Evaluation of surveillance strategies for bovine tuberculosis (Mycobacterium bovis) using an individual based epidemiological model

https://doi.org/10.1016/j.prevetmed.2004.12.002Get rights and content

Abstract

The Netherlands holds the bovine tuberculosis-free (BTB-free) status according to European Union standards, but in recent years small outbreaks of the infection have occurred. After the last outbreak in 1999 with 10 infected herds the question raised if the current surveillance system, visual inspection of carcasses at the slaughterhouse, is efficient enough to detect infected cattle in time and to maintain the official BTB-free status.

Through epidemiological modelling, the risk of a major outbreak is quantified, using one of six surveillance strategies. These are the currently used visual inspection of carcasses at the slaughterhouse (SL), the ELISA test on blood samples of carcasses at the slaughterhouse (ELISA-B), the γ-interferon test on blood samples of carcasses at the slaughterhouse (GAMMA-B), comparative tuberculination of the herd (CT), the combined method of single and comparative tuberculination of the herd (ST + CT) and the ELISA test on samples of bulk milk (ELISA-M). Test frequency of the last three methods was varied as well.

A stochastic individual based model (IBM) was developed to simulate a chain of infected herds, where each individual animal is followed in time. The model mimics the nation-wide situation after the introduction of one infected animal into one herd. BTB-transmission is simulated with an S-E1-E2-I state transition model. Output is time until detection of the infection, prevalence in the detected herd and the number of infected herds at the time of detection. For the assessment 500 simulations were used, representing 500 BTB-introductions. Model robustness to parameter values was analysed with Monte Carlo elasticity analysis, for which 1000 simulations were used.

Results of median time until detection and median number of infected farms at detection for SL (302 weeks and seven farms) were in agreement with estimates from an outbreak in the Netherlands in 1999. ELISA-B and GAMMA-B performed better than SL with a much lower median time until detection (189 and 97 weeks, respectively). The results for the tuberculination methods (ST + CT and CT) and ELISA-M depended heavily on the frequency in which the tests were performed. The tuberculination methods ST + CT and CT yield comparable results and detect the infection sooner than SL, also at the lowest tested frequency of once in 5 years. ELISA-M is comparable with SL at frequencies of once in 4 or 5 years, and this test works well at frequencies of once a year or higher. Our study results are used for an economical optimisation analysis of the six surveillance strategies.

Introduction

Due to effective eradication programs that started at the beginning of the 20th century, the Netherlands is practically bovine tuberculosis-free (BTB-free) since the mid 1950s. The European Union granted the Netherlands the official BTB-free status in 1992 (Emmerzaal et al., 1999). An advantage of this status is that the expensive yearly tuberculination of bovine herds is no longer obligatory. Visual inspection of the carcasses at the slaughterhouse and an efficient back tracing system remain obligatory, next to tuberculination of imported animals. The BTB-free status is discontinued if more than 0.1% of the bovine herds is infected in a calendar year, i.e. about 60 herds in the Netherlands in 2002 (Emmerzaal et al., 1999, Veling et al., 1993).

In the 1990s, nine outbreaks have occurred in Dutch bovine herds. The primary infected herds had a high prevalence (tuberculination test prevalence up to 80%) at the moment of detection, and the last outbreak in 1999 resulted in 10 infected herds. Thus, several infected animals must have passed the visual inspections without being detected (Emmerzaal et al., 1999). This raised the question whether the current surveillance method of visual inspection of carcasses is efficient enough to detect an outbreak before the number of infected herds increases above a certain level, being the danger zone of loosing the official BTB-free status.

In this study, we describe the epidemiological evaluation of six surveillance strategies. These are the current visual inspection of carcasses at the slaughterhouse (SL), the ELISA (enzyme-linked immunosorbent assay) test on blood samples of carcasses at the slaughterhouse (ELISA-B), the γ-interferon test on blood samples of carcasses at the slaughterhouse (GAMMA-B), comparative tuberculination of all animals of the herd (CT), single tuberculination of all animals with a confirmation test of positives with comparative tuberculination (ST + CT) and the ELISA test on samples of bulk milk (ELISA-M).

Because of the low specificity of single tuberculination, it is only used in combination with comparative tuberculination in many EU-countries. Positive single tuberculinated animals are re-tested with comparative tuberculination and found positive if both tests are positive. Comparative tuberculination (CT) without prior testing by single tuberculination is used in Ireland. The ELISA tests on samples of blood or bulk milk is a test currently in development. The objective of the bulk milk test is to develop a method, where samples of the milk of a herd can be used to determine whether one or more animals are infected, without testing individual animals in the herd.

In this paper, we evaluate which of the above-mentioned surveillance strategy is adequate to maintain the official BTB-free status. This is done by simulation of the BTB and cattle dynamics, as it currently exists in the Netherlands. We compare the surveillance strategy by the time until detection of the infection and by number of infected herds at the moment of detection for each surveillance strategy.

The most effective method in terms of short detection time and low number of infected farms does not have to be the cheapest method. In another paper, we describe the economical analysis of the results of this study (Van Asseldonk et al., in press).

Section snippets

Model description

We evaluated the six surveillance strategies by simulating BTB-outbreaks in a population of cattle herds. Unfortunately, available models of BTB spread are developed for countries with high incidence of the disease (e.g. the infection is endemic in Ireland, England and New Zealand) (Barlow, 1994, Barlow et al., 1997, Griffin and Williams, 1999). In these countries wildlife reservoirs for BTB (e.g. badgers in Ireland and England and possums in New Zealand) are subject of debate. In the

Results

After introduction of the infection in the first herd two possible scenarios exist. The first possibility is that the infection spreads within the herd and establishes itself. In this case, the infection can spread to other herds. The other possibility is that the infection does not establish itself in the herd (minor outbreak). Examples of these two scenarios are given in Fig. 1. In our simulations 6.8% (34) of the introductions ended in a minor outbreak. In a number of cases some of the

Discussion

The results of the simulated outbreaks for visual inspection of carcasses in the slaughterhouse were validated with detailed data of the 1999-outbreak in the Netherlands by comparison of time until detection, prevalence and number of infected farms.

For the 1999-outbreak, the time until detection was estimated between 220 and 430 weeks (Paaijmans, 2002, Van Roermund et al., 2003). The median of the simulations (302 weeks) is well within this range. According to the model, variation is large and

Acknowledgements

We thank Aline de Koeijer for contributions during the development of the model. Maarten de Gee is thankfully noticed for his useful comments on the model and thereby reducing stochasticity of the model. We thank Egbert van Nes for answering questions on the Monte Carlo elasticity analysis. Furthermore, we are grateful to Marleen Paaijmans, Annet Velthuis, Douwe Bakker and Fred van Zijderveld for their advise during parameter estimation.

References (30)

  • E.H. Van Nes et al.

    Charisma: a spatial explicit simulation model of submerged marcophytes

    Ecol. Model.

    (2003)
  • E.H. Van Nes et al.

    Alternative attractors may boost uncertainty and sensitivity in ecological models

    Ecol. Model.

    (2003)
  • P.R. Wood et al.

    In vitro immunodiagnostic assays for bovine tuberculosis

    Vet. Microbiol.

    (1994)
  • L. Corner et al.

    Efficiency of inspection procedures for the detection of tuberculous lesions in cattle

    Aust. Vet. J.

    (1990)
  • A. Emmerzaal et al.

    Het belang van een goede diagnostiek voor rundertuberculose

    Tijdschr. Diergeneesk

    (1999)
  • Cited by (59)

    • Assessing the variability in transmission of bovine tuberculosis within Spanish cattle herds

      2018, Epidemics
      Citation Excerpt :

      In fact, data obtained under experimental conditions (Neill et al., 1988, 1989; Costello et al., 1998; Dean et al., 2005) may not be representative of the infection dynamics under natural field conditions. Some authors have based their parameter estimations on data obtained from field studies, but with a low number of observations (Fischer et al., 2005; Pérez et al., 2002; Barlow et al., 1997), which may not reflect the whole complexity and variability of bTB spread among different farms. On the other hand, when local (Bekara et al., 2014; Álvarez et al., 2012a) or national-based data sets are used (O’Hare et al., 2014; Conlan et al., 2012; Kao et al., 1997), they are unlikely to contain the level of detail needed for the accurate estimation of transmission parameters.

    View all citing articles on Scopus
    View full text