (p<0.001) ( Table 1). The median age of UK-born patients fluctuated over time, from 71 years (interquartile range 60-76)

Comparison between Notified M. bovis and M. tuberculosis Patients
Univariable analysis showed that, when compared with M. tuberculosis notified patients, M. bovis notified patients were more likely to be >45 years of age, born in the United Kingdom, of an ethnic group other than that of the Indian subcontinent, live in a rural area, and work in agricultural or animal-related occupations. M. bovis patients were less likely than M. tuberculosis patients to have pulmonary disease. Multivariable analysis showed that the same factors, other than age, were independently associated with M. bovis; only those >65 years of age were more likely to have M. bovis disease. The strongest risk factor for M. bovis disease was working in an agricultural or animal-related occupation (adjusted odds ratio 29.5, 95% CI 16.9-51.6; Table 2). The model showed no interactions between explanatory variables. Analysis stratifying by place of birth (UK-born vs. non-UK-born) indicated that the same variables were significant.

M. bovis Patient Exposure to Risk Factors
Of the 272 M. bovis patients identified during 2006-2014, exposure questionnaires were completed for 241 (88.6%). Of these, 179 (74.3%) reported exposure to at least 1 risk factor for M. bovis acquisition; 78 (43.6%) reported 1 exposure, 57 (31.8%) 2 exposures, 28 (15.6%) 3 exposures, and 16 (8.9%) 4 exposures. For 6 patients, no exposure was known; for the remaining 56 patients, data were missing for >1 risk factor and the patients could not be classified as not having been exposed to a risk factor.
The most frequently reported exposure was consumption of unpasteurized milk products (65.7%, 109/166; Contact with a human TB patient was reported by 18.2% (33/181), but for most, recorded information was insufficient to identify the contact, particularly if the contact was not recent. Where known, 80.8% (21/26) of contacts occurred >5 years before TB diagnosis. From 24-loci MI-RU-VNTR strain typing data available during 2010-2014, a total of 48.7% (57/117) of patients (of which 46 were born in the United Kingdom and 9 were not) were in 15 M. bovis strain type clusters. One cluster contained exclusively patients not born in the United Kingdom and 7 exclusively born in the United Kingdom; 2 of the latter clusters contained the only epidemiologically linked human patients, each with a pair of household contacts.
Recent acquisition of infection cannot be directly measured, but the rate of M. bovis disease among children, along with their exposures, can provide an indirect indicator of recent acquisition. Exposure information was available for 5/6 M. bovis patients <15 years of age and suggested potential overseas acquisition; 5 had traveled to a country where TB incidence was high, 1 of whom had consumed unpasteurized milk while abroad. For 1 child not born in the United Kingdom, a questionnaire response was not obtained.
Overall, among those for whom location of exposure was known, 59.1% (97/164) of patients were exposed to >1 risk factor in the United Kingdom (Table 3). Among those not born in the United Kingdom, 18.0% (9/50) were known to have been exposed to a risk factor while in the United Kingdom, but 4 of the 9 also were exposed outside the United Kingdom.

Discussion
Our findings confirm that M. bovis disease remains rare among humans in England, Wales, and Northern Ireland. Over the study period, the annual rate of M. bovis disease and the proportion of culture-confirmed TB cases with M. bovis identified as the cause displayed a small but statistically significant increase; annual case numbers for the past 10 years were similar to those for the early 1990s (4). Although speciation has improved from the use of strain typing results (19,20), this improvement is unlikely to account for all of the increase identified. Although the previous study by Jalava et al. (4) and our study overlap by 2 years, our results benefit from improved matching (17) between case notification and culture results from 2002 on, thereby providing improved accuracy for reporting annual case numbers.
We identified, unlike previous studies (4,5), that although the number of M. bovis patients not born in the United Kingdom remained low and fluctuated over time, the annual number of cases in this group increased slightly over time. Our finding may be confounded by better recording of place of birth but is not unexpected given the increase during this period in the overall number of TB patients not born in the United Kingdom (13). Similar to previous findings (5), our findings indicate that most M. bovis patients not born in the United Kingdom lived in urban areas, specifically London. These patients originated mostly from low-income countries where TB incidence is high and therefore are at higher risk for human-to-human transmission and animal-to-human transmission because of limited detection of M. bovis in animals and less frequent milk pasteurization (24). Thus, infection was probably acquired before arrival in the United Kingdom and less likely to be related to exposure to risk factors while in the United Kingdom. Unfortunately, speciation is not routinely conducted in many high TB burden, low-income countries (16), so it is difficult to identify in which countries incidence of M. bovis is high and whether the trends in country of birth for M. bovis patients not born in the United Kingdom reflect the global incidence of the disease (24,25).
We identified a decrease over time in the proportion of UK-born patients >65 years of age and a decrease in the median age, although neither was statistically significant. Previously, most cases in UK-born patients were the result of reactivation of infection acquired before the large rollout of pasteurization (by the 1960s) (4), when M. bovis incidence was higher. Given the length of time that widespread pasteurization has been in place, progressively fewer cases among the older population are expected as the cohort exposed before pasteurization decreases. It was unexpected that, despite this decrease, the number of M. bovis cases occurring in the UK-born population did not reduce over the study period. Instead, the number and proportion of younger UK-born patients increased slightly. Numbers remain small, and it is not possible to yet detect any change in exposures; however, in recent years, the media have reported increased public demand for unpasteurized milk, which, if contaminated, could result in more human infections.
No data are available to quantify unpasteurized milk production or consumption within England, Wales, and Northern Ireland. However, results from a 2012 survey of adult consumer attitudes about unpasteurized milk (26) showed that 33% of respondents had consumed unpasteurized milk but only 3% currently consumed unpasteurized milk. Although the proportion who had ever consumed unpasteurized milk was highest among older age groups (18-24 years, 31%; 25-44 years, 28%; 45-64 years, 38%; >65 years, 40%), the proportion of current consumers was higher among younger age groups (18-24 years, 7%; 25-44 years, 4%; 45-64, 1%; >65, 1%). It is possible that increased consumption of unpasteurized milk, as reported by the media, is contributing to the small increase in M. bovis cases and may contribute to a change in demographics of patients over time. Although we do not have evidence to confirm, this hypothesis could be explored further through a formal observation study. The time between unpasteurized milk consumption and onset of TB disease among the M. bovis patients in our cohort emphasizes that the effects of current unpasteurized milk consumption may not be observed for many years.
The results of combining routine 24-loci MIRU-VN-TR typing of M. bovis from humans with epidemiologic data provide evidence of only occasional human-to-human M. bovis transmission; despite extensive follow-up of the 57 clustered cases, only 2 instances of 2 cases being epidemiologically linked were found. Only 1 prior occurrence of MIRU-VNTR-confirmed (using 15-loci typing) human-to-human transmission of M. bovis in the United Kingdom has been documented (5,27); it occurred before the rollout of routine prospective 24-loci MIRU-VNTR  typing. There are also few examples of human-to-human M. bovis transmission in countries other than those included in this study (28,29), suggesting that such transmission is rarely identified. Overall, the proportion of clustering among M. bovis cases (49%) was slightly lower than that of the overall proportion among all TB cases (56%) observed in England (13). This analysis also presents findings consistent with those previously reported (4). Although the proportion of cases among the older UK-born population seems to be decreasing, over the study period this group accounted for most cases. Our comparative analysis confirmed that the demographic profile of M. bovis patients differs from that of M. tuberculosis patients. The consumption of unpasteurized milk remained the most frequently reported exposure, and M. bovis patients were more likely than M. tuberculosis patients to work or have worked in agricultural and animal-related occupations. These findings are reassuring and show that M. bovis disease is still largely limited to those with recognized risk factors for infection. Few incidents involving animal-to-human transmission on farms (1,30,31) and a single incident of M. bovis transmission from a pet to its owners 32,33) have occurred during the study period. Most animal-to-human transmission remains sporadic, and implementation of additional specific interventions beyond those currently in place (1,2) would be difficult.
A high proportion of UK-born patients lived in rural areas, especially across the South West and Midlands of England, where M. bovis incidence among cattle is high (Figure 3, panel B). Most of these patients reported consumption of unpasteurized milk or contact with animals. However, human patients without such exposures and who reside in these areas where M. bovis cattle incidence is high should continue to be monitored and thoroughly investigated to ensure that lesser known exposures are not missed.
Similar to our study, a study in the Netherlands identified that the highest proportion of M. bovis cases occurred in the older native population (50%), followed by the foreign-born population (40%) (34). In comparison with our study, studies from the United States found that being foreign born (in particular, being of Hispanic ethnicity) and younger were independently associated with M. bovis when compared with M. tuberculosis (35,36). The difference in demographic characteristics of M. bovis patients in the United States and in England, Wales, and Northern Ireland may be explained by the fact that M. bovis in cattle or wildlife is not frequently reported in the United States (37,38) but is more common in neighboring Mexico (39,40). Thus, the epidemiology of human M. bovis in England, Wales, and Northern Ireland continues to be driven by the past and, to some extent, present prevalence of disease in cattle. Given advances in molecular techniques, improved understanding of animal-to-human transmission will require linking the genotyping results from animals with M. bovis infection in England, Wales, and Northern Ireland with data from humans. Globally, zoonotic TB should be tackled, and the needs of those affected by M. bovis disease, namely those in animal-related occupations and those consuming unpasteurized milk from infected animals, should be addressed. The implementation of methods to identify M. bovis where culture is not possible have been highlighted as essential (16,41,42). Although findings from England, Wales, and Northern Ireland cannot be extrapolated even to other high-income countries, much less to high TB burden, low-income countries, our study does illustrate the value of monitoring M. bovis disease and the data required to do so.
Our study does have some limitations. The exposure questionnaires return rate was 89%, and some responses were missing, which could lead to some error in the estimation of exposures; in addition, nonresponders were more likely to be urban dwellers. Our comparison of M. bovis and M. tuberculosis patients was limited because exposure questionnaire information was only collected for M. bovis patients; therefore, animal-related exposures, travel to countries with high TB incidence, and contact with human TB patients could not be included in the analysis. In addition, patients not born in the United Kingdom, most of whom belong to Indian subcontinent ethnic groups, are more likely missed in analysis because a higher proportion have exclusively extrapulmonary disease (43), for which culture confirmation is lower. Approximately 60% of TB cases in England, Wales, and Northern Ireland are culture confirmed; therefore, the estimated M. bovis incidence presented in this article is probably an underestimate. The proportion of TB cases culture confirmed over time has remained relatively stable (44), so underascertainment should not affect changes in the number or proportion of TB cases caused by M. bovis.
In conclusion, we found that M. bovis disease continues to account for a small number and low proportion of total TB cases in England, Wales, and Northern Ireland. The proportion of culture-confirmed TB cases caused by M. bovis has increased slightly, and the age of UK-born patients has decreased. The reasons are not fully understood, and trends should continue to be monitored. For most patients, exposure to risk factors for M. bovis acquisition (e.g., unpasteurized milk consumption, farm work, or contact with a human TB patient) were known. The current control measures in place to prevent animal-to-human spread seem to be effective; such spread occurs in a few isolated incidents and sporadic events. However, to increase understanding of M. bovis transmission in England, Wales, and Northern Ireland, we recommend strengthening collaboration between animal and human health, including linking genotyping results.