Longer-term risk of Mycobacterium bovis in Irish cattle following an inconclusive diagnosis to the single intradermal comparative tuberculin test

Preventive Veterinary Medicine 100 (2011) 147–154

Contents lists available at ScienceDirect

Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed

Longer-term risk of Mycobacterium bovis in Irish cattle following an inconclusive diagnosis to the single intradermal comparative tuberculin test T.A. Clegg a,∗ , M. Good b , A. Duignan b , R. Doyle b , M. Blake b , S.J. More a a Centre for Veterinary Epidemiology and Risk Analysis, UCD School of Agriculture, Food Science and Veterinary Medicine, University College Dublin, Belfield, Dublin 4, Ireland b Department of Agriculture, Fisheries and Food, Agriculture House, Kildare St., Dublin 2, Ireland

a r t i c l e

i n f o

Article history: Received 12 October 2010 Received in revised form 24 February 2011 Accepted 25 February 2011 Keywords: Mycobacterium bovis Tuberculosis Inconclusive Ireland Parametric survival analysis

a b s t r a c t In Ireland, new bovine tuberculosis (bTB) cases are detected using both field and abattoir surveillance. During field surveillance, an animal may be deemed a ‘standard inconclusive reactor’ (SIR) to the single intradermal comparative tuberculin test (SICTT) if the bovine response is >2 mm, and from 1 to 4 mm greater than the avian response. Little is known about the future infection risk posed by SIR animals that pass a subsequent retest, socalled ‘transient SIR’ (TIR) animals. The objective of this study was to critically evaluate the future bTB status of TIR animals, by examining the future risk of bTB diagnosis over the 4 years following initial SIR diagnosis and clearance at the subsequent retest. The study included all TIRs that were identified as SIRs in 2005 in otherwise free herds at tests with no other reactors at that test and that were clear at the subsequent retest. The analysis was restricted to cows that were neither sold, other than direct to slaughter, nor exported from the herd during the follow up period (to the end of 2009). Five control cows were randomly selected from each study herd. A parametric survival model with shared frailties, to account for clustering within herds, was developed to model time from passing a retest to future bTB diagnosis. The final parametric survival model contained the variables: TIR status in 2005, inconclusive status during the follow-up period, location, herd restricted during the study, time since last restriction within the herd and age. The time ratio for the TIR status variable was significant (p < 0.001) indicating that on average the time to diagnosis with bTB for TIRs was 78% shorter compared to the non-TIRs. The frailty term was significant (p < 0.001) indicating that animals within some herds were more likely to become reactors compared to other herds. These results have important implications for national policy and future management of TIR animals. Further, private veterinary practitioners and their clients should be aware of the increased risk associated with TIRs. © 2011 Elsevier B.V. All rights reserved.

1. Introduction In Ireland, new cases of bovine tuberculosis (bTB, caused by infection with Mycobacterium bovis) are detected using

∗ Corresponding author. Tel.: +353 1 716 6142; fax: +353 1 716 6147. E-mail address: [email protected] (T.A. Clegg). 0167-5877/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.prevetmed.2011.02.015

both field and abattoir surveillance (More and Good, 2006). Field surveillance consists of the annual testing of cattle using the single intradermal comparative tuberculin test (SICTT). In the SICTT, bovine and avian tuberculins are used in combination to assess, measure and compare the skin response at 72 [±4] h following intradermal injection (Monaghan et al., 1994). Using the standard interpretation of the SICTT, as described in European Council Directive

148

T.A. Clegg et al. / Preventive Veterinary Medicine 100 (2011) 147–154

64/432/EEC, the test result is classified as either positive (the animal is deemed a ‘standard reactor’), inconclusive (‘standard inconclusive reactor’, essentially a suspect) or negative. An animal may be deemed a ‘standard inconclusive reactor’ (SIR) to the SICTT if the bovine response is >2 mm, and from 1 to 4 mm greater than the avian response. In accordance with Directive 64/432/EEC, all herds with one or more SIR identified will have their status suspended (trade other than animals to slaughter is not allowed), and are designated as either ‘high’ or ‘low’ risk based on herd and area history, the contiguity profile of the herd and the herd profile of the herd of origin of the SIR(s) if relevant. Low risk herds may be allowed to trade within Ireland (exports are not allowed) if granted permission by the veterinary inspector (VI). The herd owner then has three choices for the management of the SIR: • to have the animal retested after a minimum period of 42 days i.e. an inconclusive reactor retest (option 1), • to slaughter the SIR and, provided the animal has no visible lesions, have a full herd test 42 days after the SIR leaves the herd (option 2), or • to slaughter the SIR and have the lymph nodes cultured for M. bovis (option 3). Little is currently known about the M. bovis infection risk posed by either SIR or transient SIR (TIR) animals, the latter being SIR animals that were negative at the subsequent retest. This is of scientific interest, but also has important policy and resource implications. Evidence of increased risk among TIRs of future bTB diagnosis will influence changes in policy with respect to TIRs. In a preliminary (unpublished) analysis conducted in county Sligo, an increased infection risk associated with TIR animals was suspected, based on the test history of a large number of animals (from approximately 100,000 animals tested in that county in 2006) during subsequent field and abattoir surveillance. Martin et al. (2002) found that the risk for an animal of being deemed a reactor was greater at the retest of an SIR animal (91.9 reactors per 1000 animals tested) compared to the annual test (1.5 animals per 1000 animals tested). Based on our understanding of the SICTT, and of diagnostic tests more generally, a TIR could be a noninfected animal returning a suspect result, often following exposure to environmental mycobacteria or infection with other mycobacteria, e.g. M. avium subsp. paratuberculosis (De la Rua-Domenech et al., 2006; Álvarez et al., 2009). Conversely, a TIR could be a bTB infected animal returning a suspect, rather than a positive result, due to a broad range of factors that relate to the animal, such as co-infection with or exposure to other mycobacteria, the tuberculin and/or the method of administration (De la Rua-Domenech et al., 2006). The SICTT is a subjective test, and issues such as the tester’s level of risk aversion or tolerance, contribute to problems with test interpretation (Monaghan et al., 1994). There are many estimates of the sensitivity and specificity of the SICTT (De la Rua-Domenech et al., 2006); under Irish conditions, a recent study estimated sensitivity and specificity as 56.8–67.5% and 99.5%, respectively (Clegg et al., 2011).

The objective of this study was to critically evaluate the future bTB status of TIR animals, by examining the future risk of bTB diagnosis over the 4 years following clearance at the retest subsequent to initial SIR diagnosis. 2. Materials and methods 2.1. The study animals A SIR was defined as any animal at the SICTT with a bovine response that was >2 mm, and from 1 to 4 mm greater than the avian response. The case animals, TIRs, were identified as follows: we first identified all SIRs that were disclosed at a full-herd test in 2005 in Ireland in otherwise bTB-free herds with no other standard reactors at that test. The study was then restricted to TIRs that tested negative at the subsequent retest, to those that had not moved from the disclosing herd (either sold or exported), other than direct to slaughter, during ‘the study period’ (from the date of the subsequent retest through to the end of 2009), and to cows (female animals from either dairy or suckler herds). We randomly selected five control animals that were present in the herd at the time that the TIR was detected. These were cows present in the herd and test negative at the time that the TIR was disclosed and likewise that did not move from the herd, except direct to slaughter, during the study period. Animals were followed from the time of the retest until the end of 2009. The outcome was whether the animal was subsequently diagnosed with bTB (either as a SICTT reactor or based on gross inspection at slaughter). Animals were censored at slaughter or at the end of the study, whichever was the earliest. 2.2. Data sources Three national databases were used: the Animal Health Computer System (AHCS), with tuberculin testing data of herds (all results) and animals (at test dates when inconclusive reactor or reactor animals were disclosed) since 1989; the Cattle Movement and Monitoring System (CMMS), recording calf registrations, cattle movements (farm-tofarm, via a market and to the abattoir) and on-farm deaths in Ireland since 2000; and a database of laboratory testing results (available since 2000) from the national abattoir surveillance programme (histopathology and culture results from all animals with a suspect lesion(s) detected at slaughter). 2.3. Data analyses 2.3.1. Univariable analysis The proportion of animals that were positive for bTB at the end of the study period were compared for each of the categorical risk factors (listed below) using a chi-square test or Fisher’s exact test when the sample size was too small. Similarly, the median values of each of the continuous variables were compared by bTB status at the end of the study using a Wilcoxon rank-sum test. Survival time to the end of the study was compared by TIR status at the start of the study using a log-rank test and other tests of equal-

T.A. Clegg et al. / Preventive Veterinary Medicine 100 (2011) 147–154

ity of the survival functions, namely wilcoxon, tarone-ware and peto-peto (which do not assume proportionality of the hazard functions). The proportion of animals that were slaughtered prior to the end of the study period was compared by TIR status using a chi-square test or Fisher’s exact test when the sample size was too small. Animals that were not slaughtered were censored at the end of 2009. Time to slaughter was compared by TIR status using a log-rank test and other tests of equality described above.

2.3.2. Multivariable analysis A survival model was developed to model the time to bTB diagnosis. The following independent variables were considered: • TIR status at the start of the study (STD INC): (TIR = 1 [case animals]; non-TIR = 0 [control animals]); • SIR diagnosis during the study period (FUTURE IC) [a timevarying predictor]: for all animals, 0 at the start of the study and increasing in increments of 1 whenever an animal was deemed a SIR; • TIR diagnosis prior to the study period (PRE IC): number of times the animal was deemed a TIR within 5 years prior to the start of the study; • Time spent in index herd prior to the SIR test (YRSINHERD); • Animal had been bought into the herd (BOUGHTIN): Animal was bought into the herd (BOUGHTIN = 1) since the beginning of 2000; • Age (in years) at the retest (time of entry to the study) (AGE): ranged from 1 to 9 with all animals aged ≥9 coded as 9, age was centred by subtracting the mean age. For animals born prior to 1996 (1822 animals) the date of birth was not recorded, therefore all animals with a missing birth date were assigned to the oldest age category (≥9 years); • Herd type (TYPE): dairy or suckler; • Previous history for brought in animals (MOVE TIMELAST): time (in days) since the animal had previously been in a bTB restricted herd; • Herd size (HSIZE): size of the herd at study start; • Location: District Electoral Division (DED; 3444 in Ireland): the herd incidence rate (ir; proportion of herds restricted) within each DED in the previous year (DEDir2004) or the number of reactors detected per 1000 animal tests (APT) within each DED either during the previous year or the previous 5 years (DEDapt1yr, DEDapt5yr) or county (26 counties): the herd incidence rate within each county in the previous year (countyir2004) or the APT within each county either during the previous year or the previous 5 years (countyapt1yr, countyapt5yr); • Previous history of the herd (TIME LASTRES): number of years since last bTB outbreak (restriction) in the herd. For herds not previously restricted, the time since last restriction was set to 20 years; • Month (MONTH): the month of the initial SIR test. Month was also categorised into seasons and tested;

149

• Herd restricted during the study period (HERD RES) [a timevarying predictor]: restricted = 1, unrestricted = 0. A parametric survival model was developed using STREG in STATA version 11 (StataCorp LP, College Station, Texas, USA). In order to account for clustering within herds, a survival model using a shared frailty effect was developed. Only parametric survival models were considered, due to the excessive time (>12 h) needed to run a Cox proportional hazards model with a shared frailty effect. Parametric models have the advantage that they use the time data as a continuous variable and do not lose any information, unlike proportional hazard models that only use the rank occurrence of events. In order to determine the correct distribution for the baseline hazard five different distributions were compared: exponential, weibull, log-logistic, log-normal and gompertz. The appropriate parametric distribution was determined by comparing the log likelihood and AIC from a model which included just the STD INC variable as a single predictor. In addition, the shape of the baseline hazard was examined by building a piecewise constant exponential regression model (Dohoo et al., 2009). Interpretation of shared frailty parametric models is easier when an accelerated failure time (AFT) metric is used. In this metric the effect of the independent variables were measured using time ratios (TR) which were interpreted as the relative effect on the mean time to event. A TR < 1 was interpreted as a shorter mean time to event. For frailty models measured using an AFT metric the TR did not change over time for subjects with different frailties (Meadows et al., 2007). A backward selection procedure was used to eliminate terms from the model based on a likelihood ratio test (p > 0.05). Continuous variables were also categorised into 4 groups based on the corresponding quartiles. Whether to treat variables as continuous, categorical or to transform the variable was tested by comparing univariate models using the AIC. To examine the appropriate functional form of a variable a plot of the lowess smooth of martingale residuals against transformations of the covariate were used. Additionally the choice of which location variable was most appropriate was determined by comparing each of the location variables (DEDapt1yr, DEDapt5yr, DEDir2004, countyapt1yr, countyapt5yr, countyir2004) using the AIC from the univariate models. Similarly the choice of animal/herd history variables (MOVE TIMELAST and TIME LASTRES, respectively) and a combination of these variables (the time since the most recent breakdown either in the current herd or a herd an animal was previously in) was determined using the AIC from the univariate models. Likewise the choice between time in herd (YRSINHERD), age and a combination of these two variables (the time in the herd for bought-in animals and age for homebred animals) was determined in the same way. An interaction between STD INC and FUTURE IC was also considered. The full multivariable model contained 12 variables which included the selected location variable; the selected animal/herd history variable; the appropriate time in the herd variable; the interaction between STD INC and FUTURE IC and the other risk factors listed above.

150

T.A. Clegg et al. / Preventive Veterinary Medicine 100 (2011) 147–154

Table 1 The univariable association between bTB infection status at the end of the study period (31 December 2009) and categorical independent variables. Variable

Class

STD INC

0 1 0 1 2 0 1 2 0 1 0 1 1 2 3 4 5 6 7 8 9 10 11 12

FUTURE IC

PRE SI

HERD RES BOUGHTIN MONTH

No. of animals 11,028 2216 12,986 251 7 12,947 290 7 10,428 2816 9932 3312 311 356 494 1499 1565 1577 1964 1792 1725 1082 551 328

No. of bTB positivea 285 207 451 40 1 477 14 1 386 106 373 119 7 25 23 54 48 56 96 57 50 31 18 27

% bTB positivea

p-Value (chi-square test)

2.58 9.34 3.47 15.94 14.29 3.68 4.83 14.29 3.70 3.76 3.76 3.59 2.25 7.02 4.66 3.60 3.07 3.55 4.89 3.18 2.90 2.87 3.27 8.23

<0.001 <0.001

0.014b

0.876 0.668 <0.001

a The number and proportion of animals that become reactors to the SICTT or have tuberculous lesions at slaughter prior to the end of the study (31 December 2009). b Due to small numbers the Fishers exact test was used.

3. Results A total of 2216 case and 11,028 control animals from 1653 herds were enrolled. By the end of the study period, 207 (9.3%) and 285 (2.6%) case and control animals, respectively, were diagnosed with bTB. At post-mortem examination during routine conditions of slaughterhouse surveillance, the proportion of TIRs and non-TIR animals with gross evidence of bTB was 33.5% and 25.4%, respectively (difference not significant, p = 0.067). Univariable analysis found that bTB disease status varied significantly by STD INC, FUTURE IC, PRE IC, MONTH, and DEDapt1yr (Tables 1 and 2) and was borderline significantly different by YRSINHERD. The Kaplan–Meier probabilities of surviving to the end of 2009 without a bTB diagnosis for case and control animals are presented in Table 3. The log-rank tests and other tests of equality of the survival functions, namely wilcoxon, tarone-ware and peto-peto were all significant (p < 0.001), suggesting that the survival estimates for the TIRs and non-TIRs were significantly different. At the end of the study period, significantly (p < 0.001) more TIRs had been slaughtered (51.4%) compared to non-TIRs (45.8%). Time to slaughter (Fig. 1) was significantly shorter for TIRs compared to non-TIRs (p < 0.001, for wilcoxon, tarone-ware and peto-peto tests). The final parametric survival model of the time to bTB diagnosis (Table 4) contained the variables: TIR status (STD INC), future inconclusive status (FUTURE IC), APT within the DED in the previous year (DEDapt1yr; as a categorical variable), herd restricted during the study period (HERD RES), previous history of the herd (TIME LASTRES; as a categorical variable) and age (AGE). The interaction

between STD INC and FUTURE IC was tested within the model but found to be not significant (p = 0.153). According to the AIC, a log-normal baseline distribution gave the best fitting model. This distribution also appears to be appropriate based on the baseline hazard generated from a piecewise constant exponential model (Fig. 2). The time ratio (TR) for the STD INC variable was significant (p < 0.001; Table 4); on average, the time to bTB diagnosis was 78% shorter for case compared to control animals. Similarly the TR for the FUTURE IC variable was significant (p < 0.001; Table 4); time to bTB diagnosis was 94% faster in those animals that were deemed SIR animals during the study. The frailty term was significant (p < 0.001) indicating that both TIR and non-TIR animals within some herds were more likely to become reactors compared to other herds. A model with an inverse gaussian shared frailty was slightly better than one with a gamma shared frailty according to the log likelihood (−1945 versus −1941). The increase in the predicted hazard function near the start of the study was more pronounced for the TIRs compared to the non-TIRs, the hazard then decreased for the TIRs after around 500 days (Fig. 3). 4. Discussion 4.1. Key finding Animals that were deemed SICTT inconclusive reactors in 2005 and passed a subsequent retest (TIRs) were significantly more likely to be diagnosed at a later date with bTB (during subsequent testing; either field or abattoir surveillance) than control animals. The proportion of TIRs diagnosed with bTB during the study was almost four

T.A. Clegg et al. / Preventive Veterinary Medicine 100 (2011) 147–154

151

Table 2 The univariable association between bTB infection status at the end of the study (31 December 2009) and continuous independent variables. Variable

bTb status

Median

Q25

Q75

Age (years)

Negative Positive Total Negative Positive Total Negative Positive Total Negative Positive Total Negative Positive Total Negative Positive Total

5 5

3 3

7 7

2.88 2.63

1.54 1.58

4.01 3.69

1198 1237

816 878

1562 1524

1575 1698

716 545

2807 3654

103 115

61 65

154 164

1.43 2.04

0.40 0.95

3.13 3.81

Yrsinherd (years) (brought in animals only)

Mover timelast (days) (brought in animals only)

Time lastres (days) (herds previously restricted only)

HSIZE (no. of animals)

DEDapt1yr (reactors per 1000 tests)

Number of animals 12,752 492 13,244 3193 119 3312 463 17 480 9384 384 9768 12,752 492 13,244 12,752 492 13,244

p-Value (wilcoxon) 0.069

0.053

0.786

0.725

0.075

<0.001

Table 3 Kaplan–Meier probability of surviving for 4 years without a bTB diagnosis, by TIR status. Standard inconclusive reactor (TIR) status Control animals (non-TIR) Case animals (TIR) a

Number at risk 11,028 2216

Survival probability

Std. error

p-Valuea

0.963 0.873

0.002 0.009

<0.001

Log-rank test of equality of survival functions.

times higher compared with non-TIRs (9.34% versus 2.58%). Indeed, during the study period the time to bTB detection was 78% shorter among TIRs compared to non-TIR animals. Although the risk of a TIR being diagnosed with bTB decreased after around 500 days (Fig. 3), the hazard was always greater than that of non-TIR animals throughout the study period. Therefore, closer follow up and testing of these animals would lead to more rapid removal of those that were truly infected. In recent years, there has been increasing recognition of the importance of local bTB persistence in Ireland, as reflected in the strong association between previous history of bTB in cattle herds and the risk of future M. bovis infection (Griffin et al., 2005; Olea-Popelka et al., 2008), and spatial clustering of M. bovis infected cattle herds (Kelly and More, 2011). Latent infection is well-recognised in human infection with M. tuberculosis (Stewart et al., 2003), and

Fig. 1. Kaplan–Meier survival estimates of time to slaughter for case (TIR) and control (non-TIR) animals.

residual infection is increasingly recognised as an important feature of bovine infection with M. bovis in Ireland. There is increasing understanding of reasons for local bTB persistence, which include residual infection in cattle and an infected wildlife reservoir. Increased bTB risk among TIR animals is largely an indication of residual infection (ongoing infection in animals, but with negative or equivocal test results). Residual infection is the most likely explanation for: the presence of a single infected animal in an attested herd detected during abattoir surveillance (Olea-Popelka et al., 2008); for increased bTB risk following purchase of animals with past bTB exposure (Wolfe et al., 2009) and for ongoing spatial clustering in areas following badger removal (Kelly and More, 2011).

Fig. 2. Baseline hazard from a piecewise constant exponential model. The baseline hazard is forced to remain constant across four time intervals and allowed to vary between time intervals.

152

T.A. Clegg et al. / Preventive Veterinary Medicine 100 (2011) 147–154

increase in the sensitivity of the SICTT (Clegg et al., 2011; De la Rua-Domenech et al., 2006). This does not invalidate the observed association between TIR and bTB status, which was observed after controlling for the effect of HERD RES. The study was restricted to dairy and suckler cows, with 75% being older than 3.6 years. Time to a positive bTB diagnosis was 6% faster for each additional year of age. Age has previously been identified as a risk factor for bTB (Clegg et al., 2008) with a similar increase in risk of bTB found as age increased.

4.3. Key points of caution

Fig. 3. Predicted hazard functions from the final log-normal survival model for case (TIR) and control (non-TIR) animals.

4.2. Significant variables within the final model A range of independent variables have been included, and controlled for, in the final model. Each of these variables is plausibly associated with future infection risk, and many have been described previously. Infection history in a location (DEDapt1yr) is long recognised as an important risk factor for bTB risk (Griffin et al., 1996; O’Sullivan and O’Keeffe, 1998; Olea-Popelka et al., 2004; Clegg et al., 2008), potentially as a consequence of long-standing infection in cattle (residually infected cattle, local cattle-to-cattle transmission) (Green et al., 2008; Kelly and More, 2011) or wildlife (Kelly and More, 2011). The inconclusive reactor status during the study (FUTURE IC) was found to be a significant risk factor (p < 0.001) with the time to a positive bTB diagnosis occurring 94% faster in those animals which is consistent with the increased future bTB risk of SIR animals. The inclusion of HERD RES, reflecting herd restriction(s) during the study period (by definition, preceding a positive bTB result in the study animals), was a reflection of an ongoing bTB herd problem. In such situations, severe interpretation will have been applied, leading to a marked

The current study was restricted to a well-characterised subset of all SIR animals detected annually in Ireland. By necessity, the study focuses on TIRs in otherwise bTB-free herds. In bTB-infected herds (such as a herd with 2 or more standard SICTT), animals were tested at a minimum of 60day intervals until the whole herd was negative on two consecutive occasions, a ‘severe interpretation’ of the SICTT was applied to these herds and SIRs were deemed reactors and removed from the herd. For simplicity of analysis, the study was restricted to dairy and suckler cows and to animals that remained in the herd following the retest. Since cows tend to live longer than other animal classes, they are the animals that would be at greatest risk of M. bovis infection. Consequently, any future changes to testing policy regarding inconclusive reactor status would have the greatest impact on cows. The study results appear robust despite some evidence of risk adverse behaviour towards TIR animals by both farmers and testing veterinarians. Despite the potential for differential misclassification, we have good reason to believe that the study results are robust with any bias leading to an apparent reduction of the true measure of effect. We present three detailed illustrations to highlight this point. Firstly, time to slaughter was significantly shorter for TIR compared to non-TIR animals (Fig. 1), consistent with concern among farmers about the future viability of these

Table 4 Parametric log-normal survival model for time to bTB detection. Variable

TIR status at the start of the study (STD INC) Inconclusive diagnosis during the study (FUTURE IC) APT in the DED in the previous year (DEDapt1yr)

Herd restricted during the study period (HERD RES) Days since last restriction in herd (TIME LASTRES)

Age (AGE) Variance estimates a b c d

Wald test. Likelihood ratio test. Standard deviation of the log-normal distribution. Variance of the unobserved frailty parameter.

Class

1: <0.5 2: 0.5–1.6 3: 1.6–3.2 4: >3.2 1: <366 2: 366–730 3: 731–1095 4: >1095 Sigmac Thetad

95% CI p-Valuea

p-Valueb

0.27 0.13

<0.001 <0.001

<0.001 <0.001 0.043

0.50 0.41 0.45 0.23

1.06 0.86 0.97 0.44

0.101 0.006 0.035 <0.001

1.21 0.95 1.28 0.90 1.69 5.64

4.29 3.45 3.21 0.98 1.99 11.26

0.011 0.069 0.003 0.002

Time ratio

Lower

0.22 0.06 Referent 0.73 0.59 0.66 0.32 Referent 2.28 1.82 2.02 0.94 1.83 7.97

0.18 0.03

Upper

<0.001

<0.001 0.015

0.002

T.A. Clegg et al. / Preventive Veterinary Medicine 100 (2011) 147–154

animals. Consequently, TIRs were more likely to be subject to abattoir than field surveillance compared to non-TIR animals. However, the sensitivity of abattoir surveillance is known to be much lower than field surveillance (De la RuaDomenech et al., 2006), leading to an underestimate of both the number of TIRs (compared with non-TIRs) infected and the difference between future bTB risk in TIRs and nonTIRs. Secondly, among TIRs that were not slaughtered, it is feasible that these animals were at increased risk of a false positive result during field surveillance, as a consequence of concern among farmers and testing veterinarians about the future viability of these animals. If this is true, however, we would expect a higher rate of false positives among TIRs than non TIR animals, as reflected in the lesion rate at post-mortem. However, this was not the case with the lesion rate among TIRs (33.5%), being higher (but not significantly so, p = 0.067) than non-TIR animals (25.4%). Finally, the study was restricted to animals that remained in the herd throughout the study period. In accompanying work (Clegg et al., in preparation), the risk of selling an animal following a retest was significantly higher for TIRs compared to their herdmates (after accounting for differences in herd type, animal class and month of retest), reflecting concern among farmers of the future viability of these animals. Farmers may be selling those TIRs they perceive to be of greater risk such as TIRs with previous inconclusive readings, with larger bovine reactions or older animals. If this is the case, the TIRs under study were at lower risk, further contributing to an underestimate of the difference in future bTB risk between TIR and non-TIR animals. 4.4. Methodological issues The baseline hazard is measured when all variables in the model are zero. As seen in Fig. 2, the hazard is initially low, increases rapidly and then decreases slowly over time. This pattern is consistent with the baseline hazard for a log-normal model. The pattern is also biologically plausible. There will be an initial time lag before animals are tested again or slaughtered following the retest, which would explain the initial low hazard. The initial increase would reflect a residual infection risk that remains in the herd following the retest (assuming that some/the majority of inconclusive reactors were truly infected). This risk would decrease over time as most of the animals that were (or later become) infected or which fail the SICTT are removed from the herd over time. 4.5. Policy implications These results have important implications for farmers, national policy makers and bTB testers. These and earlier results provide strong evidence of M. bovis infection in some SICTT inconclusive reactor animals, in support of imperfect test sensitivity and residual M. bovis infection in individual animals in the Irish cattle population. In addition Coad et al. (2010) has found that successive short-interval tests can lead to progressive desensitisation giving rise to the possibility of a false negative result. We acknowledge that the SICTT has a much higher sensitivity when used as a herd-level test. Although such conclusions

153

are not new, they do provide support for an evaluation of relevant aspects of current policy and management, particularly with respect to TIRs in the context of both herd and animal management. To minimise future bTB risk, immediate slaughter of all SIRs is preferred. As an alternative, animal-level restriction, such that TIR animals remain on the farm of initial disclosure, apart from direct consignment to slaughter, may be justified. In all situations other than immediate slaughter, TIRs should be closely monitored at all future tests. Although the risk of a TIR being diagnosed with bTB decreased after around 500 days, it is important to note that the hazard was always greater than that of nonTIR animals throughout the study period. Private veterinary practitioners (PVPs) and farmers should be made aware of the increased risk associated with TIRs, despite these animals having a negative result at the retest. At future testing events, PVPs should have access to the inconclusive reactor testing history of the animals in order to monitor their future test results. Acknowledgements Advice from Prof. Wayne Martin during the study design is gratefully acknowledged. References Álvarez, J., de Juan, L., Bezos, L., Romero, B., Sáez, J.L., Marqués, S., Domínguez, C., Mínguez, O., Fernández-Mardomingo, B., Mateos, A., Domínguez, L., Aranaz, A., 2009. Effect of paratuberculosis on the diagnosis of bovine tuberculosis in a cattle herd with a mixed infection using interferon-gamma detection assay. Vet. Microbiol. 135, 389–393. Clegg, T.A., More, S.J., Higgins, I.M., Good, M., Blake, M., Williams, D.H., 2008. Potential infection-control benefit for Ireland from premovement testing of cattle for tuberculosis. Prev. Vet. Med. 84, 94–111. Clegg, T.A., Duignan, A., Whelan, C., Gormley, E., Good, M., Clarke, J., Toft, N., More, S.J., 2011. Using latent class analysis to estimate the test characteristics of the interferon-␥ test, the single intradermal comparative tuberculin test and a multiplex immunoassay under Irish conditions. Vet. Microbiol., doi:10.1016/j.vetmic.2011.02.027. Clegg, T.A., Good, M., Duignan, A., Doyle, R., More, S.J. Shorter-term risk of Mycobacterium bovis in Irish cattle following an inconclusive diagnosis to the single intradermal comparative tuberculin test, in preparation. Coad, M., Clifford, D., Rhodes, S.G., Hewinson, R.G., Vordermeier, H.M., Whelan, A.O., 2010. Repeat tuberculin skin testing leads to desensitisation in naturally infected tuberculous cattle which is associated with elevated interleukin-10 and decreased interleukin-1 beta responses. Vet. Res. 41:14, doi:10.1051/vetres/2009062. De la Rua-Domenech, R., Goodchild, A.T., Vordermeier, H.M., Hewinson, R.G., Christiansen, K.H., Clifton-Hadley, R.S., 2006. Ante mortem diagnosis of tuberculosis in cattle: a review of the tuberculin tests, ␥interferon assay and other ancillary diagnostic techniques. Res. Vet. Sci. 81, 190–210. Dohoo, I., Martin, W., Stryhm, H., 2009. Veterinary Epidemiologic Research, 2nd ed. VER Inc., Charlottetown, PEI, Canada. Green, D.M., Kiss, I.Z., Mitchell, A.P., Kao, R.R., 2008. Estimates for local and movement-based transmission of bovine tuberculosis in British cattle. Proc. R. Soc. Lond. B Biol. Sci. 275, 1001–1005. Griffin, J.M., Martin, S.W., Thorburn, M.A., Eves, J.A., Hammond, R.F., 1996. 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. 27, 75–87. Griffin, J.M., Williams, D.H., Kelly, G.E., Clegg, T.A., O’Boyle, I., Collins, J.D., More, S.J., 2005. The impact of badger removal on the control of tuberculosis in cattle herds in Ireland. Prev. Vet. Med. 67, 237–266. Kelly, G.E., More, S.J., 2011. Spatial clustering of TB-infected cattle herds prior to and following proactive badger removal. Epidem. Infect, doi:10.1017/S0950268810002323.

154

T.A. Clegg et al. / Preventive Veterinary Medicine 100 (2011) 147–154

Martin, S.W., O’Keeffe, J.J., White, P.W., Edge, V.L., Collins, J.D., 2002. Factors associated with the risk of, and animal level response to, Mycobacterium bovis in Irish cattle, 1993–98. In: Selected Papers 2000–2001, Veterinary Epidemiology and Tuberculosis Investigation Unit, University College Dublin, Dublin, pp. 1–8. Meadows, C., Rajala-Schultz, P.J., Frazer, G.S., Phillips, G., Meiring, R.W., Hoblet, K.H., 2007. Evaluation of a contract breeding management program in selected Ohio dairy herds with event-time analysis II. Parametric frailty models. Prev. Vet. Med. 80, 89–102. Monaghan, M.L., Doherty, M.L., Collins, J.D., Kazda, J.F., Quinn, P.J., 1994. The tuberculin test. Vet. Microbiol. 40, 111–124. More, S.J., Good, M., 2006. The tuberculosis eradication programme in Ireland: a review of scientific and policy advances since 1988. Vet. Microbiol. 112, 239–251. Olea-Popelka, F.J., White, P.W., Collins, J.D., O’Keeffe, J., Kelton, D.F., Martin, S.W., 2004. Breakdown severity during a bovine tuberculosis episode as a predictor of future herd breakdowns in Ireland. Prev. Vet. Med. 63, 163–172.

Olea-Popelka, F.J., Costello, E., White, P., McGrath, G., Collins, J.D., O’Keeffe, J., Kelton, D.F., Berke, O., More, S.J., Martin, S.W., 2008. Risk factors for disclosure of additional tuberculous cattle in attested-clear herds that had an animal with a confirmed lesion of tuberculosis at slaughter during 2003 in Ireland. Prev. Vet. Med. 85, 81–91. O’Sullivan, M.C., O’Keeffe, J.J., 1998. Bovine tuberculosis: statistical modelling of single animal breakdowns. In: Selected Papers 1997, Tuberculosis Investigation Unit, University College Dublin, Dublin, pp. 43–49. Stewart, G.R., Robertson, B.D., Young, D.B., 2003. Tuberculosis: a problem of persistence. Nat. Rev. Microbiol. 1, 97–105. Wolfe, D.M., Berke, O., More, S.J., Kelton, D.F., White, P.W., O’Keeffe, J., Martin, S.W., 2009. The risk of a positive test for bovine tuberculosis in cattle purchased from herds with and without a recent history of bovine tuberculosis in Ireland. Prev. Vet. Med. 92, 99– 105.