Copyright is owned by the Author of the thesis. Permission is given for a copy to be downloaded by an individual for the purpose of research and private study only. The thesis may not be reproduced elsewhere without the permission of the Author. MODEL APPLICATIONS OF DECISION SUPPORT SYSTEMS IN MEAT HYGIENE PROGRAMS A THESIS PRESENTED IN PARTIAL FULFIIMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF VETERINARY SCIENCE AT MASSEY UNIVERSITY PETRUS BERNARDUS VAN DER LOGT December, 1996 ABSTRACT Current systems to ensure safety of meat are to a large degree based on a "procedural" approach which specifies what inspection actions will be taken to protect human health. Both knowledge and disease priorities have changed substantially over recent decades, and moreover the scale of the problems created by any breakdowns in protection has escalated greatly, as food trading and consumption patterns have changed. It is now recognized that meat hygiene needs to focus primarily on ways by which the risk that product will represent a hazard to human health can be effectively reduced, rather than merely ensuring compliance with a defined set of procedures. In addition to human food safety, meat inspection has the potential to contribute information to improve animal health on a national and a local scale. This thesis examines example issues in order to identify possible approaches to the development of decision support systems which assist in protecting meat consumers and improving the health of livestock on farms. The main areas which were explored for this purpose were respiratory disease in lambs and chemical residues in slaughter animals. A literature review of pneumonia and pleurisy in lambs showed that numerous factors have been proposed as predisposing causes for these diseases, but there was surprisingly little valid experimental or observational research evidence to support such statements. A hazard analysis was performed for the micro-organisms which have been isolated from pneumonic lungs. The major commonly detected organisms did not appear to cause a risk to healthy people. However there were a number of micro-organisms which are isolated on occasion from pneumonic and sometimes from healthy sheep lungs that might cause human disease. A case-control study was carried out as an exploratory means to identify risk factors and to generate hypotheses about causal processes. A number of risk factors were initially identified at univariate level. At the second stage the importance of some of these risk factors was quantified in a logistic regression model. Finally a third stage analysis showed the interactions between the factors in a logistic path model, which consisted of three clusters. One cluster included characteristics of the farm and paddocks, one cluster included the yards and practices in the yards, and a third cluster included the types and number of animals on the farm. Two intervention studies were subsequently carried out to evaluate the effect of making various management modifications on the prevalence of pneumonia and pleurisy at slaughter. One intervention study evaluated the time lambs spent in the yards after weaning and the use or oral or injectable drenches. The second intervention study evaluated the use of oral versus injectable drenches and the use of a shower dip versus a wand. The intervention studies showed an effect of time in the yards on pneumonia. There was some association between time in the yards and acute localised pleurisy but none between the other measures tested and respiratory disease. The studies showed clear temporal patterns with regard to pleurisy and pneumonia and enabled comparisons to be made between farms. A study of inspection for pleurisy at slaughterhouses was analysed. The analysis identified the temporal patterns of certain types of pleurisy. Comparisons were made between four participating premises. The sensitivity and specificity of meat inspection for the various types of pleurisy was analysed. The pleurisy data over an eleven year period of the entire country were analysed. Differences were shown between islands and regions. The potential for development of components of a decision support system for pneumonia and pleurisy was illustrated with a number of examples. An important component was to determine how farmers could be assisted in improving the health of their lambs with regard to pleurisy. Ideas to improve farmer involvement were developed. The principles of a decision support system which evaluated the issue of cross-contamination due to handling of product by the inspector were developed. Epidemiological principles of chemical residues in slaughter animals were investigated. A number of statistical quality control tests were applied to known data sets to evaluate what sample sizes would be required to detect changing trends or spatial patems. Temporal simulations were performed to determine how well clusters in time could be detected. The Moving Average approach was used and it appeared that with the given data set sample sizes well beyond those feasible to achieve would be required. Spatial analyses with a number of different statistics were performed. In this case also, large sample sizes were required for reliable results. It was concluded that use of a risk analysis model to define a risk-reduction strategy targeted to avoid any significant risk to the consumer offered a much more effective tool than a fixed sampling system. This model combines a range of possible risk reduction measures in various mixes, and determines whether or not each of the tested strategies achieves the goal of making it very improbable that a consumer would be exposed to sufficient levels of chemical residues in food to even constitute some minimal public health risk. 11 ACKNOWLEDGEMENTS There have been many people contributing in a variety of ways to this thesis to whom I am very grateful. I would like to thank all of them for enabling me to carry out these studies and in particular: Professor Roger Morris, my chief supervisor, without whose input this work would not have been possible. His foresight was indispensable. Our discussions were very stimulating and lead me to many new insights in epidemiology. Steve Hathaway and Per Madie, my other supervisors, each with their own unique area of expertise, contributed invaluable input into this thesis. Steve Hathaway is thanked for the development and the supervision of an inspection trial regarding pleurisy in lambs. A number of statistical issues were encountered during various projects and Dirk Pfeiffer would always be prepared to advise constructively in resolving them. Discussions with Dave West were a great help in getting a fuller understanding of sheep diseases and farm management. Brent Patterson, Warwick Deighton and staff on the properties who carried out the intervention studies and voluntarily put in extra time and effort. The meat inspectors and veterinarians at the AFFCO Rangiuru plant for inspecting and recording the diseases of the lambs during the intervention studies. AFFCO New Zealand for their cooperation with the case-control study, and the sheep farmers in the North Island who willingly completed and returned the questionnaire. Mallinckrodt Veterinary Limited for donating the animal remedies which were used in the intervention studies. The New Zealand Ministry of Agriculture for sponsoring me while I undertook this study. Finally, I would like to thank my wife Rae for her support during this study. 111 Abstract Acknowledgements Table of Contents List of Figures List of Tables TABLE OF CONTENTS Chapter 1: GENERAL INTRODUCTION Aim of Food Safety Programmes Conventional Meat Inspection Systems Process Control Risk Analysis Animal Health Surveillance Data Collection, Analysis and Feedback Future Developments Projects in the Thesis Chapter 2: PNEUMONIA AND PLEURISY IN LAMBS Literature Review of Pneumonia and Pleurisy in Sheep Introduction Clinical signs Page 11 lV XU xiv 1 1 1 2 3 3 3 4 4 6 6 6 6 lV Pathology Micro-organisms Micro-organisms in the upper respiratory tract Risk factors Vaccination and treatment Post mortem inspection and processing Hazard Analysis of Pneumonia and Pleurisy in Lambs with Regard to Public Health Introduction Materials and methods Classification as a food-borne agent Interpretation of the literature Results Discussion Case-control Study of Pleurisy in Lambs Introduction Materials and methods Collection of data Statistical analysis Results Discussion Intervention Study of Pleurisy and Pneumonia At Farm A Introduction Materials and methods 7 8 12 13 14 15 16 16 16 16 18 19 27 30 30 30 30 31 31 38 47 47 47 V Weaning and first drench Second drench Third drench or slaughter Fourth drench Second draft for slaughter Inspection and procedures for diseased lambs Analytical methods Results Information collected at the slaughterhouse Chi-squared tests to evaluate effect of treatments Log-linear modelling to evaluate interactions between treatments and time of slaughter Relationships Relationship with Dictyocaulus viviparus Discussion Intervention Study of Pleurisy and Pneumonia At Farm B Introduction Materials and methods Treatments before trials started Fourth drench Fly strike treatment and fifth drench Subsequent drenches Inspection and procedures for diseased lambs 48 48 49 49 49 49 50 50 50 51 52 54 55 56 58 58 58 58 59 59 59 59 Vl Analytical methods Results Observations of diseased stock Information collected at the slaughterhouse Effects of treatment Log-linear modelling to evaluate interactions between treatments and time of slaughter Temporal analysis of pleurisy and pneumonia Relationships Comparison between Farm A and Farm B Discussion Analysis of Lamb Pleurisy Inspection Trials Introduction Materials and methods Results of analysis of prevalence Discussion of analysis of prevalence Materials and methods for sensitivity and specificity Results of sensitivity and specificity analysis Discussion of sensitivity and specificity analysis Analysis of Pleurisy Data of the MAF Disease and Defect Database Introduction Materials and methods Results Comparison between islands 60 60 60 60 61 63 63 64 65 65 67 67 67 68 70 74 75 76 78 78 78 80 80 vu Comparison between regions 80 Comparison between premises and their regions 81 Predictions of the prevalence of pleurisy based on values 81 of other premises Effect of month and year 82 Discussion 84 Decision Support Systems to Evaluate Pleurisy 86 in Slaughter Lambs Introduction 86 DSS for pleurisy of any degree of severity 87 Comparison of point prevalence and cumulative prevalence 88 Comparison of cumulative prevalence including consideration 90 of stock slaughter pattern Performance of a farm over time 92 DSS to evaluate the probability of developing severe problems 94 The development of a feedback system 94 DSS for food safety 95 Discussion 98 CHAPTER 3: CHEMICAL RESIDUES IN SLAUGHTER ANIMALS Introduction 100 100 100 101 101 Objectives and priorities of sampling for residues Non-compliances Maximum Residue Levels (MRLs) viii Reaction to non-complying levels Categories of chemicals that leave residues Animal remedies Environmental contaminants Chemicals that occur naturally in the environment Chemical compounds discussed in the thesis Analytical Methods Baselines Livestock classes Random sampling Temporal analysis Plotting ANO VA and chi-squared tests Quality control charts and statistics Sequential sampling Regression techniques Spatial Analysis Display of geographical patterns ANOVA and chi-squared tests Spatial statistics Time-space analysis Evaluation of Current Sampling Plans Introduction 102 102 102 103 103 103 104 104 105 107 108 109 109 109 110 111 111 112 112 112 115 116 116 ix Methods and materials of sampling plans 117 Comparison between levels 117 Random and stratified sampling 117 Sequential sampling 118 Results of sampling plans 118 Sample sizes for normal distribution with copper 118 Sample sizes for binomial distribution with 120 ivermectin/milbemycin Sequential sampling with copper and ivermectin/milbemycin 120 Discussion of sampling plans 120 Materials and methods of temporal simulations 124 Results of temporal simulations 125 Discussion of temporal simulations 128 Materials and methods of spatial simulations 129 Results of spatial simulations 132 Simulation of a normal situation 133 Simulation of clustering at two locations at a low level (L2) 135 Simulation of clustering at two locations at a high level (H2) 137 Simulation of clustering at six locations at a low level (L6) 139 Simulation of clustering at six locations at a high level (H6) 142 Simulation of clustering at 16 locations at a low level (LI 6) 144 Simulation of clustering at 16 locations at a high level (H 16) 147 Discussion of spatial simulations 149 Risk-based Control System of Chemical Residues 151 X Introduction Model Stage 1 Good handling practice Comments on stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Targeted high risk animals Targeted high riskfarms Reduction by targeted sampling Farm certification programme Discussion CHAPTER 4: GENERAL DISCUSSION APPENDICES REFERENCES 151 152 152 153 155 155 156 157 157 167 170 204 XI LIST OF FIGURES Page Figure 2.1 Null hypothesis of cluster "Yards" 41 Figure 2.2 Final cluster "Yards" 42 Figure 2.3 Null hypothesis of cluster "Farm" 43 Figure 2.4 Final cluster "Farm" 44 Figure 2.5 Null hypothesis of cluster "Livestock" 45 Figure 2.6 Final cluster "Livestock" 46 Figure 2.7 Percentage prevalence of Major pleurisy at four slaughterhouses 71 Figure 2.8 Percentage prevalence of Minor pleurisy at four slaughterhouses 71 Figure 2.9 The Ration of Major pleurisy/Minor pleurisy at four slaughterhouses 72 Figure 2.10 Percentage prevalence of Acute pleurisy at four slaughterhouses 72 Figure 2.11 Percentage prevalence of Septicaemia at four slaughterhouses 73 Figure 2.12 Example of a component of a DSS to evaluate the existence 90 of pleurisy problems regardless of slaughter pattern Figure 2.13 Example of a component of a DSS to evaluate the existence 91 of pleurisy problems with consideration of the slaughter pattern Figure 2.14 Graphical assistance for farmer to evaluate his performance 92 with regard to pleurisy and slaughter of his lambs Figure 2.15 Component of a DSS to evaluate the probability of a farm 94 developing problems Figure 3.1 Illustration of the concept of providing assurances over a 119 moving time frame Figure 3.2 Sequential sampling plan for copper in sheep 123 xii Figure 3.3 Sequential sampling plan for ivermectin/milbemycin 123 Figure 3.4 Explanation of difference of non-complying days and runs 125 Figure 3.5 Histogram distribution of cadmium 129 Figure 3.6 Histogram distribution if original cadmium values 130 are multiplied by 1.10 Figure 3.7 Histogram distribution if original cadmium values 130 are multiplied by 1.25 Figure 3.8 Screen - Risk reduction strategy 160 Figure 3.9 Screen - Good handling practice (1) 160 Figure 3.10 Screen - Good handling practice (2) 161 Figure 3.11 Screen - Good handling practice (3) 161 Figure 3.12 Screen - Targeted high risk animal programme (1) 162 Figure 3.13 Screen - Targeted high risk animal programme (2) 162 Figure 3.14 Screen - Targeted high risk farms (1) 163 Figure 3.15 Screen - Targeted high risk farms (2) 163 Figure 3.16 Screen - Targeted sampling (1) 164 Figure 3.17 Screen - Targeted sampling (2) 164 Figure 3.18 Screen - Targeted sampling (3) 165 Figure 3.19 Screen - Farm certification programme (1) 165 Figure 3.20 Screen - Farm certification programme (2) 166 Xlll LIST OF TABLES Page Table 2.1 Micro-organisms reported to have been isolated from the ovine lower 10 respiratory tract in New Zealand and overseas with their references Table 2.2 Criteria for including micro-organisms in or excluding from food 18 safety programmes Table 2.3 Codes and descriptions used in the multivariate analysis 33 Table 2.4 Chi-squared values, p-values, and numbers of missing values of 34 statistically significant variables at univariate level Table 2.5 Beta coefficients, Standard errors (SE), Odds ratios (OR) and the 90% 35 confidence intervals of variables included in the logistic regression model at the second stage Table 2.6 Associations, Beta coefficients, Standard errors (SE), Odds ratios (OR) 36 and 90% confidence limits of significant paths in the variable cluster "Yards" Table 2.7 Associations, Beta coefficients, Standard errors (SE), Odds ratios (OR) 37 and 90% confidence limits of significant paths in the variable cluster "Farms" Table 2.8 Associations, Beta coefficients, Standard errors (SE), Odds ratios (OR) 38 and 90% confidence limits of significant paths in the variable cluster "Livestock" Table 2.9 Number of lambs that were slaughtered and respiratory 51 pathology that was recorded at Farm A Table 2.10 Chi-squared tests comparing the effects of treatments by using 53 cumulative prevalences ofrespiratory pathology on 1/3/95 Table 2.11 Log-linear model which evaluates the interactions between treatment 54 and the time of slaughter for Farm A Table 2.12 Combinations of pathology of lungs and pleura at Farm A 55 Table 2.13 Relationship of pleurisy and pneumonia with Dictyocaulus viviparus 56 XIV Table 2.14 Nwnber of lambs that were slaughtered and respiratory pathology 62 that was recorded at Fann B Table 2.15 Log-linear model which evaluates the interactions between treatment 63 and the time of slaughter at Fann B Table 2.16 Combinations of pathology of lungs and pleura at Farm B 64 Table 2.17 Comparison of pathology between two farms early in March 1995 65 Table 2.18 Prevalence of pleural lesions and the ratio of Minor pleurisy / 69 Major pleurisy Table 2.19 Codes used for determining meat inspection performance characteristics 74 Table 2.20 Meat inspection performance characteristics of pleurisy 75 Table 2.21 Meat inspection performance characteristics of Minor pleurisy 75 Table 2.22 Meat inspection performance characteristics of Major pleurisy 76 Table 2.23 Meat inspection performance characteristics of Acute pleurisy 76 Table 2.24 Premises of which pleurisy data were compared 79 Table 2.25 Nwnber of slaughtered lambs and percentage prevalence by island 80 Table 2.26 Nwnber of slaughtered lambs and percentage prevalence by region 81 Table 2.27 Percentage prevalence by slaughterhouse and neighbouring area 82 Table 2.28 P-values and R-squared values oflinear regressions to evaluate 83 the month and year factor Table 2.29 Kruskal-Wallis one-way non-parametric ANOVA to analyse 83 year effects Table 2.30 Fictitious data of pleurisy and slaughtered lambs of Farm A and Area X 89 Table 2.31 Evaluation of fann performance over time 93 Table 3.1 Required sample size for testing copper in sheep based on 119 historical data Table 3.2 Required sample size for testing ivermectin/milbemycin in bulls 120 based on historical data xv Table 3.3 Expected sample size for sequential sampling of copper 121 in sheep Table 3.4 Expected sample size for sequential sampling of 121 ivermectin/milbemycin in bulls Table 3.5 Upper Control Limit used for temporal simulation of 124 levamisole non-compliances in lambs Table 3.6 Number of violative days per time period 126 Table 3.7 Number of violative runs per time period 127 Table 3.8 Frequency distribution of cadmium values in hoggets for 129 a normal situation Table 3.9 Frequency distribution of cadmium values which were 130 1.1 * normal situation Table 3.10 Frequency distribution of cadmium values which were 131 1.25 * normal situation Table 3.11 I and c, p-values for simulation of normal situation 133 Table 3.12 G ( d), p-values for simulation of normal situation 133 Table 3.13 G;( d), clustered locations for normal situation 134 Table 3.14 I and c, p-values for simulated clustering at L2 level 135 Table 3.15 G ( d), p-values for simulated clustering at L2 level 135 Table 3.16 G;( d), clustered locations for simulation at L2 level 136 Table 3.17 I and c, p- values for simulated clustering at H2 level 137 Table 3.18 G ( d), p-values for simulated clustering at H2 level 138 Table 3.19 G;( d), p-values for simulated clustering at H2 level 138 Table 3.20 I and c, p- values for simulated clustering at L6 level 140 Table 3.21 G ( d), p-values for simulated clustering at L6 level 140 Table 3.22 G;( d), clustered locations for simulation at L6 level 141 XVI Table 3.23 I and c, p-values for for simulated clustering at H6 142 Table 3.24 G (d), p-values for simulated clustering at H6 level 143 Table 2.25 GJ d), clustered locations for simulation at H6 level 143 Table 3.26 I and c, p- values for simulated clustering at L 16 level 145 Table 3.27 G (d), p-values for simulated clustering at L16 level 145 Table 3.28 GJ d), clustered locations for simulation at L 16 level 146 Table 3.29 I and c, p- values for simulated clustering at Hl 6 level 147 Table 3.30 G (d), p-values for simulated clustering at H16 level 148 Table 3.31 GJd), clustered locations for simualtion at H16 level 148 Table 3.32 Fictitious weighted risk factors 154 XVll CHAPTER! INTRODUCTION Aim of Food Safety Programs The aim of food safety programmes is to provide food which has a minimal risk of producing diseases or other adverse effects on the consumer, at a reasonable price. This definition conveys the concept that the safety of food should be seen in the context of product price, ie additional safety can be bought at the cost of an increased price for the product. There should be an appreciation that there is a balance between the desire to eat various foods and the risk that is inherently taken in the process of eating each type of food. The definition is intended to express the idea that few human activities are risk-free, and that actions taken to reduce or eliminate risks must balance the additional benefit from safety programmes against the additional costs. The critical components which should form the basis for designing meat safety programmes are explained below. They consist of process control, data collection, analysis/feedback, and risk assessment. In contrast, conventional food safety programmes rely on inspection of the carcasses of individual animals and are commonly called meat inspection systems. Conventional Meat Inspection Systems As an established concept 'meat inspection' has been central to efforts aimed at guaranteeing protection of the consumer. This term may have become counterproductive, because it emphasises what should be a component only of a more comprehensive food safety programme which includes process control and protection against chemical residues. Inspection of animals in slaughterhouses is intended to protect the public health, but current inspection procedures support past disease priorities and outdated epidemiological understanding. They do not accurately reflect current concepts of product safety, and how it should be achieved. In New Zealand the vast majority of sheep, cattle, pigs, deer, goats and horses that are slaughtered for human consumption are subjected to an ante mortem and a post mortem inspection which are clinical and pathological evaluations. Usually both evaluations are carried out rapidly. However in case of abnormalities there is scope to detain live animals or carcasses and tissues for a thorough inspection, with laboratory backup if necessary. Those animals or parts of animals which are deemed unfit for human consumption will be condemned. Although the 1 intention is to remove product from the food chain which presents a public health hazard, in reality only tissues which display abnormalities are condemned. Many conditions of public health significance cannot be detected by current procedures. Only some of the animal diseases that are used to determine carcass disposition at meat inspection have public health significance. Generally meat inspection systems in the Western world are 'procedure' driven. A product is inspected at the end of the production process. The systems are very occupied with compliance with their rules rather than with their more fundamental aims. These meat inspection systems are frequently based on the systems that were developed late last century. Many of the conditions that were considered to be significant at the time are no longer considered to be important. This is to some degree because the prevalence of a number of diseases in livestock has decreased. However there is an increasing awareness that some of the conditions which cannot be detected by conventional meat inspection are of great human health importance. These conditions include the presence of pathogenic micro-organisms and residues of chemicals. There are other important functions which meat inspection already performs to some degree and which can be strengthened in new food safety programmes. They include surveillance of animal health and production, and defects of processing and marketing importance. These components are not concerned with food safety. Where these issues are addressed successfully farmers and meat processors will benefit from them. The ability to carry out post mortem examinations of large numbers of animals is a strong point of meat inspection. The role of meat inspection regarding animal health surveillance is explored below. This type of surveillance should not be considered as a stand-alone system. There are inherent weaknesses in the collection of the data, and it is biased in various respects. However an appreciation of the scope of these deficiencies will facilitate the incorporation of slaughterhouse data with other systems, ultimately leading to an overall animal disease surveillance system that will give a sufficiently accurate 'picture' of the situation. The removal of defects such as pleurisy, arthritic joints, and abnormally pigmented meat from carcasses is a cost to meat processing. Meat inspection systems can assist in quantifying the cost oflabour and discarded product thereby providing valuable feedback to both farmers and meat processors. Such feedback is essential where diseases can be prevented on the farm, while the only remedy available to meat processors is trimming. Process Control The main tasks facing food safety programmes in relation to infectious diseases are to exclude pathogens from the food chain to a reasonable degree and to limit the opportunities for these pathogens to multiply during processing and storage. The phrases that have been coined to describe these concepts are 'pre-harvest food safety' and HACCP (Hazard Analysis Critical Control Point). In the case of pre-harvest food safety, risk factors which contribute to the existence of pathogens in farm animals are identified. Subsequently farmers can be encouraged to raise livestock in such a manner that the prevalence of pathogens in the livestock population is reduced. The HACCP principles can be applied to both farming and the meat processing industry. In the meat processing industry it focuses on the areas where processing can go wrong, 2 resulting in contamination of product and the multiplication of pathogens. Examples of such critical points are contact between skin, ingesta or faeces and the carcass, contact between meat and food handlers, and temperature abuse. Risk Analysis There is a need to develop risk analysis tools which will be able to assess the risk which pathogens, procedures and their interaction, pose to public health. This may be performed in a qualitative or a quantitative sense. There is a need to generate a greater appreciation among the general public and decision makers that with the present production and processing methods a nil-risk policy is not feasible. In fact it has never existed. The perception in the past that meat was "safe" if inspected and handled properly, was based on incomplete knowledge. New understanding which has become available over recent decades demonstrates the limitation of the inspection approach. Risk analysis will need to be tied in with economic models. If unequivocally "safe" meat cannot be produced, then questions need to be asked about what are the economic options for producing meat of different degrees of "safety", to satisfy different market needs. Animal Health Surveillance Over the years the importance of disease surveillance has shifted from easily identifiable animal diseases to endemic diseases which are not clearly defined and which are strongly multifactorial. Monitoring for exotic diseases such as foot and mouth disease (FMD) through food safety programmes that are based at slaughterhouses is an important part of the overall national FMD surveillance programme. Most livestock will finally be slaughtered in a slaughterhouse. This provides an opportunity to assess the health of the national flock and individual farm flocks. The data for certain diseases can be acquired more easily and cheaply in a slaughterhouse than anywhere else. These data can then be used to assess the health status of the national herd and to monitor the effect of control or eradication campaigns. Especially in the case of endemic diseases such as pneumonia in sheep, and Johne's disease, slaughterhouse data will be invaluable to individual farmers. Data Collection, Analysis and Feedback The objectives of data collection have to be clearly defined before the actual data collection is started. It would need to be considered whether food safety, animal health or product quality 3 issues are considered. The success of process control, risk analysis and animal health surveillance depends heavily on the availability of data. These data need to be comprehensive and of good quality. Systems need to be in place to collect relevant data and to retrieve them on-line. In addition there is a need to analyse the data in a consistent, statistically sound manner. Computer hardware has reached the stage where it is able to perform the above functions in such a manner that data can be used for practical purposes. There is a need now to develop the appropriate software. Major problems still exist such as identifying geographical areas where animals have been raised before they were slaughtered. This problem may be of more importance in the cattle and sheep industry than in the pig industry. Future Developments There is a growing appreciation in the meat processing industry that control of the process of food production at all stages is as appropriate for them as it is for other industries. Issues such as human infections with Salmonella ssp., Campylobacter ssp., Listeria monocytogenes, verotoxigenic E. coli and Toxoplasma gondii in food of animal origin continue to make headlines in the media. They will not go away and the food processing industry is acutely aware of this. Bovine spongiform encephalopathy (BSE) has brought this subject even more clearly into focus at the worldwide level. Consumer rights have been clearly defined in law and are likely to be applied to instances of food poisoning more often in the future than is currently happening. In the case of Salmonella in eggs corrective action was taken by the poultry industry overseas. An increased awareness among farmers regarding their responsibility to supply healthy stock may seem desirable. However risk factors are usually poorly defined or not defined at all. Therefore the practices that farmers should comply with to supply animals carrying fewer pathogens are not very obvious. Improved identification systems of stock are currently being considered, especially for cattle. This will enhance systems such as AgriBase which is close to full implementation. In the near future the identification of stock may be less troublesome. The ratification of GATT will have two major implications. Certification based on freedom of certain diseases on defined farms or in certain areas rather than in the whole of a country has become an accepted practice. Disease will be considered in relation to zones of varying size, rather than only in relation to a whole country. This will especially have an effect on the use of geographical information systems and statistical analysis. Any restrictions in trade will need to be scientifically based. This will especially have an effect on the development of risk analysis techniques. Projects in the Thesis This thesis takes two quite different issues as examples to explore how meat safety programs can adapt to the new priorities. First it explores how existing inspection findings could be used to 4 provide information to producers, based on the example of pleurisy in lambs. Second, it examines the issue of chemical residue control, considering how a more effective surveillance system for preventing human health risks from such residues could be developed. 5