RISK ASSESSMENT FOR BOVINE WELFARE AT SLAUGHTER. A COMPARISON BETWEEN A RISK ASSESSMENT BASED ON EMPIRICAL DATA AND A RISK ASSESSMENT BASED ON EXPERT OPINION

Background. Risk assessment is a systematic process that uses modelling to estimate the likelihood of adverse effects occurring from exposure to hazards. While risk assessment is widely used to support decision making in many areas of food and feed safety (e.g. veterinary epidemiology, toxicology, eco-toxicology), it is a relatively new concept for animal welfare. Currently there are no standardised guidelines for animal welfare risk assessment. Furthermore, very little research has been conducted to assess the reliability of the existing methods, which mostly rely on qualitative data and are based on expert opinion. Objectives. The objectives of this research project were to assess the scientific robustness of existing risk assessment methods for animal welfare and identify any potential methodological flaws of these processes. Methods. The currently available methods for animal welfare risk assessment were analysed in detail. Two risk assessments for beef cattle at slaughter (in northern Italy) were performed and compared. One based on empirical data (i.e. collected in slaughterhouses) and one based on expert opinion (gathered via a questionnaire submitted to a group of 11 experts). The two new risk assessments were structured to be as similar as possible to the animal welfare risk assessments under appraisal. A list of 56 hazards potentially relevant to beef cattle at slaughter was produced via a literature review. The relevance of such hazards was assessed by a series of preliminary observations in abattoirs and by asking the 11 experts to assess it. Fourteen hazards were excluded from the subsequent analyses as never being observed during the on-site observations and indicated as not relevant by at least 5 out 11 experts. For the risk assessment based on empirical data, a novel method for performing on-site exposure assessment and likelihood of the adverse effects (by severity levels) was developed. The method was based on a precise definition of the hazards and a differentiation between adverse welfare effects and measurable indicators of adverse effects. The latter were associated to different severity classes defined qualitatively on the basis of the intensity of the behavioural responses of the animals. In the second risk assessment the approach to eliciting expert opinion was different from the existing animal welfare risk assessments (based on consensus opinion) as the experts answered the questionnaire independently. Thorough a series of risk assessment-tailored questions, the experts were asked to assess hazard exposure (for beef cattle at slaughter in northern Italy), characterise the adverse effects resulting from the exposure to the hazards and indicate the related uncertainty. Exposure assessment based on empirical data and on expert opinion was compared for 42 hazards. As 18 hazards were never detected during the on-site observations (or the number of animals exposed was < 5), adverse effect characterisation and final risk were estimated and compared for 24 hazards. Results. The results of exposure assessment based on empirical data and on expert opinion were inconsistent for 24 out of 42 hazards. Consistent results for all possible adverse effects resulting from the exposure to a hazard never occurred. Often the variability between the experts’ responses on exposure assessment and adverse effect characterisation was high. In line with the results of exposure assessment and adverse effect characterisation, the two risk estimates rarely produced comparable results. Discussion. The analysis of the available methods for animal welfare risk assessment performed in this study, the discordance of the results of the two risk assessments and the variability between the experts’ responses highlighted some inherent flaws and requirements of existing risk assessments for animal welfare. A more detailed and measurable description of the hazards should be available. Further, a clear understanding of the animal welfare outcomes and their measurement is paramount. In addition, while performing the on-site observations it was clear that interactions between hazards and different hazards intensities and durations need more consideration. The method developed for performing on-site exposure assessment and estimating the likelihood of the adverse effects proved to be very valuable to solve most of the highlighted limitations of existing animal welfare risk assessments. Conclusions. A unique and useful approach to defining the hazards for animal welfare and to assessing animal welfare in a measurable and quantifiable way was developed. In particular the method for assessing animal welfare was based on a clear differentiation between adverse welfare effects and measurable indicators of poor welfare (classified by severity levels). This approach to hazard description and welfare outcome definition and assessment is recommended for enhancing empirical research on animal welfare especially when there is a lack of empirical data for risk assessment. Furthermore, this method can lead to a standardised and harmonised approach for the evaluation of hazards and adverse effects between experts, leading to more robust risk assessments. This study also proposed an alternative method for eliciting expert opinion based on independent scoring of the risk assessment parameters. The approach showed very useful implications for identifying sources of uncertainties that are normally overlooked in existing risk assessments for animal welfare, such as difficulties in assessing the risk assessment parameters, disagreements between the experts or lack of expert knowledge. Finally this study highlighted that, independently of the data used for the risk assessment (i.e. empirical data or expert opinion), the method for either reviewing the literature or gathering expert opinion should be chosen in light of the best available practices. The process and any decisions taken should be documented to ensure greater transparency and reproducibility.