• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of biolettershomepageaboutsubmitalertseditorial board
Biol Lett. Mar 22, 2005; 1(1): 53–56.
Published online Mar 1, 2005. doi:  10.1098/rsbl.2004.0249
PMCID: PMC1629052

Herd-level risk factors associated with tuberculosis breakdowns among cattle herds in England before the 2001 foot‐and‐mouth disease epidemic

Abstract

A case–control study of the factors associated with the risk of a bovine tuberculosis (TB) breakdown in cattle herds was undertaken within the randomized badger culling trial (RBCT). TB breakdowns occurring prior to the 2001 foot-and-mouth disease epidemic in three RBCT triplets were eligible to be cases; controls were selected from the same RBCT area. Data from 151 case farms and 117 control farms were analysed using logistic regression. The strongest factors associated with an increased TB risk were movement of cattle onto the farm from markets or farm sales, operating a farm over multiple premises and the use of either covered yard or ‘other’ housing types. Spreading artificial fertilizers or farmyard manure on grazing land were both associated with decreased risk. These first case–control results from the RBCT will be followed by similar analyses as more data become available.

Keywords: bovine tuberculosis, case–control study, cattle movement, randomized badger culling trial, TB99

1. Introduction

Bovine tuberculosis (TB) is a major animal health problem in Great Britain. Over 4.5 million cattle tests were performed during 2003, resulting in the slaughter of over 23 000 cattle. In the State Veterinary Service (SVS) west region, one in seven herds experienced TB-related movement restrictions during the year, the majority owing to a TB ‘breakdown’—at least one member of a cattle herd failing the conventional TB skin test or showing evidence of TB lesions at slaughterhouse inspection (TB statistics, Defra: http://www.defra.gov.uk/animalh/tb/stats/stats_dec2003.htm). To identify factors associated with the risk of a TB breakdown, a case–control study (known as the TB99 study) was initiated in 1999 (http://www.defra.gov.uk/animalh/tb/trans/p3tb99.htm). The TB99 study was based on cattle farms enrolled in the randomized badger culling trial (RBCT; Bourne et al. 1999; Donnelly et al. 2003) and was intended to complement the findings of the RBCT.

Case–control studies in the Republic of Ireland (Griffin et al. 1993, 1996), Northern Ireland (Denny & Wilesmith 1999) and Michigan, USA (Kaneene et al. 2002) have provided useful information on herd-level predictors of Mycobacterium bovis infection (the causative agent of bovine TB) and have led to recommendations on practices expected to reduce TB risk. The present study is the first such investigation in southwest England.

This paper describes an analysis of the risk of a TB breakdown in three RBCT regions (known as triplets) active before the 2001 foot-and-mouth disease (FMD) epidemic. The FMD epidemic interrupted data collection and so provided a clearly defined subset of data for analysis.

2. Materials and methods

All TB herd breakdowns that occurred in triplets A, B and C of the RBCT (table 1), after the initial proactive badger culls (the starting point of the RBCT) and before 1 January 2001, were eligible cases. Insufficient reports were obtained from the four other active RBCT triplets before the FMD interruption to warrant inclusion. Each case herd agreeing to participate triggered the selection of up to three control herds (one contiguous and two non-contiguous) matched on RBCT area. Eligible control herds had not experienced TB-related restrictions in the 12 months prior to the case breakdown and were selected randomly to reflect the herd size distribution of breakdowns within the RBCT areas in 1998. Each herd was included in the dataset only once, with subsequent case or control reports being excluded. Both the control and case reports were excluded if a farm completed one of each. All questionnaires were completed during an interview by trained staff from the local Animal Health Office.

Table 1
Numbers of case and control reports available from each of the RBCT triplets for the various analyses.

Variables derived from the questionnaire covered a wide range of farm management, environmental and herd data and were screened using logistic regression, controlling for RBCT triplet and treatment (study design variables) and herd size to model the risk of a TB breakdown. (A full list of variables analysed is presented in the Electronic Appendix.) Non-categorical variables were ln transformed. All variables with a likelihood ratio test (LRT) p-value lesser than or equal to 0.15 compared with the forced covariates model (triplet, treatment and herd size) were made available for a final regression analysis. Models were constructed by backward elimination, considering the contribution of each variable to the variation in the data (by means of a LRT), the significance of the variable in the model and the stability of the remaining parameter estimates and standard errors when the variable was removed. Elimination of variables continued until a stable model was obtained. Interactions between all fitted and forced variables were examined, as were nonlinear terms for non-categorical variables.

The population attributable fraction (λpop: the proportion of the observed disease potentially prevented by removing exposure to each factor) was calculated to compare factors in the regression model (Levin 1953)

λpop=p(E+|D+)((ψmult1)/ψmult))

where p(E+|D+) is proportion of cases (D+) exposed (E+) to factor and ψmult, adjusted odds ratio (OR) for factor from multiple regression.

To calculate λpop, fitted non-categorical variables were converted to binary variables based on the median response. For protective factors, λpop was calculated using the inverse of ψmult and the proportion of cases not exposed to the factor (p(E|D+)=1p(E+|D+)).

The stability of the findings was examined by fitting alternative regression models. First, herd type (classified as dairy, beef or other) was included as an additional forced covariate. Second, cases were restricted to breakdowns confirmed by either culture of M. bovis or demonstration of visible lesions consistent with M. bovis.

3. Results

In total, 151 case reports and 117 control reports from unique farms were analysed; 111 of the case reports referred to confirmed TB breakdowns (table 1). Case reports were received for 87% of eligible breakdowns. Seven case reports and 46 control reports were discarded as repeat visits. Ten categorical variables (table 2) and one non-categorical variable (total number of farm premises; cases: n=151, mean=2.33, s.e.m.=0.15; controls: n=116, mean=1.74, s.e.m.=0.10) passed the univariable screening, and were used in further analyses.

Table 2
Numbers of case and control farms reporting exposure to categorical variables made available to multiple logistic regression modelling.

The effects of the forced covariates triplet and treatment reflected the relative numbers of cases and controls obtained from each triplet and treatment. There was no effect of herd size, as the controls were selected based on the expected distribution of case herd size.

Seven variables, in addition to the forced covariates, were included in the final regression model (table 3). Movement of cattle on to the farm from either markets or farm sales increased TB risk with odds ratios of 3.26 and 1.93, respectively. Two housing variables were associated with an increased risk of a breakdown, namely use of covered yard housing (OR=4.22) and use of ‘other’ housing types (OR=2.30). Operating herds over multiple premises also increased the TB risk (OR=1.79). Spreading of artificial fertilizers and farmyard manure on grazing land were both associated with a decreased TB risk (OR=0.21 and 0.42, respectively). None of the interaction terms was significant (using Bonferroni's correction for multiple tests) and neither were the nonlinear effects of herd size or number of premises. None of the factors was distinguished on the basis of λpop (range: 0.10 (use of covered yard housing) to 0.39 (use of ‘other’ housing types); table 3).

Table 3
Logistic regression model of the risk of a TB breakdown showing the prevalence of exposure to the various factors among cases and population attributable fractions.

Similar risk factors were identified and consistent effects were found in the alternative models (table 4). Firstly, the inclusion of herd type as a forced covariate did not alter the composition of the model or the parameter estimates. Secondly, restricting the case definition to include only confirmed TB breakdowns resulted in the substitution of the protective factor use of artificial fertilizer with the presence of ‘other’ soil types on the farm as a protective factor. There was no appreciable change in the OR estimates of the conserved factors in the model but the confidence limits were somewhat wider.

Table 4
Comparison of variations to the logistic regression model of the risk of a TB breakdown based on changes to the data being modelled and the forced covariates included.(Variables included in the final models are shown with 95% confidence limits of the ...

4. Discussion

A number of herd and farm management factors associated with TB risk have been identified in these analyses. Foremost were variables describing the movement of cattle on to the farm, namely sourcing cattle from markets or farm sales. Somewhat surprisingly, the risk did not increase with the number of cattle moved on to the farm or with moving specific cattle types, as observed by Griffin et al. (1993). Nonetheless, the consistent results for these factors suggest a significant risk of importing TB onto the farm. Griffin et al. (1996) did not show any risk associated with moving cattle onto the farm where a policy of pre-movement testing was in place, but instead found that cattle purchased since the last herd test were less likely to fail the subsequent test.

Operating the farm over multiple separated premises was associated with an increased TB risk beyond any effects of herd size and in the absence of an effect of total area farmed. Further study is required to determine whether this factor acts as a proxy for a combination of environmental or management variables that together increase risk. For example, more premises may include more contiguous herds or more badger setts, both identified as risks in other studies (Griffin et al. 1996; Martin et al. 1997; Denny & Wilesmith 1999).

Two housing-related factors were included in the various regression models, both associated with an increased risk of a TB breakdown: covered yards (open-walled housing for larger groups of cattle that may or may not be sub-divided) and ‘other’ housing types. The latter of these two was poorly defined and undoubtedly denotes a broad range of housing types, and so interpretation of its inclusion in the model is difficult. Kaneene et al. (2002) concluded that housing designed to minimize cattle–wildlife contacts (deer in this case) reduced TB risk. This was not demonstrated in the present study, although the importance of housing factors was highlighted. Better definition of housing types is needed, determining which cattle were kept in various housing types and for how long.

Fertilizer use on grazing land was found to be protective, irrespective of whether it was artificial fertilizer or manure. It is not clear why this is so, although there were few instances where farms did not use fertilizers on the grazing land. Griffin et al. (1993) observed that spreading slurry increased the risk of TB, although this variable did not appear in their final regression model.

Unlike the conclusions of Kaneene et al. (2002) and Griffin et al. (1993), the presence of a potential wildlife reservoir, measured here as active badger setts mapped to either the farm land or to within 1 km of the farm boundaries, made no appreciable contribution to the models. An alternative indicator of badger exposure, such as the distance to the nearest main sett or local badger density, may be more appropriate. The prevalence of TB in badgers assessed locally would be particularly relevant because prevalence can vary considerably between adjacent social groups (Rogers et al. 2003).

The design of the TB99 study was to include three controls per case, but this was never realized and there was therefore a potential for bias owing to the non-random loss of potential controls. This was investigated using farm data available from the SVS and RBCT databases by comparing those farms included as controls with those herds in the RBCT areas not included in the case–control study (see the Electronic Appendix). As expected from the selection procedure, control farms were larger than farms not included in the study in terms of herd size and farm area ( p<0.01). Levels of consent to the RBCT differed between control farms and non-participating farms ( p<0.01). There were no differences between the two groups with respect to herd type ( p=0.08), number of land parcels ( p>0.1) or the numbers of badger setts on the farm land ( p>0.1) or within 1 km ( p>0.1).

Possible causal interpretations of these conclusions must be treated cautiously, however, especially because of biases, potential unobserved confounders and the possibility that some explanatory variables are proxies for other, more biologically relevant, variables. While we have aimed to exclude alternative explanations as far as possible, ambiguities of interpretation apply to broad observational studies of this type.

Nonetheless, these analyses have identified factors associated with the risk of a TB breakdown in three RBCT triplets before the 2001 FMD epidemic. The results are broadly consistent with findings from studies elsewhere, although some differences were found, as would be expected from this complex disease system. Further investigations of factors will be possible as additional data become available.

Acknowledgments

The authors thank all of the farmers who participated in this study, the SVS staff who collected the data, the VLA who designed and wrote the database and entered, stored and provided the data, and the ISG Secretariat for support throughout the study. This work was funded by Defra.

Supplementary Material

Herd-level risk factors associated with TB breakdowns prior to the 2001 UK Foot and Mouth Disease epidemic among cattle herds in England:

References

  • Bourne F.J, Donnelly C.A, Cox D.R, Gettinby G, McInerney J.P, Morrison W.I, Woodroffe R. MAFF Publications; London: 1999. An epidemiological investigation into bovine tuberculosis: second report of the independent scientific group on cattle TB. See http:// www.defra.gov.uk/animalh/tb/isg/report/contents.htm.
  • Denny G.O, Wilesmith J.W. Bovine tuberculosis in Northern Ireland: a case-control study of herd risk factors. Vet. Rec. 1999;144:305–310. [PubMed]
  • Donnelly C.A, Woodroffe R, Cox D.R, Bourne J, Gettinby G, Le Fevre A.M, McInerney J.P, Morrison W.I. Impact of localized badger culling on tuberculosis incidence in British cattle. Nature. 2003;426:834–837. [PubMed]
  • Griffin J.M, Hahesy T, Lynch K, Salman M.D, McCarthy J, Hurley T. The association of cattle husbandry practices, environmental factors and farmer characteristics with the occurrence of chronic bovine tuberculosis in dairy herds in the Republic of Ireland. Prev. Vet. Med. 1993;17:145–160.
  • Griffin J.M, Martin S.W, Thorburn M.A, Eves J.A, Hammond R.F. A case-control study on the association of selected risk factors with the occurrence of bovine tuberculosis in the Republic of Ireland. Prev. Vet. Med. 1996;27:217–229.
  • Kaneene J.B, Bruning-Fann C.S, Granger L.M, Miller R.A, Porter-Spalding B.A. Environmental and farm management factors associated with tuberculosis on cattle farms in northeastern Michigan. J. Am. Vet. Med. Assoc. 2002;221:837–842. [PubMed]
  • Levin M.L. The occurrence of lung cancer in man. Acta Unio Int. Cancer. 1953;9:531–541. [PubMed]
  • Martin S.W, Eves J.A, Dolan L.A, Hammond R.F, Griffin J.M, Collins J.D, Shoukri M.M. The association between the bovine tuberculosis status of herds in the East Offaly Project Area, and the distance to badger setts, 1988–1993. Prev. Vet. Med. 1997;31:113–125. [PubMed]
  • Rogers L.M, Forrester G.J, Wilson G.J, Yarnell R.W, Cheeseman C.L. The role of setts in badger (Meles meles) group size, breeding success and status of TB (Mycobacterium bovis) J. Zool. 2003;260:209–215.

Articles from Biology Letters are provided here courtesy of The Royal Society
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...